Files
test2/source/blender/nodes/NOD_derived_node_tree.hh

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

450 lines
12 KiB
C++
Raw Permalink Normal View History

/* SPDX-FileCopyrightText: 2023 Blender Authors
*
* SPDX-License-Identifier: GPL-2.0-or-later */
#pragma once
/** \file
* \ingroup nodes
*
* DerivedNodeTree makes working with (nested) node groups more convenient and safe. It does so by
* pairing nodes and sockets with a context. The context contains information about the current
* "instance" of the node or socket. A node might be "instanced" multiple times when it is in a
* node group that is used multiple times.
*/
#include "BLI_function_ref.hh"
#include "BLI_linear_allocator.hh"
Geometry Nodes: initial scattering and geometry nodes This is the initial merge from the geometry-nodes branch. Nodes: * Attribute Math * Boolean * Edge Split * Float Compare * Object Info * Point Distribute * Point Instance * Random Attribute * Random Float * Subdivision Surface * Transform * Triangulate It includes the initial evaluation of geometry node groups in the Geometry Nodes modifier. Notes on the Generic attribute access API The API adds an indirection for attribute access. That has the following benefits: * Most code does not have to care about how an attribute is stored internally. This is mainly necessary, because we have to deal with "legacy" attributes such as vertex weights and attributes that are embedded into other structs such as vertex positions. * When reading from an attribute, we generally don't care what domain the attribute is stored on. So we want to abstract away the interpolation that that adapts attributes from one domain to another domain (this is not actually implemented yet). Other possible improvements for later iterations include: * Actually implement interpolation between domains. * Don't use inheritance for the different attribute types. A single class for read access and one for write access might be enough, because we know all the ways in which attributes are stored internally. We don't want more different internal structures in the future. On the contrary, ideally we can consolidate the different storage formats in the future to reduce the need for this indirection. * Remove the need for heap allocations when creating attribute accessors. It includes commits from: * Dalai Felinto * Hans Goudey * Jacques Lucke * Léo Depoix
2020-12-02 13:25:25 +01:00
#include "BLI_vector_set.hh"
#include "BKE_node_runtime.hh"
namespace blender::nodes {
class DTreeContext;
class DerivedNodeTree;
class DNode;
class DSocket;
class DInputSocket;
class DOutputSocket;
/**
* The context attached to every node or socket in a derived node tree. It can be used to determine
* the place of a node in a hierarchy of node groups.
*
* Contexts are organized in a tree data structure to avoid having to store the entire path to the
* root node group for every node/socket.
*/
class DTreeContext {
private:
/* Null when this context is for the root node group. Otherwise it points to the context one
* level up. */
DTreeContext *parent_context_;
/* Null when this context is for the root node group. Otherwise it points to the group node in
* the parent node group that contains this context. */
const bNode *parent_node_;
/* The current node tree. */
const bNodeTree *btree_;
/* The instance key of the parent node. bke::NODE_INSTANCE_KEY_BASE for root contexts. */
bNodeInstanceKey instance_key_;
/* All the children contexts of this context. */
Map<const bNode *, DTreeContext *> children_;
DerivedNodeTree *derived_tree_;
friend DerivedNodeTree;
public:
const bNodeTree &btree() const;
const DTreeContext *parent_context() const;
const bNode *parent_node() const;
bNodeInstanceKey instance_key() const;
const DTreeContext *child_context(const bNode &node) const;
const DerivedNodeTree &derived_tree() const;
bool is_root() const;
};
/**
* A (nullable) reference to a node and the context it is in. It is unique within an entire nested
* node group hierarchy. This type is small and can be passed around by value.
*/
class DNode {
private:
const DTreeContext *context_ = nullptr;
const bNode *bnode_ = nullptr;
public:
DNode() = default;
DNode(const DTreeContext *context, const bNode *node);
const DTreeContext *context() const;
const bNode *bnode() const;
bNodeInstanceKey instance_key() const;
const bNode *operator->() const;
const bNode &operator*() const;
BLI_STRUCT_EQUALITY_OPERATORS_2(DNode, context_, bnode_)
operator bool() const;
uint64_t hash() const;
DInputSocket input(int index) const;
DOutputSocket output(int index) const;
DInputSocket input_by_identifier(StringRef identifier) const;
DOutputSocket output_by_identifier(StringRef identifier) const;
};
/**
* A (nullable) reference to a socket and the context it is in. It is unique within an entire
* nested node group hierarchy. This type is small and can be passed around by value.
