griefith/test
1
0
Fork 0
Code Issues Pull Requests Actions 1 Packages Projects Releases Wiki Activity
Files
f32d7dd0c8aeee53d5bca4e6ff2f2e96468cd062
test/source/blender/compositor/intern/COM_NodeOperationBuilder.cc

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

818 lines
24 KiB
C++
Raw Normal View History

File headers: SPDX License migration Use a shorter/simpler license convention, stops the header taking so much space. Follow the SPDX license specification: https://spdx.org/licenses - C/C++/objc/objc++ - Python - Shell Scripts - CMake, GNUmakefile While most of the source tree has been included - `./extern/` was left out. - `./intern/cycles` & `./intern/atomic` are also excluded because they use different header conventions. doc/license/SPDX-license-identifiers.txt has been added to list SPDX all used identifiers. See P2788 for the script that automated these edits. Reviewed By: brecht, mont29, sergey Ref D14069
2022-02-11 09:07:11 +11:00
/* SPDX-License-Identifier: GPL-2.0-or-later
* Copyright 2013 Blender Foundation. */
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
Cleanup: remove unused includes in Compositor And move unneeded includes in frequently used headers to source files. Slightly reduces compile time.
2021-10-13 23:00:24 +02:00
#include <set>
Cleanup: Use blender::MultiValueMap. Fixed concern raise on {93e2491ee724}.
2021-03-17 09:14:38 +01:00
#include "BLI_multi_value_map.hh"
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
#include "COM_Converter.h"
#include "COM_Debug.h"
Cleanup: remove unused includes in Compositor And move unneeded includes in frequently used headers to source files. Slightly reduces compile time.
2021-10-13 23:00:24 +02:00
#include "COM_ExecutionGroup.h"
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
#include "COM_PreviewOperation.h"
Cleanup: `make format` after SortedIncludes change
2020-03-19 09:33:03 +01:00
#include "COM_ReadBufferOperation.h"
#include "COM_SetColorOperation.h"
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
#include "COM_SetValueOperation.h"
#include "COM_SetVectorOperation.h"
Fix T40986: crash on using the viewer node inside of group nodes. Viewers were activated both inside the active group as well as the top level tree (the latter being a quick fix for getting a fallback viewer). This caused a race condition on the shared viewer image. Now the active viewer is defined at node conversion time in the converter so that only one can be active at a time without each node having to follow complicated rules for exclusion.
2014-07-08 12:48:41 +02:00
#include "COM_ViewerOperation.h"
Cleanup: `make format` after SortedIncludes change
2020-03-19 09:33:03 +01:00
#include "COM_WriteBufferOperation.h"
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
Compositor: Constant folding Currently there is no clear way to know if an operation is constant (i.e. when all rendered pixels have same values). Operations may need to get constant input values before rendering to determine their resolution or areas of interest. This is the case of scale, rotate and translate operations. Only "set operations" are known as constant but many more are constant when all their inputs are so. Such cases can be optimized by only rendering one pixel. Current solution for tiled implementation is to get first pixel from input. This works for root execution groups, others need previous groups to be rendered. On full frame implementation this is not possible, because buffers are created on rendering to reduce peak memory and there is no per pixel calls. This patch evaluates all operations that are constant into primitive operations (Value/Vector/Color) before determining resolutions. Reviewed By: jbakker Differential Revision: https://developer.blender.org/D11490
2021-07-06 16:16:08 +02:00
#include "COM_ConstantFolder.h"
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
#include "COM_NodeOperationBuilder.h" /* own include */
Cleanup: Add namespace to compositor.
2021-03-23 17:12:27 +01:00
namespace blender::compositor {
Compositor: Fix execution system unset during constant folding
2021-07-06 19:54:13 +02:00
NodeOperationBuilder::NodeOperationBuilder(const CompositorContext *context,
bNodeTree *b_nodetree,
ExecutionSystem *system)
Cleanup: replace members `m_` prefix by `_` suffix in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:04 +02:00
: context_(context), exec_system_(system), current_node_(nullptr), active_viewer_(nullptr)
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
{
Cleanup: replace members `m_` prefix by `_` suffix in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:04 +02:00
graph_.from_bNodeTree(*context, b_nodetree);
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
}
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
void NodeOperationBuilder::convert_to_operations(ExecutionSystem *system)
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
{
/* interface handle for nodes */
NodeConverter converter(this);
Cleanup: trailing space for compositor
2018-06-17 17:05:29 +02:00
Cleanup: replace members `m_` prefix by `_` suffix in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:04 +02:00
for (Node *node : graph_.nodes()) {
current_node_ = node;
Cleanup: trailing space for compositor
2018-06-17 17:05:29 +02:00
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
DebugInfo::node_to_operations(node);
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
node->convert_to_operations(converter, *context_);
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
}
Cleanup: trailing space for compositor
2018-06-17 17:05:29 +02:00
Cleanup: replace members `m_` prefix by `_` suffix in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:04 +02:00
current_node_ = nullptr;
Cleanup: trailing space for compositor
2018-06-17 17:05:29 +02:00
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
/* The input map constructed by nodes maps operation inputs to node inputs.
* Inverting yields a map of node inputs to all connected operation inputs,
* so multiple operations can use the same node input.
*/
Cleanup: Use blender::MultiValueMap. Fixed concern raise on {93e2491ee724}.
2021-03-17 09:14:38 +01:00
blender::MultiValueMap<NodeInput *, NodeOperationInput *> inverse_input_map;
Cleanup: replace members `m_` prefix by `_` suffix in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:04 +02:00
for (Map<NodeOperationInput *, NodeInput *>::MutableItem item : input_map_.items()) {
Cleanup: Use blender::MultiValueMap. Fixed concern raise on {93e2491ee724}.
2021-03-17 09:14:38 +01:00
inverse_input_map.add(item.value, item.key);
Cleanup: style, use braces for compositor
2019-04-23 11:21:22 +10:00
}
Cleanup: trailing space for compositor
2018-06-17 17:05:29 +02:00
Cleanup: replace members `m_` prefix by `_` suffix in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:04 +02:00
for (const NodeGraph::Link &link : graph_.links()) {
Cleanup: use blender::Vector.
2021-03-05 16:28:08 +01:00
NodeOutput *from = link.from;
NodeInput *to = link.to;
Cleanup: trailing space for compositor
2018-06-17 17:05:29 +02:00
Cleanup: replace members `m_` prefix by `_` suffix in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:04 +02:00
NodeOperationOutput *op_from = output_map_.lookup_default(from, nullptr);
Cleanup: replace std::map with blender::Map.
2021-03-16 15:02:37 +01:00
Cleanup: Use blender::MultiValueMap. Fixed concern raise on {93e2491ee724}.
2021-03-17 09:14:38 +01:00
const blender::Span<NodeOperationInput *> op_to_list = inverse_input_map.lookup(to);
if (!op_from || op_to_list.is_empty()) {
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
/* XXX allow this? error/debug message? */
Cleanup: comments (long lines) in compositor
2019-05-01 10:50:02 +10:00
// BLI_assert(false);
Cleanup: consistent use of tags: NOTE/TODO/FIXME/XXX Also use doxy style function reference `#` prefix chars when referencing identifiers.
2021-07-03 23:08:40 +10:00
/* XXX NOTE: this can happen with certain nodes (e.g. OutputFile)
Fix T39799: Backdrop (compositor) ignores alpha. This issue is because of a somewhat "special" behavior in old code, which got lost during rB09874df: There was a variant of the `relinkConnections` function which would leave the socket completely unconnected. This is not a valid state really (given that each unconnected input must otherwise connected to a constant `Set` type node), but was used as a way to distinguish connected alpha/depth sockets in composite and viewer output nodes. https://developer.blender.org/diffusion/B/browse/master/source/blender/compositor/intern/COM_InputSocket.cpp;28a829893c702918afc5ac1945a06eaefa611594$69 After the large cleanup patch ({D309}) every socket is now automatically connected to a constant, such that `getInputSocketReader` will never return a NULL pointer. This breaks the previous test method, which needs to be replaced by more explicit flags. Luckily this was done only for very few output nodes (Composite, Viewer, Output-File). These now use the regular SetValueOperation default in case "use alpha" is disabled, but set this to an explicit 1.0 value instead of mapping to the node socket.
2014-04-25 12:00:35 +02:00
* which only generate operations in certain circumstances (rendering)
* just let this pass silently for now ...
*/
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
continue;
}
Cleanup: trailing space for compositor
2018-06-17 17:05:29 +02:00
Cleanup: Use blender::MultiValueMap. Fixed concern raise on {93e2491ee724}.
2021-03-17 09:14:38 +01:00
for (NodeOperationInput *op_to : op_to_list) {
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
add_link(op_from, op_to);
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
}
}
Cleanup: trailing space for compositor
2018-06-17 17:05:29 +02:00
Fix T44149: Compositing : Node Groups do not work correctly Input constants are to be connected before removing proxies, otherwise node groups might give totally different result. This is a regression and to be put into final release.