*
* A #DSocket can represent an input or an output socket. If the type of a socket is known at
* compile time is preferable to use #DInputSocket or #DOutputSocket instead.
*/
class DSocket {
protected:
const DTreeContext *context_ = nullptr;
const bNodeSocket *bsocket_ = nullptr;
public:
DSocket() = default;
DSocket(const DTreeContext *context, const bNodeSocket *socket);
DSocket(const DInputSocket &input_socket);
DSocket(const DOutputSocket &output_socket);
const DTreeContext *context() const;
const bNodeSocket *bsocket() const;
const bNodeSocket *operator->() const;
const bNodeSocket &operator*() const;
BLI_STRUCT_EQUALITY_OPERATORS_2(DSocket, context_, bsocket_)
operator bool() const;
uint64_t hash() const;
DNode node() const;
};
/** A (nullable) reference to an input socket and the context it is in. */
class DInputSocket : public DSocket {
public:
DInputSocket() = default;
DInputSocket(const DTreeContext *context, const bNodeSocket *socket);
explicit DInputSocket(const DSocket &base_socket);
DOutputSocket get_corresponding_group_node_output() const;
Vector<DOutputSocket, 4> get_corresponding_group_input_sockets() const;
/**
* Call `origin_fn` for every "real" origin socket. "Real" means that reroutes, muted nodes
* and node groups are handled by this function. Origin sockets are ones where a node gets its
* inputs from.
*/
void foreach_origin_socket(FunctionRef<void(DSocket)> origin_fn) const;
};
/** A (nullable) reference to an output socket and the context it is in. */
class DOutputSocket : public DSocket {
public:
DOutputSocket() = default;
DOutputSocket(const DTreeContext *context, const bNodeSocket *socket);
explicit DOutputSocket(const DSocket &base_socket);
DInputSocket get_corresponding_group_node_input() const;
DInputSocket get_active_corresponding_group_output_socket() const;
struct TargetSocketPathInfo {
/** All sockets on the path from the current to the final target sockets, excluding `this`. */
Vector<DSocket, 16> sockets;
};
using ForeachTargetSocketFn =
FunctionRef<void(DInputSocket, const TargetSocketPathInfo &path_info)>;
/**
* Calls `target_fn` for every "real" target socket. "Real" means that reroutes, muted nodes
* and node groups are handled by this function. Target sockets are on the nodes that use the
* value from this socket.
*/
void foreach_target_socket(ForeachTargetSocketFn target_fn) const;
private:
void foreach_target_socket(ForeachTargetSocketFn target_fn,
TargetSocketPathInfo &path_info) const;
};
class DerivedNodeTree {
private:
LinearAllocator<> allocator_;
DTreeContext *root_context_;
VectorSet<const bNodeTree *> used_btrees_;
public:
/**
* Construct a new derived node tree for a given root node tree. The generated derived node tree
* does not own the used node tree refs (so that those can be used by others as well). The caller
* has to make sure that the node tree refs added to #node_tree_refs live at least as long as the
* derived node tree.