2015-03-27 17:25:12 +05:00
add_operation_input_constants();
Cleanup: trailing space for compositor
2018-06-17 17:05:29 +02:00
Compositor: File output node sockets When using RNA to alter the type of socket only the type was changed. the typeinfo was not updated. Internally the File Output Node used RNA to update the sockettype. making the socket invalid. When users save the file and reopened the typeinfo was used. Also the color of the node was determined via the typeinfo. Another thing that happened was that the socket conversion was ignored when empty node groups were present. The empty node groups were optimized away before the needed data conversion was determined. Reviewed By: brecht Differential Revision: https://developer.blender.org/D4938
2019-05-24 14:08:21 +02:00
resolve_proxies();
Fix T67358: Compositor Disabled Node Config -> SegFault The issue was caused by wrong conversion happening after some of the proxies are removed. Easiest solution is to first remove proxies and then add required converter operations. Thanks Jeroen for the review!
2020-06-19 14:49:36 +02:00
add_datatype_conversions();
Cleanup: replace members `m_` prefix by `_` suffix in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:04 +02:00
if (context_->get_execution_model() == eExecutionModel::FullFrame) {
Compositor: Merge equal operations Some operations can take a lot of time to execute and any duplication should be avoided. This patch implements a compile step that detects operations with the same type, inputs and parameters that produce the same result and merge them. Now operations can generate a hash that represents their output result. They only need to implement `hash_output_params` and hash any parameter that affects the output result. Reviewed By: jbakker Differential Revision: https://developer.blender.org/D12341
2021-09-04 16:59:02 +02:00
save_graphviz("compositor_prior_folding");
Compositor: Constant folding Currently there is no clear way to know if an operation is constant (i.e. when all rendered pixels have same values). Operations may need to get constant input values before rendering to determine their resolution or areas of interest. This is the case of scale, rotate and translate operations. Only "set operations" are known as constant but many more are constant when all their inputs are so. Such cases can be optimized by only rendering one pixel. Current solution for tiled implementation is to get first pixel from input. This works for root execution groups, others need previous groups to be rendered. On full frame implementation this is not possible, because buffers are created on rendering to reduce peak memory and there is no per pixel calls. This patch evaluates all operations that are constant into primitive operations (Value/Vector/Color) before determining resolutions. Reviewed By: jbakker Differential Revision: https://developer.blender.org/D11490
2021-07-06 16:16:08 +02:00
ConstantFolder folder(*this);
folder.fold_operations();
}
Compositor: Replace resolution concept by canvas This is a code refactor in preparation of supporting canvas compositing. See {D12466}. No functional changes, all canvases are at (0,0) position matching tiled implementation. Differential Revision: https://developer.blender.org/D12465
2021-09-28 19:32:49 +02:00
determine_canvases();
Cleanup: trailing space for compositor
2018-06-17 17:05:29 +02:00
Compositor: Merge equal operations Some operations can take a lot of time to execute and any duplication should be avoided. This patch implements a compile step that detects operations with the same type, inputs and parameters that produce the same result and merge them. Now operations can generate a hash that represents their output result. They only need to implement `hash_output_params` and hash any parameter that affects the output result. Reviewed By: jbakker Differential Revision: https://developer.blender.org/D12341
2021-09-04 16:59:02 +02:00
save_graphviz("compositor_prior_merging");
merge_equal_operations();
Cleanup: replace members `m_` prefix by `_` suffix in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:04 +02:00
if (context_->get_execution_model() == eExecutionModel::Tiled) {
Compositor: Full-frame base system This patch adds the base code needed to make the full-frame system work for both current tiled/per-pixel implementation of operations and full-frame. Two execution models: - Tiled: Current implementation. Renders execution groups in tiles from outputs to input. Not all operations are buffered. Runs the tiled/per-pixel implementation. - FullFrame: All operations are buffered. Fully renders operations from inputs to outputs. Runs full-frame implementation of operations if available otherwise the current tiled/per-pixel. Creates output buffers on first read and free them as soon as all its readers have finished, reducing peak memory usage of complex/long trees. Operations are multi-threaded but do not run in parallel as Tiled (will be done in another patch). This should allow us to convert operations to full-frame in small steps with the system already working and solve the problem of high memory usage. FullFrame breaking changes respect Tiled system, mainly: - Translate, Rotate, Scale, and Transform take effect immediately instead of next buffered operation. - Any sampling is always done over inputs instead of last buffered operation. Reviewed By: jbakker Differential Revision: https://developer.blender.org/D11113
2021-06-01 10:25:38 +02:00
/* surround complex ops with read/write buffer */
add_complex_operation_buffers();
}
Cleanup: trailing space for compositor
2018-06-17 17:05:29 +02:00
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
/* links not available from here on */
Cleanup: replace members `m_` prefix by `_` suffix in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:04 +02:00
/* XXX make links_ a local variable to avoid confusion! */
links_.clear();
Cleanup: trailing space for compositor
2018-06-17 17:05:29 +02:00
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
prune_operations();
Cleanup: trailing space for compositor
2018-06-17 17:05:29 +02:00
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
/* ensure topological (link-based) order of nodes */
Cleanup: use C++ comments for disabled code Also ensure space around text in C-comment blocks.
2022-04-13 13:46:22 +10:00
// sort_operations(); /* not needed yet. */
Cleanup: trailing space for compositor
2018-06-17 17:05:29 +02:00
Cleanup: replace members `m_` prefix by `_` suffix in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:04 +02:00
if (context_->get_execution_model() == eExecutionModel::Tiled) {
Compositor: Full-frame base system This patch adds the base code needed to make the full-frame system work for both current tiled/per-pixel implementation of operations and full-frame. Two execution models: - Tiled: Current implementation. Renders execution groups in tiles from outputs to input. Not all operations are buffered. Runs the tiled/per-pixel implementation. - FullFrame: All operations are buffered. Fully renders operations from inputs to outputs. Runs full-frame implementation of operations if available otherwise the current tiled/per-pixel. Creates output buffers on first read and free them as soon as all its readers have finished, reducing peak memory usage of complex/long trees. Operations are multi-threaded but do not run in parallel as Tiled (will be done in another patch). This should allow us to convert operations to full-frame in small steps with the system already working and solve the problem of high memory usage. FullFrame breaking changes respect Tiled system, mainly: - Translate, Rotate, Scale, and Transform take effect immediately instead of next buffered operation. - Any sampling is always done over inputs instead of last buffered operation. Reviewed By: jbakker Differential Revision: https://developer.blender.org/D11113
2021-06-01 10:25:38 +02:00
/* create execution groups */
group_operations();
}
Cleanup: trailing space for compositor
2018-06-17 17:05:29 +02:00
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
/* transfer resulting operations to the system */
Cleanup: replace members `m_` prefix by `_` suffix in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:04 +02:00
system->set_operations(operations_, groups_);
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
}
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
void NodeOperationBuilder::add_operation(NodeOperation *operation)
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
{
Cleanup: replace members `m_` prefix by `_` suffix in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:04 +02:00
operation->set_id(operations_.size());
operations_.append(operation);
if (current_node_) {
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
operation->set_name(current_node_->get_bnode()->name);
Compositor: Debug stream operator. Stream operators for NodeOperator and ExecutionGroup to help during debugging.
2021-04-02 15:24:34 +02:00
}
Cleanup: replace members `m_` prefix by `_` suffix in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:04 +02:00
operation->set_execution_model(context_->get_execution_model());
Compositor: Fix execution system unset during constant folding
2021-07-06 19:54:13 +02:00
operation->set_execution_system(exec_system_);
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
}
Compositor: Constant folding Currently there is no clear way to know if an operation is constant (i.e. when all rendered pixels have same values). Operations may need to get constant input values before rendering to determine their resolution or areas of interest. This is the case of scale, rotate and translate operations. Only "set operations" are known as constant but many more are constant when all their inputs are so. Such cases can be optimized by only rendering one pixel. Current solution for tiled implementation is to get first pixel from input. This works for root execution groups, others need previous groups to be rendered. On full frame implementation this is not possible, because buffers are created on rendering to reduce peak memory and there is no per pixel calls. This patch evaluates all operations that are constant into primitive operations (Value/Vector/Color) before determining resolutions. Reviewed By: jbakker Differential Revision: https://developer.blender.org/D11490
2021-07-06 16:16:08 +02:00
void NodeOperationBuilder::replace_operation_with_constant(NodeOperation *operation,
ConstantOperation *constant_operation)
{
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
BLI_assert(constant_operation->get_number_of_input_sockets() == 0);
Compositor: Merge equal operations Some operations can take a lot of time to execute and any duplication should be avoided. This patch implements a compile step that detects operations with the same type, inputs and parameters that produce the same result and merge them. Now operations can generate a hash that represents their output result. They only need to implement `hash_output_params` and hash any parameter that affects the output result. Reviewed By: jbakker Differential Revision: https://developer.blender.org/D12341
2021-09-04 16:59:02 +02:00
unlink_inputs_and_relink_outputs(operation, constant_operation);
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
add_operation(constant_operation);
Compositor: Merge equal operations Some operations can take a lot of time to execute and any duplication should be avoided. This patch implements a compile step that detects operations with the same type, inputs and parameters that produce the same result and merge them. Now operations can generate a hash that represents their output result. They only need to implement `hash_output_params` and hash any parameter that affects the output result. Reviewed By: jbakker Differential Revision: https://developer.blender.org/D12341
2021-09-04 16:59:02 +02:00
}
void NodeOperationBuilder::unlink_inputs_and_relink_outputs(NodeOperation *unlinked_op,
NodeOperation *linked_op)
{
Compositor: Constant folding Currently there is no clear way to know if an operation is constant (i.e. when all rendered pixels have same values). Operations may need to get constant input values before rendering to determine their resolution or areas of interest. This is the case of scale, rotate and translate operations. Only "set operations" are known as constant but many more are constant when all their inputs are so. Such cases can be optimized by only rendering one pixel. Current solution for tiled implementation is to get first pixel from input. This works for root execution groups, others need previous groups to be rendered. On full frame implementation this is not possible, because buffers are created on rendering to reduce peak memory and there is no per pixel calls. This patch evaluates all operations that are constant into primitive operations (Value/Vector/Color) before determining resolutions. Reviewed By: jbakker Differential Revision: https://developer.blender.org/D11490
2021-07-06 16:16:08 +02:00
int i = 0;
Cleanup: replace members `m_` prefix by `_` suffix in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:04 +02:00
while (i < links_.size()) {
Link &link = links_[i];
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
if (&link.to()->get_operation() == unlinked_op) {
link.to()->set_link(nullptr);
Cleanup: replace members `m_` prefix by `_` suffix in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:04 +02:00
links_.remove(i);
Compositor: Constant folding Currently there is no clear way to know if an operation is constant (i.e. when all rendered pixels have same values). Operations may need to get constant input values before rendering to determine their resolution or areas of interest. This is the case of scale, rotate and translate operations. Only "set operations" are known as constant but many more are constant when all their inputs are so. Such cases can be optimized by only rendering one pixel. Current solution for tiled implementation is to get first pixel from input. This works for root execution groups, others need previous groups to be rendered. On full frame implementation this is not possible, because buffers are created on rendering to reduce peak memory and there is no per pixel calls. This patch evaluates all operations that are constant into primitive operations (Value/Vector/Color) before determining resolutions. Reviewed By: jbakker Differential Revision: https://developer.blender.org/D11490
2021-07-06 16:16:08 +02:00
continue;
}
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
if (&link.from()->get_operation() == unlinked_op) {
link.to()->set_link(linked_op->get_output_socket());
links_[i] = Link(linked_op->get_output_socket(), link.to());
Compositor: Constant folding Currently there is no clear way to know if an operation is constant (i.e. when all rendered pixels have same values). Operations may need to get constant input values before rendering to determine their resolution or areas of interest. This is the case of scale, rotate and translate operations. Only "set operations" are known as constant but many more are constant when all their inputs are so. Such cases can be optimized by only rendering one pixel. Current solution for tiled implementation is to get first pixel from input. This works for root execution groups, others need previous groups to be rendered. On full frame implementation this is not possible, because buffers are created on rendering to reduce peak memory and there is no per pixel calls. This patch evaluates all operations that are constant into primitive operations (Value/Vector/Color) before determining resolutions. Reviewed By: jbakker Differential Revision: https://developer.blender.org/D11490
2021-07-06 16:16:08 +02:00
}
i++;
}
}
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
void NodeOperationBuilder::map_input_socket(NodeInput *node_socket,
NodeOperationInput *operation_socket)
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
{
Cleanup: replace members `m_` prefix by `_` suffix in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:04 +02:00
BLI_assert(current_node_);
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
BLI_assert(node_socket->get_node() == current_node_);
Cleanup: trailing space for compositor
2018-06-17 17:05:29 +02:00
Cleanup: consistent use of tags: NOTE/TODO/FIXME/XXX Also use doxy style function reference `#` prefix chars when referencing identifiers.
2021-07-03 23:08:40 +10:00
/* NOTE: this maps operation sockets to node sockets.
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
* for resolving links the map will be inverted first in convert_to_operations,
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
* to get a list of links for each node input socket.
*/
Cleanup: replace members `m_` prefix by `_` suffix in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:04 +02:00
input_map_.add_new(operation_socket, node_socket);
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
}
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
void NodeOperationBuilder::map_output_socket(NodeOutput *node_socket,
NodeOperationOutput *operation_socket)
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
{
Cleanup: replace members `m_` prefix by `_` suffix in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:04 +02:00
BLI_assert(current_node_);
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
BLI_assert(node_socket->get_node() == current_node_);
Cleanup: trailing space for compositor
2018-06-17 17:05:29 +02:00
Cleanup: replace members `m_` prefix by `_` suffix in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:04 +02:00
output_map_.add_new(node_socket, operation_socket);
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
}
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
void NodeOperationBuilder::add_link(NodeOperationOutput *from, NodeOperationInput *to)
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
{
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
if (to->is_connected()) {
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
return;
Cleanup: style, use braces for compositor
2019-04-23 11:21:22 +10:00
}
Cleanup: trailing space for compositor
2018-06-17 17:05:29 +02:00
Cleanup: replace members `m_` prefix by `_` suffix in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:04 +02:00
links_.append(Link(from, to));
Cleanup: trailing space for compositor
2018-06-17 17:05:29 +02:00
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
/* register with the input */
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
to->set_link(from);
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
}
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
void NodeOperationBuilder::remove_input_link(NodeOperationInput *to)
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
{
Cleanup: Replace std::vector with blender::Vector.
2021-03-16 14:15:36 +01:00
int index = 0;
Cleanup: replace members `m_` prefix by `_` suffix in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:04 +02:00
for (Link &link : links_) {
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
if (link.to() == to) {
/* unregister with the input */
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
to->set_link(nullptr);
Cleanup: trailing space for compositor
2018-06-17 17:05:29 +02:00
Cleanup: replace members `m_` prefix by `_` suffix in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:04 +02:00
links_.remove(index);
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
return;
}
Cleanup: Replace std::vector with blender::Vector.
2021-03-16 14:15:36 +01:00
index++;
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
}
}
PreviewOperation *NodeOperationBuilder::make_preview_operation() const
{
Cleanup: replace members `m_` prefix by `_` suffix in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:04 +02:00
BLI_assert(current_node_);
Cleanup: trailing space for compositor
2018-06-17 17:05:29 +02:00
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
if (!(current_node_->get_bnode()->flag & NODE_PREVIEW)) {
Cleanup: Clang-Tidy modernize-use-nullptr Replace `NULL` with `nullptr` in C++ code. No functional changes.
2020-11-06 17:49:09 +01:00
return nullptr;
Cleanup: style, use braces for compositor
2019-04-23 11:21:22 +10:00
}
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
/* previews only in the active group */
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
if (!current_node_->is_in_active_group()) {
Cleanup: Clang-Tidy modernize-use-nullptr Replace `NULL` with `nullptr` in C++ code. No functional changes.
2020-11-06 17:49:09 +01:00
return nullptr;
Cleanup: style, use braces for compositor
2019-04-23 11:21:22 +10:00
}
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
/* do not calculate previews of hidden nodes */
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
if (current_node_->get_bnode()->flag & NODE_HIDDEN) {
Cleanup: Clang-Tidy modernize-use-nullptr Replace `NULL` with `nullptr` in C++ code. No functional changes.
2020-11-06 17:49:09 +01:00
return nullptr;
Cleanup: style, use braces for compositor
2019-04-23 11:21:22 +10:00
}
Cleanup: trailing space for compositor
2018-06-17 17:05:29 +02:00
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
bNodeInstanceHash *previews = context_->get_preview_hash();
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
if (previews) {
Cleanup: refactor passing of color management settings for image save Make a copy of ImageFormatData that contains the effective color management settings, and pass that along to the various functions. This will make it possible to add more complex logic later. For compositing nodes, passing along view and display settings through many functions made it harder to add additional settings, so just get those from the scene now. Differential Revision: https://developer.blender.org/D14401
2022-03-11 18:21:05 +01:00
Scene *scene = context_->get_scene();
PreviewOperation *operation = new PreviewOperation(&scene->view_settings,
&scene->display_settings,
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
current_node_->get_bnode()->preview_xsize,
current_node_->get_bnode()->preview_ysize);
operation->set_bnodetree(context_->get_bnodetree());
operation->verify_preview(previews, current_node_->get_instance_key());
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
return operation;
}
Cleanup: trailing space for compositor
2018-06-17 17:05:29 +02:00
Cleanup: Clang-Tidy modernize-use-nullptr Replace `NULL` with `nullptr` in C++ code. No functional changes.
2020-11-06 17:49:09 +01:00
return nullptr;
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
}
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
void NodeOperationBuilder::add_preview(NodeOperationOutput *output)
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
{
PreviewOperation *operation = make_preview_operation();
if (operation) {
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
add_operation(operation);
Cleanup: trailing space for compositor
2018-06-17 17:05:29 +02:00
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
add_link(output, operation->get_input_socket(0));
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
}
}
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
void NodeOperationBuilder::add_node_input_preview(NodeInput *input)
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
{
PreviewOperation *operation = make_preview_operation();
if (operation) {
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
add_operation(operation);
Cleanup: trailing space for compositor
2018-06-17 17:05:29 +02:00
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
map_input_socket(input, operation->get_input_socket(0));
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
}
}
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
void NodeOperationBuilder::register_viewer(ViewerOperation *viewer)
Fix T40986: crash on using the viewer node inside of group nodes. Viewers were activated both inside the active group as well as the top level tree (the latter being a quick fix for getting a fallback viewer). This caused a race condition on the shared viewer image. Now the active viewer is defined at node conversion time in the converter so that only one can be active at a time without each node having to follow complicated rules for exclusion.