*/
DerivedNodeTree(const bNodeTree &btree);
~DerivedNodeTree();
const DTreeContext &root_context() const;
Span<const bNodeTree *> used_btrees() const;
Geometry Nodes: initial scattering and geometry nodes This is the initial merge from the geometry-nodes branch. Nodes: * Attribute Math * Boolean * Edge Split * Float Compare * Object Info * Point Distribute * Point Instance * Random Attribute * Random Float * Subdivision Surface * Transform * Triangulate It includes the initial evaluation of geometry node groups in the Geometry Nodes modifier. Notes on the Generic attribute access API The API adds an indirection for attribute access. That has the following benefits: * Most code does not have to care about how an attribute is stored internally. This is mainly necessary, because we have to deal with "legacy" attributes such as vertex weights and attributes that are embedded into other structs such as vertex positions. * When reading from an attribute, we generally don't care what domain the attribute is stored on. So we want to abstract away the interpolation that that adapts attributes from one domain to another domain (this is not actually implemented yet). Other possible improvements for later iterations include: * Actually implement interpolation between domains. * Don't use inheritance for the different attribute types. A single class for read access and one for write access might be enough, because we know all the ways in which attributes are stored internally. We don't want more different internal structures in the future. On the contrary, ideally we can consolidate the different storage formats in the future to reduce the need for this indirection. * Remove the need for heap allocations when creating attribute accessors. It includes commits from: * Dalai Felinto * Hans Goudey * Jacques Lucke * Léo Depoix
2020-12-02 13:25:25 +01:00
/** Returns the active context for the node tree. The active context represents the node tree
* currently being edited. In most cases, that would be the top level node tree itself, but in
* the case where the user is editing the node tree of a node group, the active context would be
* a representation of the node tree of that node group. Note that the context also stores the
* group node that the user selected to edit the node tree, so the context fully represents a
* particular instance of the node group. */
const DTreeContext &active_context() const;
/**
* \return True when there is a link cycle. Unavailable sockets are ignored.
*/
Geometry Nodes: initial scattering and geometry nodes This is the initial merge from the geometry-nodes branch. Nodes: * Attribute Math * Boolean * Edge Split * Float Compare * Object Info * Point Distribute * Point Instance * Random Attribute * Random Float * Subdivision Surface * Transform * Triangulate It includes the initial evaluation of geometry node groups in the Geometry Nodes modifier. Notes on the Generic attribute access API The API adds an indirection for attribute access. That has the following benefits: * Most code does not have to care about how an attribute is stored internally. This is mainly necessary, because we have to deal with "legacy" attributes such as vertex weights and attributes that are embedded into other structs such as vertex positions. * When reading from an attribute, we generally don't care what domain the attribute is stored on. So we want to abstract away the interpolation that that adapts attributes from one domain to another domain (this is not actually implemented yet). Other possible improvements for later iterations include: * Actually implement interpolation between domains. * Don't use inheritance for the different attribute types. A single class for read access and one for write access might be enough, because we know all the ways in which attributes are stored internally. We don't want more different internal structures in the future. On the contrary, ideally we can consolidate the different storage formats in the future to reduce the need for this indirection. * Remove the need for heap allocations when creating attribute accessors. It includes commits from: * Dalai Felinto * Hans Goudey * Jacques Lucke * Léo Depoix
2020-12-02 13:25:25 +01:00
bool has_link_cycles() const;
bool has_undefined_nodes_or_sockets() const;
/** Calls the given callback on all nodes in the (possibly nested) derived node tree. */
void foreach_node(FunctionRef<void(DNode)> callback) const;
/** Generates a graph in dot format. The generated graph has all node groups inlined. */
std::string to_dot() const;
private:
DTreeContext &construct_context_recursively(DTreeContext *parent_context,
const bNode *parent_node,
const bNodeTree &btree,
const bNodeInstanceKey instance_key);
void destruct_context_recursively(DTreeContext *context);
void foreach_node_in_context_recursive(const DTreeContext &context,
FunctionRef<void(DNode)> callback) const;
};