2014-07-08 12:48:41 +02:00
{
Cleanup: replace members `m_` prefix by `_` suffix in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:04 +02:00
if (active_viewer_) {
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
if (current_node_->is_in_active_group()) {
Fix T40986: crash on using the viewer node inside of group nodes. Viewers were activated both inside the active group as well as the top level tree (the latter being a quick fix for getting a fallback viewer). This caused a race condition on the shared viewer image. Now the active viewer is defined at node conversion time in the converter so that only one can be active at a time without each node having to follow complicated rules for exclusion.
2014-07-08 12:48:41 +02:00
/* deactivate previous viewer */
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
active_viewer_->set_active(false);
Cleanup: trailing space for compositor
2018-06-17 17:05:29 +02:00
Cleanup: replace members `m_` prefix by `_` suffix in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:04 +02:00
active_viewer_ = viewer;
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
viewer->set_active(true);
Fix T40986: crash on using the viewer node inside of group nodes. Viewers were activated both inside the active group as well as the top level tree (the latter being a quick fix for getting a fallback viewer). This caused a race condition on the shared viewer image. Now the active viewer is defined at node conversion time in the converter so that only one can be active at a time without each node having to follow complicated rules for exclusion.
2014-07-08 12:48:41 +02:00
}
}
else {
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
if (current_node_->get_bnodetree() == context_->get_bnodetree()) {
Cleanup: replace members `m_` prefix by `_` suffix in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:04 +02:00
active_viewer_ = viewer;
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
viewer->set_active(true);
Fix T40986: crash on using the viewer node inside of group nodes. Viewers were activated both inside the active group as well as the top level tree (the latter being a quick fix for getting a fallback viewer). This caused a race condition on the shared viewer image. Now the active viewer is defined at node conversion time in the converter so that only one can be active at a time without each node having to follow complicated rules for exclusion.
2014-07-08 12:48:41 +02:00
}
}
}
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
/****************************
**** Optimization Steps ****
****************************/
void NodeOperationBuilder::add_datatype_conversions()
{
Cleanup: remove loading blender namespace from Vector.
2021-04-02 16:13:27 +02:00
Vector<Link> convert_links;
Cleanup: replace members `m_` prefix by `_` suffix in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:04 +02:00
for (const Link &link : links_) {
Fix T41067: Muted nodes still do data type conversions. Proxy operations from muted nodes would still create conversion operations where the datatypes don't match, which creates unexpected behavior. Arguably datatype conversion could still happen even when the main operation is muted, but this would be a design change and so is disabled now.
2014-07-15 10:55:49 +02:00
/* proxy operations can skip data type conversion */
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
NodeOperation *from_op = &link.from()->get_operation();
NodeOperation *to_op = &link.to()->get_operation();
Cleanup: Replace virtual methods with bitflags.
2021-03-30 08:23:09 +02:00
if (!(from_op->get_flags().use_datatype_conversion ||
to_op->get_flags().use_datatype_conversion)) {
Fix T41067: Muted nodes still do data type conversions. Proxy operations from muted nodes would still create conversion operations where the datatypes don't match, which creates unexpected behavior. Arguably datatype conversion could still happen even when the main operation is muted, but this would be a design change and so is disabled now.
2014-07-15 10:55:49 +02:00
continue;
Cleanup: style, use braces for compositor
2019-04-23 11:21:22 +10:00
}
Cleanup: trailing space for compositor
2018-06-17 17:05:29 +02:00
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
if (link.from()->get_data_type() != link.to()->get_data_type()) {
Cleanup: Replace std::vector with blender::Vector.
2021-03-16 14:15:36 +01:00
convert_links.append(link);
Cleanup: style, use braces for compositor
2019-04-23 11:21:22 +10:00
}
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
}
Cleanup: Modernize iterators.
2021-03-16 13:47:30 +01:00
for (const Link &link : convert_links) {
Cleanup: COM_convert_data_types parameters.
2021-03-05 10:54:29 +01:00
NodeOperation *converter = COM_convert_data_type(*link.from(), *link.to());
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
if (converter) {
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
add_operation(converter);
Cleanup: trailing space for compositor
2018-06-17 17:05:29 +02:00
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
remove_input_link(link.to());
add_link(link.from(), converter->get_input_socket(0));
add_link(converter->get_output_socket(0), link.to());
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
}
}
}
void NodeOperationBuilder::add_operation_input_constants()
{
Cleanup: consistent use of tags: NOTE/TODO/FIXME/XXX Also use doxy style function reference `#` prefix chars when referencing identifiers.
2021-07-03 23:08:40 +10:00
/* NOTE: unconnected inputs cached first to avoid modifying
Cleanup: replace members `m_` prefix by `_` suffix in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:04 +02:00
* operations_ while iterating over it
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
*/
Cleanup: remove loading blender namespace from Vector.
2021-04-02 16:13:27 +02:00
Vector<NodeOperationInput *> pending_inputs;
Cleanup: replace members `m_` prefix by `_` suffix in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:04 +02:00
for (NodeOperation *op : operations_) {
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
for (int k = 0; k < op->get_number_of_input_sockets(); ++k) {
NodeOperationInput *input = op->get_input_socket(k);
if (!input->is_connected()) {
Cleanup: Replace std::vector with blender::Vector.
2021-03-19 08:11:11 +01:00
pending_inputs.append(input);
Cleanup: style, use braces for compositor
2019-04-23 11:21:22 +10:00
}
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
}
}
Cleanup: Replace std::vector with blender::Vector.
2021-03-19 08:11:11 +01:00
for (NodeOperationInput *input : pending_inputs) {
Cleanup: replace members `m_` prefix by `_` suffix in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:04 +02:00
add_input_constant_value(input, input_map_.lookup_default(input, nullptr));
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
}
}
ClangFormat: apply to source, most of intern Apply clang format as proposed in T53211. For details on usage and instructions for migrating branches without conflicts, see: https://wiki.blender.org/wiki/Tools/ClangFormat
2019-04-17 06:17:24 +02:00
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
void NodeOperationBuilder::add_input_constant_value(NodeOperationInput *input,
Cleanup: add const.
2021-03-19 08:18:00 +01:00
const NodeInput *node_input)
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
{
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
switch (input->get_data_type()) {
Cleanup: Use enum class for DataType.
2021-03-19 14:26:24 +01:00
case DataType::Value: {
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
float value;
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
if (node_input && node_input->get_bnode_socket()) {
value = node_input->get_editor_value_float();
Cleanup: style, use braces for compositor
2019-04-23 11:21:22 +10:00
}
else {
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
value = 0.0f;
Cleanup: style, use braces for compositor
2019-04-23 11:21:22 +10:00
}
ClangFormat: apply to source, most of intern Apply clang format as proposed in T53211. For details on usage and instructions for migrating branches without conflicts, see: https://wiki.blender.org/wiki/Tools/ClangFormat
2019-04-17 06:17:24 +02:00
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
SetValueOperation *op = new SetValueOperation();
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
op->set_value(value);
add_operation(op);
add_link(op->get_output_socket(), input);
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
break;
}
Cleanup: Use enum class for DataType.
2021-03-19 14:26:24 +01:00
case DataType::Color: {
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
float value[4];
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
if (node_input && node_input->get_bnode_socket()) {
node_input->get_editor_value_color(value);
Cleanup: style, use braces for compositor
2019-04-23 11:21:22 +10:00
}
else {
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
zero_v4(value);
Cleanup: style, use braces for compositor
2019-04-23 11:21:22 +10:00
}
ClangFormat: apply to source, most of intern Apply clang format as proposed in T53211. For details on usage and instructions for migrating branches without conflicts, see: https://wiki.blender.org/wiki/Tools/ClangFormat
2019-04-17 06:17:24 +02:00
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
SetColorOperation *op = new SetColorOperation();
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
op->set_channels(value);
add_operation(op);
add_link(op->get_output_socket(), input);
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
break;
}
Cleanup: Use enum class for DataType.
2021-03-19 14:26:24 +01:00
case DataType::Vector: {
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
float value[3];
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
if (node_input && node_input->get_bnode_socket()) {
node_input->get_editor_value_vector(value);
Cleanup: style, use braces for compositor
2019-04-23 11:21:22 +10:00
}
else {
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
zero_v3(value);
Cleanup: style, use braces for compositor
2019-04-23 11:21:22 +10:00
}
ClangFormat: apply to source, most of intern Apply clang format as proposed in T53211. For details on usage and instructions for migrating branches without conflicts, see: https://wiki.blender.org/wiki/Tools/ClangFormat
2019-04-17 06:17:24 +02:00
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
SetVectorOperation *op = new SetVectorOperation();
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
op->set_vector(value);
add_operation(op);
add_link(op->get_output_socket(), input);
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
break;
}
}
}
void NodeOperationBuilder::resolve_proxies()
{
Cleanup: remove loading blender namespace from Vector.