Geometry Nodes: initial scattering and geometry nodes This is the initial merge from the geometry-nodes branch. Nodes: * Attribute Math * Boolean * Edge Split * Float Compare * Object Info * Point Distribute * Point Instance * Random Attribute * Random Float * Subdivision Surface * Transform * Triangulate It includes the initial evaluation of geometry node groups in the Geometry Nodes modifier. Notes on the Generic attribute access API The API adds an indirection for attribute access. That has the following benefits: * Most code does not have to care about how an attribute is stored internally. This is mainly necessary, because we have to deal with "legacy" attributes such as vertex weights and attributes that are embedded into other structs such as vertex positions. * When reading from an attribute, we generally don't care what domain the attribute is stored on. So we want to abstract away the interpolation that that adapts attributes from one domain to another domain (this is not actually implemented yet). Other possible improvements for later iterations include: * Actually implement interpolation between domains. * Don't use inheritance for the different attribute types. A single class for read access and one for write access might be enough, because we know all the ways in which attributes are stored internally. We don't want more different internal structures in the future. On the contrary, ideally we can consolidate the different storage formats in the future to reduce the need for this indirection. * Remove the need for heap allocations when creating attribute accessors. It includes commits from: * Dalai Felinto * Hans Goudey * Jacques Lucke * Léo Depoix
2020-12-02 13:25:25 +01:00
namespace derived_node_tree_types {
using nodes::DerivedNodeTree;
using nodes::DInputSocket;
using nodes::DNode;
using nodes::DOutputSocket;
using nodes::DSocket;
using nodes::DTreeContext;
} // namespace derived_node_tree_types
Geometry Nodes: initial scattering and geometry nodes This is the initial merge from the geometry-nodes branch. Nodes: * Attribute Math * Boolean * Edge Split * Float Compare * Object Info * Point Distribute * Point Instance * Random Attribute * Random Float * Subdivision Surface * Transform * Triangulate It includes the initial evaluation of geometry node groups in the Geometry Nodes modifier. Notes on the Generic attribute access API The API adds an indirection for attribute access. That has the following benefits: * Most code does not have to care about how an attribute is stored internally. This is mainly necessary, because we have to deal with "legacy" attributes such as vertex weights and attributes that are embedded into other structs such as vertex positions. * When reading from an attribute, we generally don't care what domain the attribute is stored on. So we want to abstract away the interpolation that that adapts attributes from one domain to another domain (this is not actually implemented yet). Other possible improvements for later iterations include: * Actually implement interpolation between domains. * Don't use inheritance for the different attribute types. A single class for read access and one for write access might be enough, because we know all the ways in which attributes are stored internally. We don't want more different internal structures in the future. On the contrary, ideally we can consolidate the different storage formats in the future to reduce the need for this indirection. * Remove the need for heap allocations when creating attribute accessors. It includes commits from: * Dalai Felinto * Hans Goudey * Jacques Lucke * Léo Depoix
2020-12-02 13:25:25 +01:00
2021-10-05 11:10:25 +11:00
/* -------------------------------------------------------------------- */
/** \name #DTreeContext Inline Methods
* \{ */
inline const bNodeTree &DTreeContext::btree() const
{
return *btree_;
}
inline const DTreeContext *DTreeContext::parent_context() const
{
return parent_context_;
}
inline const bNode *DTreeContext::parent_node() const
{
return parent_node_;
}
inline bNodeInstanceKey DTreeContext::instance_key() const
{
return instance_key_;
}
inline const DTreeContext *DTreeContext::child_context(const bNode &node) const
{
return children_.lookup_default(&node, nullptr);
}
inline const DerivedNodeTree &DTreeContext::derived_tree() const
{
return *derived_tree_;
}
inline bool DTreeContext::is_root() const
{
return parent_context_ == nullptr;
}
2021-10-05 11:10:25 +11:00
/** \} */
2021-10-05 11:10:25 +11:00
/* -------------------------------------------------------------------- */
/** \name #DNode Inline Methods
* \{ */
inline DNode::DNode(const DTreeContext *context, const bNode *bnode)
: context_(context), bnode_(bnode)
{
BLI_assert(bnode == nullptr || bnode->runtime->owner_tree == &context->btree());
}
inline const DTreeContext *DNode::context() const
{
return context_;
}
inline const bNode *DNode::bnode() const
{
return bnode_;
}
inline DNode::operator bool() const
{
return bnode_ != nullptr;
}
inline const bNode *DNode::operator->() const
{
return bnode_;
}
inline const bNode &DNode::operator*() const
{
BLI_assert(bnode_ != nullptr);
return *bnode_;
}