2021-04-02 16:13:27 +02:00
Vector<Link> proxy_links;
Cleanup: replace members `m_` prefix by `_` suffix in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:04 +02:00
for (const Link &link : links_) {
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
/* don't replace links from proxy to proxy, since we may need them for replacing others! */
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
if (link.from()->get_operation().get_flags().is_proxy_operation &&
!link.to()->get_operation().get_flags().is_proxy_operation) {
Cleanup: Replace std::vector with blender::Vector.
2021-03-16 14:15:36 +01:00
proxy_links.append(link);
ClangFormat: apply to source, most of intern Apply clang format as proposed in T53211. For details on usage and instructions for migrating branches without conflicts, see: https://wiki.blender.org/wiki/Tools/ClangFormat
2019-04-17 06:17:24 +02:00
}
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
}
ClangFormat: apply to source, most of intern Apply clang format as proposed in T53211. For details on usage and instructions for migrating branches without conflicts, see: https://wiki.blender.org/wiki/Tools/ClangFormat
2019-04-17 06:17:24 +02:00
Cleanup: Replace std::vector with blender::Vector.
2021-03-16 14:15:36 +01:00
for (const Link &link : proxy_links) {
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
NodeOperationInput *to = link.to();
NodeOperationOutput *from = link.from();
do {
/* walk upstream bypassing the proxy operation */
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
from = from->get_operation().get_input_socket(0)->get_link();
} while (from && from->get_operation().get_flags().is_proxy_operation);
ClangFormat: apply to source, most of intern Apply clang format as proposed in T53211. For details on usage and instructions for migrating branches without conflicts, see: https://wiki.blender.org/wiki/Tools/ClangFormat
2019-04-17 06:17:24 +02:00
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
remove_input_link(to);
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
/* we may not have a final proxy input link,
* in that case it just gets dropped
*/
Cleanup: style, use braces for compositor
2019-04-23 11:21:22 +10:00
if (from) {
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
add_link(from, to);
Cleanup: style, use braces for compositor
2019-04-23 11:21:22 +10:00
}
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
}
}
Compositor: Replace resolution concept by canvas This is a code refactor in preparation of supporting canvas compositing. See {D12466}. No functional changes, all canvases are at (0,0) position matching tiled implementation. Differential Revision: https://developer.blender.org/D12465
2021-09-28 19:32:49 +02:00
void NodeOperationBuilder::determine_canvases()
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
{
Compositor: Replace resolution concept by canvas This is a code refactor in preparation of supporting canvas compositing. See {D12466}. No functional changes, all canvases are at (0,0) position matching tiled implementation. Differential Revision: https://developer.blender.org/D12465
2021-09-28 19:32:49 +02:00
/* Determine all canvas areas of the operations. */
const rcti &preferred_area = COM_AREA_NONE;
Cleanup: replace members `m_` prefix by `_` suffix in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:04 +02:00
for (NodeOperation *op : operations_) {
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
if (op->is_output_operation(context_->is_rendering()) &&
!op->get_flags().is_preview_operation) {
Compositor: Replace resolution concept by canvas This is a code refactor in preparation of supporting canvas compositing. See {D12466}. No functional changes, all canvases are at (0,0) position matching tiled implementation. Differential Revision: https://developer.blender.org/D12465
2021-09-28 19:32:49 +02:00
rcti canvas = COM_AREA_NONE;
op->determine_canvas(preferred_area, canvas);
op->set_canvas(canvas);
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
}
}
ClangFormat: apply to source, most of intern Apply clang format as proposed in T53211. For details on usage and instructions for migrating branches without conflicts, see: https://wiki.blender.org/wiki/Tools/ClangFormat
2019-04-17 06:17:24 +02:00
Cleanup: replace members `m_` prefix by `_` suffix in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:04 +02:00
for (NodeOperation *op : operations_) {
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
if (op->is_output_operation(context_->is_rendering()) &&
op->get_flags().is_preview_operation) {
Compositor: Replace resolution concept by canvas This is a code refactor in preparation of supporting canvas compositing. See {D12466}. No functional changes, all canvases are at (0,0) position matching tiled implementation. Differential Revision: https://developer.blender.org/D12465
2021-09-28 19:32:49 +02:00
rcti canvas = COM_AREA_NONE;
op->determine_canvas(preferred_area, canvas);
op->set_canvas(canvas);
ClangFormat: apply to source, most of intern Apply clang format as proposed in T53211. For details on usage and instructions for migrating branches without conflicts, see: https://wiki.blender.org/wiki/Tools/ClangFormat
2019-04-17 06:17:24 +02:00
}
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
}
ClangFormat: apply to source, most of intern Apply clang format as proposed in T53211. For details on usage and instructions for migrating branches without conflicts, see: https://wiki.blender.org/wiki/Tools/ClangFormat
2019-04-17 06:17:24 +02:00
Compositor: Replace resolution concept by canvas This is a code refactor in preparation of supporting canvas compositing. See {D12466}. No functional changes, all canvases are at (0,0) position matching tiled implementation. Differential Revision: https://developer.blender.org/D12465
2021-09-28 19:32:49 +02:00
/* Convert operation canvases when needed. */
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
{
Cleanup: remove loading blender namespace from Vector.
2021-04-02 16:13:27 +02:00
Vector<Link> convert_links;
Cleanup: replace members `m_` prefix by `_` suffix in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:04 +02:00
for (const Link &link : links_) {
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
if (link.to()->get_resize_mode() != ResizeMode::None) {
const rcti &from_canvas = link.from()->get_operation().get_canvas();
const rcti &to_canvas = link.to()->get_operation().get_canvas();
Compositor: Add support for canvas compositing This commit adds functionality for operations that require pixel translation or resizing on "Full Frame" mode, allowing to adjust their canvas. It fixes most cropping issues in translate, scale, rotate and transform nodes by adjusting their canvas to the result, instead of the input canvas. Operations output buffer is still always on (0,0) position for easier image algorithm implementation, even when the canvas is not. Current limitations (will be addressed on bcon2): - Displayed translation in Viewer node is limited to 6000px. - When scaling up the canvas size is limited to the scene resolution size x 1.5 . From that point it crops. If none of these limitations are hit, the Viewer node displays the full input with any translation. Differential Revision: https://developer.blender.org/D12466
2021-09-28 19:33:06 +02:00
bool needs_conversion;
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
if (link.to()->get_resize_mode() == ResizeMode::Align) {
Compositor: Add support for canvas compositing This commit adds functionality for operations that require pixel translation or resizing on "Full Frame" mode, allowing to adjust their canvas. It fixes most cropping issues in translate, scale, rotate and transform nodes by adjusting their canvas to the result, instead of the input canvas. Operations output buffer is still always on (0,0) position for easier image algorithm implementation, even when the canvas is not. Current limitations (will be addressed on bcon2): - Displayed translation in Viewer node is limited to 6000px. - When scaling up the canvas size is limited to the scene resolution size x 1.5 . From that point it crops. If none of these limitations are hit, the Viewer node displays the full input with any translation. Differential Revision: https://developer.blender.org/D12466
2021-09-28 19:33:06 +02:00
needs_conversion = from_canvas.xmin != to_canvas.xmin ||
from_canvas.ymin != to_canvas.ymin;
}
else {
needs_conversion = !BLI_rcti_compare(&from_canvas, &to_canvas);
}
if (needs_conversion) {
Cleanup: Replace std::vector with blender::Vector.
2021-03-16 14:15:36 +01:00
convert_links.append(link);
Cleanup: style, use braces for compositor
2019-04-23 11:21:22 +10:00
}
ClangFormat: apply to source, most of intern Apply clang format as proposed in T53211. For details on usage and instructions for migrating branches without conflicts, see: https://wiki.blender.org/wiki/Tools/ClangFormat
2019-04-17 06:17:24 +02:00
}
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
}
Cleanup: Replace std::vector with blender::Vector.
2021-03-16 14:15:36 +01:00
for (const Link &link : convert_links) {
Compositor: Replace resolution concept by canvas This is a code refactor in preparation of supporting canvas compositing. See {D12466}. No functional changes, all canvases are at (0,0) position matching tiled implementation. Differential Revision: https://developer.blender.org/D12465
2021-09-28 19:32:49 +02:00
COM_convert_canvas(*this, link.from(), link.to());
ClangFormat: apply to source, most of intern Apply clang format as proposed in T53211. For details on usage and instructions for migrating branches without conflicts, see: https://wiki.blender.org/wiki/Tools/ClangFormat
2019-04-17 06:17:24 +02:00
}
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
}
}
ClangFormat: apply to source, most of intern Apply clang format as proposed in T53211. For details on usage and instructions for migrating branches without conflicts, see: https://wiki.blender.org/wiki/Tools/ClangFormat
2019-04-17 06:17:24 +02:00
Compositor: Merge equal operations Some operations can take a lot of time to execute and any duplication should be avoided. This patch implements a compile step that detects operations with the same type, inputs and parameters that produce the same result and merge them. Now operations can generate a hash that represents their output result. They only need to implement `hash_output_params` and hash any parameter that affects the output result. Reviewed By: jbakker Differential Revision: https://developer.blender.org/D12341
2021-09-04 16:59:02 +02:00
static Vector<NodeOperationHash> generate_hashes(Span<NodeOperation *> operations)
{
Vector<NodeOperationHash> hashes;
for (NodeOperation *op : operations) {
std::optional<NodeOperationHash> hash = op->generate_hash();
if (hash) {
hashes.append(std::move(*hash));
}
}
return hashes;
}
void NodeOperationBuilder::merge_equal_operations()
{
Cleanup: improve code clarity Addresses D12341 review.
2021-09-05 14:25:14 +02:00
bool check_for_next_merge = true;
while (check_for_next_merge) {
Compositor: Merge equal operations Some operations can take a lot of time to execute and any duplication should be avoided. This patch implements a compile step that detects operations with the same type, inputs and parameters that produce the same result and merge them. Now operations can generate a hash that represents their output result. They only need to implement `hash_output_params` and hash any parameter that affects the output result. Reviewed By: jbakker Differential Revision: https://developer.blender.org/D12341
2021-09-04 16:59:02 +02:00
/* Re-generate hashes with any change. */
Cleanup: replace members `m_` prefix by `_` suffix in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:04 +02:00
Vector<NodeOperationHash> hashes = generate_hashes(operations_);
Compositor: Merge equal operations Some operations can take a lot of time to execute and any duplication should be avoided. This patch implements a compile step that detects operations with the same type, inputs and parameters that produce the same result and merge them. Now operations can generate a hash that represents their output result. They only need to implement `hash_output_params` and hash any parameter that affects the output result. Reviewed By: jbakker Differential Revision: https://developer.blender.org/D12341
2021-09-04 16:59:02 +02:00
/* Make hashes be consecutive when they are equal. */
std::sort(hashes.begin(), hashes.end());
Cleanup: improve code clarity Addresses D12341 review.
2021-09-05 14:25:14 +02:00
bool any_merged = false;
Compositor: Merge equal operations Some operations can take a lot of time to execute and any duplication should be avoided. This patch implements a compile step that detects operations with the same type, inputs and parameters that produce the same result and merge them. Now operations can generate a hash that represents their output result. They only need to implement `hash_output_params` and hash any parameter that affects the output result. Reviewed By: jbakker Differential Revision: https://developer.blender.org/D12341
2021-09-04 16:59:02 +02:00
const NodeOperationHash *prev_hash = nullptr;
for (const NodeOperationHash &hash : hashes) {
if (prev_hash && *prev_hash == hash) {
merge_equal_operations(prev_hash->get_operation(), hash.get_operation());
any_merged = true;
}
prev_hash = &hash;
}
Cleanup: improve code clarity Addresses D12341 review.
2021-09-05 14:25:14 +02:00
check_for_next_merge = any_merged;
Compositor: Merge equal operations Some operations can take a lot of time to execute and any duplication should be avoided. This patch implements a compile step that detects operations with the same type, inputs and parameters that produce the same result and merge them. Now operations can generate a hash that represents their output result. They only need to implement `hash_output_params` and hash any parameter that affects the output result. Reviewed By: jbakker Differential Revision: https://developer.blender.org/D12341
2021-09-04 16:59:02 +02:00
}
}
void NodeOperationBuilder::merge_equal_operations(NodeOperation *from, NodeOperation *into)
{
unlink_inputs_and_relink_outputs(from, into);
Cleanup: replace members `m_` prefix by `_` suffix in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:04 +02:00
operations_.remove_first_occurrence_and_reorder(from);
Compositor: Merge equal operations Some operations can take a lot of time to execute and any duplication should be avoided. This patch implements a compile step that detects operations with the same type, inputs and parameters that produce the same result and merge them. Now operations can generate a hash that represents their output result. They only need to implement `hash_output_params` and hash any parameter that affects the output result. Reviewed By: jbakker Differential Revision: https://developer.blender.org/D12341
2021-09-04 16:59:02 +02:00
delete from;
}
Cleanup: remove loading blender namespace from Vector.
2021-04-02 16:13:27 +02:00
Vector<NodeOperationInput *> NodeOperationBuilder::cache_output_links(
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
NodeOperationOutput *output) const
{
Cleanup: remove loading blender namespace from Vector.
2021-04-02 16:13:27 +02:00
Vector<NodeOperationInput *> inputs;
Cleanup: replace members `m_` prefix by `_` suffix in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:04 +02:00
for (const Link &link : links_) {
Cleanup: style, use braces for compositor
2019-04-23 11:21:22 +10:00
if (link.from() == output) {
Cleanup: Replace std::vector with blender::Vector.
2021-03-16 14:15:36 +01:00
inputs.append(link.to());
Cleanup: style, use braces for compositor
2019-04-23 11:21:22 +10:00
}
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
}
return inputs;
}
ClangFormat: apply to source, most of intern Apply clang format as proposed in T53211. For details on usage and instructions for migrating branches without conflicts, see: https://wiki.blender.org/wiki/Tools/ClangFormat
2019-04-17 06:17:24 +02:00
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
WriteBufferOperation *NodeOperationBuilder::find_attached_write_buffer_operation(
NodeOperationOutput *output) const
{
Cleanup: replace members `m_` prefix by `_` suffix in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:04 +02:00
for (const Link &link : links_) {
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
if (link.from() == output) {
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
NodeOperation &op = link.to()->get_operation();
Cleanup: Replace `is...Operation()` methods with a flag.
2021-03-29 16:54:02 +02:00
if (op.get_flags().is_write_buffer_operation) {
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
return (WriteBufferOperation *)(&op);
Cleanup: style, use braces for compositor
2019-04-23 11:21:22 +10:00
}
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
}
}
Cleanup: Clang-Tidy modernize-use-nullptr Replace `NULL` with `nullptr` in C++ code. No functional changes.
2020-11-06 17:49:09 +01:00
return nullptr;
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
}
Compositor: Code cleanup, prepare for strict C++ flags
2015-03-27 15:49:07 +05:00
void NodeOperationBuilder::add_input_buffers(NodeOperation * /*operation*/,
NodeOperationInput *input)
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
{
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
if (!input->is_connected()) {
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
return;
Cleanup: style, use braces for compositor
2019-04-23 11:21:22 +10:00
}
Cleanup: trailing space for compositor
2018-06-17 17:05:29 +02:00
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
NodeOperationOutput *output = input->get_link();
if (output->get_operation().get_flags().is_read_buffer_operation) {
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
/* input is already buffered, no need to add another */
return;
}
Cleanup: trailing space for compositor
2018-06-17 17:05:29 +02:00
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
/* this link will be replaced below */
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
remove_input_link(input);
Cleanup: trailing space for compositor
2018-06-17 17:05:29 +02:00
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
/* check of other end already has write operation, otherwise add a new one */
WriteBufferOperation *writeoperation = find_attached_write_buffer_operation(output);
if (!writeoperation) {
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
writeoperation = new WriteBufferOperation(output->get_data_type());
writeoperation->set_bnodetree(context_->get_bnodetree());
add_operation(writeoperation);
Cleanup: trailing space for compositor
2018-06-17 17:05:29 +02:00
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
add_link(output, writeoperation->get_input_socket(0));
Cleanup: trailing space for compositor
2018-06-17 17:05:29 +02:00
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
writeoperation->read_resolution_from_input_socket();
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
}
Cleanup: trailing space for compositor
2018-06-17 17:05:29 +02:00
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
/* add readbuffer op for the input */
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
ReadBufferOperation *readoperation = new ReadBufferOperation(output->get_data_type());
readoperation->set_memory_proxy(writeoperation->get_memory_proxy());
this->add_operation(readoperation);
Cleanup: trailing space for compositor
2018-06-17 17:05:29 +02:00
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
add_link(readoperation->get_output_socket(), input);
Cleanup: trailing space for compositor
2018-06-17 17:05:29 +02:00
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
readoperation->read_resolution_from_write_buffer();
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
}
void NodeOperationBuilder::add_output_buffers(NodeOperation *operation,
NodeOperationOutput *output)
{
/* cache connected sockets, so we can safely remove links first before replacing them */
Cleanup: remove loading blender namespace from Vector.
2021-04-02 16:13:27 +02:00
Vector<NodeOperationInput *> targets = cache_output_links(output);
Cleanup: Replace std::vector with blender::Vector.
2021-03-16 14:15:36 +01:00
if (targets.is_empty()) {
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
return;
Cleanup: style, use braces for compositor
2019-04-23 11:21:22 +10:00
}
Cleanup: trailing space for compositor
2018-06-17 17:05:29 +02:00
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
WriteBufferOperation *write_operation = nullptr;
Cleanup: Replace std::vector with blender::Vector.
2021-03-16 14:15:36 +01:00
for (NodeOperationInput *target : targets) {
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
/* try to find existing write buffer operation */
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
if (target->get_operation().get_flags().is_write_buffer_operation) {
BLI_assert(write_operation == nullptr); /* there should only be one write op connected */
write_operation = (WriteBufferOperation *)(&target->get_operation());
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
}
else {
/* remove all links to other nodes */
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
remove_input_link(target);
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
}
}
Cleanup: trailing space for compositor
2018-06-17 17:05:29 +02:00
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
/* if no write buffer operation exists yet, create a new one */
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
if (!write_operation) {
write_operation = new WriteBufferOperation(operation->get_output_socket()->get_data_type());
write_operation->set_bnodetree(context_->get_bnodetree());
add_operation(write_operation);
Cleanup: trailing space for compositor
2018-06-17 17:05:29 +02:00
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
add_link(output, write_operation->get_input_socket(0));
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
}
Cleanup: trailing space for compositor
2018-06-17 17:05:29 +02:00
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
write_operation->read_resolution_from_input_socket();
Cleanup: trailing space for compositor
2018-06-17 17:05:29 +02:00
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
/* add readbuffer op for every former connected input */
Cleanup: Replace std::vector with blender::Vector.
2021-03-16 14:15:36 +01:00
for (NodeOperationInput *target : targets) {
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
if (&target->get_operation() == write_operation) {
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
continue; /* skip existing write op links */
Cleanup: style, use braces for compositor
2019-04-23 11:21:22 +10:00
}
Cleanup: trailing space for compositor
2018-06-17 17:05:29 +02:00
D627: Memory usage optimization for the compositor. The compostor used a fixed size of 4 floats to hold pixel data. this patch will select size of a pixel based on its type. It uses 1 float for Value, 3 float for vector and 4 floats for color data types. When benchmarking on shots (opening shot of caminandes) we get a reduction of memory of 30% and a tiny speedup as less data transformations needs to take place (but these are negligable. More information of the patch can be found on https://developer.blender.org/D627 and http://wiki.blender.org/index.php/Dev:Ref/Proposals/Compositor2014_p1.1_TD Developers: jbakker & mdewanchand Thanks for Sergey for his indept review.
2015-01-19 18:13:26 +01:00
ReadBufferOperation *readoperation = new ReadBufferOperation(
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
operation->get_output_socket()->get_data_type());
readoperation->set_memory_proxy(write_operation->get_memory_proxy());
add_operation(readoperation);
Cleanup: trailing space for compositor
2018-06-17 17:05:29 +02:00
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
add_link(readoperation->get_output_socket(), target);
Cleanup: trailing space for compositor
2018-06-17 17:05:29 +02:00
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
readoperation->read_resolution_from_write_buffer();
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
}
}
void NodeOperationBuilder::add_complex_operation_buffers()
{
Cleanup: consistent use of tags: NOTE/TODO/FIXME/XXX Also use doxy style function reference `#` prefix chars when referencing identifiers.
2021-07-03 23:08:40 +10:00
/* NOTE: complex ops and get cached here first, since adding operations
Cleanup: replace members `m_` prefix by `_` suffix in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:04 +02:00
* will invalidate iterators over the main operations_
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
*/
Cleanup: remove loading blender namespace from Vector.
2021-04-02 16:13:27 +02:00
Vector<NodeOperation *> complex_ops;
Cleanup: replace members `m_` prefix by `_` suffix in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:04 +02:00
for (NodeOperation *operation : operations_) {
Cleanup: Use Bitflags For Booleans.
2021-03-29 16:45:49 +02:00
if (operation->get_flags().complex) {
Cleanup: Use blender::Vector.
2021-03-05 16:45:11 +01:00
complex_ops.append(operation);
Cleanup: style, use braces for compositor
2019-04-23 11:21:22 +10:00
}
}
Cleanup: trailing space for compositor
2018-06-17 17:05:29 +02:00
Cleanup: Use blender::Vector.
2021-03-05 16:45:11 +01:00
for (NodeOperation *op : complex_ops) {
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
DebugInfo::operation_read_write_buffer(op);
Cleanup: trailing space for compositor
2018-06-17 17:05:29 +02:00
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
for (int index = 0; index < op->get_number_of_input_sockets(); index++) {
add_input_buffers(op, op->get_input_socket(index));
Cleanup: style, use braces for compositor
2019-04-23 11:21:22 +10:00
}
Cleanup: trailing space for compositor
2018-06-17 17:05:29 +02:00
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
for (int index = 0; index < op->get_number_of_output_sockets(); index++) {
add_output_buffers(op, op->get_output_socket(index));
Cleanup: style, use braces for compositor
2019-04-23 11:21:22 +10:00
}
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
}
}
Cleanup: Clang-Tidy, modernize-use-using Replace `typedef` with `using` in C++ code. In the case of `typedef struct SomeName { ... } SomeName;` I removed the `typedef` altogether, as this is unnecessary in C++. Such cases have been rewritten to `struct SomeName { ... };` No functional changes.
2020-12-04 12:46:43 +01:00
using Tags = std::set<NodeOperation *>;
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
static void find_reachable_operations_recursive(Tags &reachable, NodeOperation *op)
{
Cleanup: style, use braces for compositor
2019-04-23 11:21:22 +10:00
if (reachable.find(op) != reachable.end()) {
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
return;
Cleanup: style, use braces for compositor
2019-04-23 11:21:22 +10:00
}
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
reachable.insert(op);
Cleanup: trailing space for compositor
2018-06-17 17:05:29 +02:00
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
for (int i = 0; i < op->get_number_of_input_sockets(); i++) {
NodeOperationInput *input = op->get_input_socket(i);
if (input->is_connected()) {
find_reachable_operations_recursive(reachable, &input->get_link()->get_operation());
Cleanup: style, use braces for compositor
2019-04-23 11:21:22 +10:00
}
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
}
Cleanup: trailing space for compositor
2018-06-17 17:05:29 +02:00
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
/* associated write-buffer operations are executed as well */
Cleanup: Replace `is...Operation()` methods with a flag.
2021-03-29 16:54:02 +02:00
if (op->get_flags().is_read_buffer_operation) {
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
ReadBufferOperation *read_op = (ReadBufferOperation *)op;
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
MemoryProxy *memproxy = read_op->get_memory_proxy();
find_reachable_operations_recursive(reachable, memproxy->get_write_buffer_operation());
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
}
}
void NodeOperationBuilder::prune_operations()
{
Tags reachable;
Cleanup: replace members `m_` prefix by `_` suffix in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:04 +02:00
for (NodeOperation *op : operations_) {
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
/* output operations are primary executed operations */
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
if (op->is_output_operation(context_->is_rendering())) {
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
find_reachable_operations_recursive(reachable, op);
Cleanup: style, use braces for compositor
2019-04-23 11:21:22 +10:00
}
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
}
Cleanup: trailing space for compositor
2018-06-17 17:05:29 +02:00
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
/* delete unreachable operations */
Cleanup: remove loading blender namespace from Vector.
2021-04-02 16:13:27 +02:00
Vector<NodeOperation *> reachable_ops;
Cleanup: replace members `m_` prefix by `_` suffix in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:04 +02:00
for (NodeOperation *op : operations_) {
Cleanup: style, use braces for compositor
2019-04-23 11:21:22 +10:00
if (reachable.find(op) != reachable.end()) {
Cleanup: Use blender::Vector.
2021-03-05 16:45:11 +01:00
reachable_ops.append(op);
Cleanup: style, use braces for compositor
2019-04-23 11:21:22 +10:00
}
else {
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
delete op;
Cleanup: style, use braces for compositor
2019-04-23 11:21:22 +10:00
}
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
}
/* finally replace the operations list with the pruned list */
Cleanup: replace members `m_` prefix by `_` suffix in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:04 +02:00
operations_ = reachable_ops;
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
}
/* topological (depth-first) sorting of operations */
Cleanup: remove loading blender namespace from Vector.
2021-04-02 16:13:27 +02:00
static void sort_operations_recursive(Vector<NodeOperation *> &sorted,
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
Tags &visited,
NodeOperation *op)
{
Cleanup: style, use braces for compositor
2019-04-23 11:21:22 +10:00
if (visited.find(op) != visited.end()) {
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
return;
Cleanup: style, use braces for compositor
2019-04-23 11:21:22 +10:00
}
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
visited.insert(op);
Cleanup: trailing space for compositor
2018-06-17 17:05:29 +02:00
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
for (int i = 0; i < op->get_number_of_input_sockets(); i++) {
NodeOperationInput *input = op->get_input_socket(i);
if (input->is_connected()) {
sort_operations_recursive(sorted, visited, &input->get_link()->get_operation());
Cleanup: style, use braces for compositor
2019-04-23 11:21:22 +10:00
}
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
}
Cleanup: trailing space for compositor
2018-06-17 17:05:29 +02:00
Cleanup: Use blender::Vector.
2021-03-05 16:45:11 +01:00
sorted.append(op);
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
}
void NodeOperationBuilder::sort_operations()
{
Cleanup: remove loading blender namespace from Vector.
2021-04-02 16:13:27 +02:00
Vector<NodeOperation *> sorted;
Cleanup: replace members `m_` prefix by `_` suffix in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:04 +02:00
sorted.reserve(operations_.size());
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
Tags visited;
Cleanup: trailing space for compositor
2018-06-17 17:05:29 +02:00
Cleanup: replace members `m_` prefix by `_` suffix in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:04 +02:00
for (NodeOperation *operation : operations_) {
Cleanup: Use blender::Vector.
2021-03-05 16:45:11 +01:00
sort_operations_recursive(sorted, visited, operation);
Cleanup: style, use braces for compositor
2019-04-23 11:21:22 +10:00
}
Cleanup: trailing space for compositor
2018-06-17 17:05:29 +02:00
Cleanup: replace members `m_` prefix by `_` suffix in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:04 +02:00
operations_ = sorted;
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
}
static void add_group_operations_recursive(Tags &visited, NodeOperation *op, ExecutionGroup *group)
{
Cleanup: style, use braces for compositor
2019-04-23 11:21:22 +10:00
if (visited.find(op) != visited.end()) {
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
return;
Cleanup: style, use braces for compositor
2019-04-23 11:21:22 +10:00
}
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
visited.insert(op);
Cleanup: trailing space for compositor
2018-06-17 17:05:29 +02:00
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
if (!group->add_operation(op)) {
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
return;
Cleanup: style, use braces for compositor
2019-04-23 11:21:22 +10:00
}
Cleanup: trailing space for compositor
2018-06-17 17:05:29 +02:00
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
/* add all eligible input ops to the group */
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
for (int i = 0; i < op->get_number_of_input_sockets(); i++) {
NodeOperationInput *input = op->get_input_socket(i);
if (input->is_connected()) {
add_group_operations_recursive(visited, &input->get_link()->get_operation(), group);
Cleanup: style, use braces for compositor
2019-04-23 11:21:22 +10:00
}
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
}
}
ExecutionGroup *NodeOperationBuilder::make_group(NodeOperation *op)
{
Cleanup: replace members `m_` prefix by `_` suffix in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:04 +02:00
ExecutionGroup *group = new ExecutionGroup(groups_.size());
groups_.append(group);
Cleanup: trailing space for compositor
2018-06-17 17:05:29 +02:00
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
Tags visited;
add_group_operations_recursive(visited, op, group);
Cleanup: trailing space for compositor
2018-06-17 17:05:29 +02:00
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
return group;
}
void NodeOperationBuilder::group_operations()
{
Cleanup: replace members `m_` prefix by `_` suffix in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:04 +02:00
for (NodeOperation *op : operations_) {
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
if (op->is_output_operation(context_->is_rendering())) {
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
ExecutionGroup *group = make_group(op);
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
group->set_output_execution_group(true);
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
}
ClangFormat: apply to source, most of intern Apply clang format as proposed in T53211. For details on usage and instructions for migrating branches without conflicts, see: https://wiki.blender.org/wiki/Tools/ClangFormat
2019-04-17 06:17:24 +02:00
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
/* add new groups for associated memory proxies where needed */
Cleanup: Replace `is...Operation()` methods with a flag.
2021-03-29 16:54:02 +02:00
if (op->get_flags().is_read_buffer_operation) {
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
ReadBufferOperation *read_op = (ReadBufferOperation *)op;
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
MemoryProxy *memproxy = read_op->get_memory_proxy();
ClangFormat: apply to source, most of intern Apply clang format as proposed in T53211. For details on usage and instructions for migrating branches without conflicts, see: https://wiki.blender.org/wiki/Tools/ClangFormat
2019-04-17 06:17:24 +02:00
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
if (memproxy->get_executor() == nullptr) {
ExecutionGroup *group = make_group(memproxy->get_write_buffer_operation());
memproxy->set_executor(group);
Structural cleanup and improvements for the compositor. Many parts of the compositor are unnecessarily complicated. This patch aims at reducing the complexity of writing nodes and making the code more transparent. == Separating Nodes and Operations == Currently these are both mixed in the same graph, even though they have very different purposes and are used at distinct stages in the compositing process. The patch introduces dedicated graph classes for nodes and for operations. This removes the need for a lot of special case checks (isOperation etc.) and explicit type casts. It simplifies the code since it becomes clear at every stage what type of node we are dealing with. The compiler can use static typing to avoid common bugs from mixing up these types and fewer runtime sanity checks are needed. == Simplified Node Conversion == Converting nodes to operations was previously based on "relinking", i.e. nodes would start with by mirroring links in the Blender DNA node trees, then add operations and redirect these links to them. This was very hard to follow in many cases and required a lot of attention to avoid invalid states. Now there is a helper class called the NodeConverter, which is passed to nodes and implements a much simpler API for this process. Nodes can add operations and explicit connections as before, but defining "external" links to the inputs/outputs of the original node now uses mapping instead of directly modifying link data. Input data (node graph) and result (operations graph) are cleanly separated. == Removed Redundant Data Structures == A few redundant data structures have been removed, notably the SocketConnection. These are only needed temporarily during graph construction. For executing the compositor operations it is perfectly sufficient to store only the direct input link pointers. A common pointer indirection is avoided this way (which might also give a little performance improvement). == Avoid virtual recursive functions == Recursive virtual functions are evil. They are very hard to follow during debugging. At least in the parts this patch is concerned with these functions have been replaced by a non-virtual recursive core function (which might then call virtual non-recursive functions if needed). See for example NodeOperationBuilder::group_operations.
2014-04-15 16:06:12 +02:00
}
}
}
}
Cleanup: Add namespace to compositor.
2021-03-23 17:12:27 +01:00
Compositor: Merge equal operations Some operations can take a lot of time to execute and any duplication should be avoided. This patch implements a compile step that detects operations with the same type, inputs and parameters that produce the same result and merge them. Now operations can generate a hash that represents their output result. They only need to implement `hash_output_params` and hash any parameter that affects the output result. Reviewed By: jbakker Differential Revision: https://developer.blender.org/D12341
2021-09-04 16:59:02 +02:00
void NodeOperationBuilder::save_graphviz(StringRefNull name)
{
if (COM_EXPORT_GRAPHVIZ) {
Cleanup: replace members `m_` prefix by `_` suffix in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:04 +02:00
exec_system_->set_operations(operations_, groups_);
Compositor: Merge equal operations Some operations can take a lot of time to execute and any duplication should be avoided. This patch implements a compile step that detects operations with the same type, inputs and parameters that produce the same result and merge them. Now operations can generate a hash that represents their output result. They only need to implement `hash_output_params` and hash any parameter that affects the output result. Reviewed By: jbakker Differential Revision: https://developer.blender.org/D12341
2021-09-04 16:59:02 +02:00
DebugInfo::graphviz(exec_system_, name);
}
}
Compositor: Add Streaming Operator for NodeOperationBuilder. For debugging purposes to convert the internal state of the NodeOperationBuilder to a graphviz. Usage: std::cout << *this << "\n"; Inside any method of the NodeOperationBuilder.
2021-04-06 12:04:47 +02:00
std::ostream &operator<<(std::ostream &os, const NodeOperationBuilder &builder)
{
os << "# Builder start\n";
os << "digraph G {\n";
os << " rankdir=LR;\n";
os << " node [shape=box];\n";
for (const NodeOperation *operation : builder.get_operations()) {
os << " op" << operation->get_id() << " [label=\"" << *operation << "\"];\n";
}
os << "\n";
for (const NodeOperationBuilder::Link &link : builder.get_links()) {
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
os << " op" << link.from()->get_operation().get_id() << " -> op"
<< link.to()->get_operation().get_id() << ";\n";
Compositor: Add Streaming Operator for NodeOperationBuilder. For debugging purposes to convert the internal state of the NodeOperationBuilder to a graphviz. Usage: std::cout << *this << "\n"; Inside any method of the NodeOperationBuilder.
2021-04-06 12:04:47 +02:00
}
for (const NodeOperation *operation : builder.get_operations()) {
if (operation->get_flags().is_read_buffer_operation) {
const ReadBufferOperation &read_operation = static_cast<const ReadBufferOperation &>(
*operation);
const WriteBufferOperation &write_operation =
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
*read_operation.get_memory_proxy()->get_write_buffer_operation();
Compositor: Add Streaming Operator for NodeOperationBuilder. For debugging purposes to convert the internal state of the NodeOperationBuilder to a graphviz. Usage: std::cout << *this << "\n"; Inside any method of the NodeOperationBuilder.
2021-04-06 12:04:47 +02:00
os << " op" << write_operation.get_id() << " -> op" << read_operation.get_id() << ";\n";
}
}
os << "}\n";
os << "# Builder end\n";
return os;
}
std::ostream &operator<<(std::ostream &os, const NodeOperationBuilder::Link &link)
{
Cleanup: convert camelCase naming to snake_case in Compositor To convert old code to the current convention and use a single code style.
2021-10-13 23:01:15 +02:00
os << link.from()->get_operation().get_id() << " -> " << link.to()->get_operation().get_id();
Compositor: Add Streaming Operator for NodeOperationBuilder. For debugging purposes to convert the internal state of the NodeOperationBuilder to a graphviz. Usage: std::cout << *this << "\n"; Inside any method of the NodeOperationBuilder.
2021-04-06 12:04:47 +02:00
return os;
}
Cleanup: Add namespace to compositor.
2021-03-23 17:12:27 +01:00
} // namespace blender::compositor
Reference in New Issue Copy Permalink