inline uint64_t DNode::hash() const
{
return get_default_hash(context_, bnode_);
}
inline DInputSocket DNode::input(int index) const
{
return {context_, &bnode_->input_socket(index)};
}
inline DOutputSocket DNode::output(int index) const
{
return {context_, &bnode_->output_socket(index)};
}
inline DInputSocket DNode::input_by_identifier(StringRef identifier) const
{
return {context_, bnode_->input_by_identifier(identifier)};
}
inline DOutputSocket DNode::output_by_identifier(StringRef identifier) const
{
return {context_, bnode_->output_by_identifier(identifier)};
}
2021-10-05 11:10:25 +11:00
/** \} */
2021-10-05 11:10:25 +11:00
/* -------------------------------------------------------------------- */
/** \name #DSocket Inline Methods
* \{ */
inline DSocket::DSocket(const DTreeContext *context, const bNodeSocket *bsocket)
: context_(context), bsocket_(bsocket)
{
BLI_assert(bsocket == nullptr ||
bsocket->runtime->owner_node->runtime->owner_tree == &context->btree());
}
inline DSocket::DSocket(const DInputSocket &input_socket)
: DSocket(input_socket.context_, input_socket.bsocket_)
{
}
inline DSocket::DSocket(const DOutputSocket &output_socket)
: DSocket(output_socket.context_, output_socket.bsocket_)
{
}
inline const DTreeContext *DSocket::context() const
{
return context_;
}
inline const bNodeSocket *DSocket::bsocket() const
{
return bsocket_;
}
inline DSocket::operator bool() const
{
return bsocket_ != nullptr;
}
inline const bNodeSocket *DSocket::operator->() const
{
return bsocket_;
}
inline const bNodeSocket &DSocket::operator*() const
{
BLI_assert(bsocket_ != nullptr);
return *bsocket_;
}
inline uint64_t DSocket::hash() const
{
return get_default_hash(context_, bsocket_);
}
inline DNode DSocket::node() const
{
BLI_assert(bsocket_ != nullptr);
return {context_, bsocket_->runtime->owner_node};
}
2021-10-05 11:10:25 +11:00
/** \} */
2021-10-05 11:10:25 +11:00
/* -------------------------------------------------------------------- */
/** \name #DInputSocket Inline Methods
* \{ */
inline DInputSocket::DInputSocket(const DTreeContext *context, const bNodeSocket *bsocket)
: DSocket(context, bsocket)
{
}
inline DInputSocket::DInputSocket(const DSocket &base_socket) : DSocket(base_socket)
{
BLI_assert(base_socket->is_input());
}
2021-10-05 11:10:25 +11:00
/** \} */
2021-10-05 11:10:25 +11:00
/* -------------------------------------------------------------------- */
/** \name #DOutputSocket Inline Methods
* \{ */
Geometry Nodes: initial scattering and geometry nodes This is the initial merge from the geometry-nodes branch. Nodes: * Attribute Math * Boolean * Edge Split * Float Compare * Object Info * Point Distribute * Point Instance * Random Attribute * Random Float * Subdivision Surface * Transform * Triangulate It includes the initial evaluation of geometry node groups in the Geometry Nodes modifier. Notes on the Generic attribute access API The API adds an indirection for attribute access. That has the following benefits: * Most code does not have to care about how an attribute is stored internally. This is mainly necessary, because we have to deal with "legacy" attributes such as vertex weights and attributes that are embedded into other structs such as vertex positions. * When reading from an attribute, we generally don't care what domain the attribute is stored on. So we want to abstract away the interpolation that that adapts attributes from one domain to another domain (this is not actually implemented yet). Other possible improvements for later iterations include: * Actually implement interpolation between domains. * Don't use inheritance for the different attribute types. A single class for read access and one for write access might be enough, because we know all the ways in which attributes are stored internally. We don't want more different internal structures in the future. On the contrary, ideally we can consolidate the different storage formats in the future to reduce the need for this indirection. * Remove the need for heap allocations when creating attribute accessors. It includes commits from: * Dalai Felinto * Hans Goudey * Jacques Lucke * Léo Depoix
2020-12-02 13:25:25 +01:00
inline DOutputSocket::DOutputSocket(const DTreeContext *context, const bNodeSocket *bsocket)
: DSocket(context, bsocket)
Geometry Nodes: initial scattering and geometry nodes This is the initial merge from the geometry-nodes branch. Nodes: * Attribute Math * Boolean * Edge Split * Float Compare * Object Info * Point Distribute * Point Instance * Random Attribute * Random Float * Subdivision Surface * Transform * Triangulate It includes the initial evaluation of geometry node groups in the Geometry Nodes modifier. Notes on the Generic attribute access API The API adds an indirection for attribute access. That has the following benefits: * Most code does not have to care about how an attribute is stored internally. This is mainly necessary, because we have to deal with "legacy" attributes such as vertex weights and attributes that are embedded into other structs such as vertex positions. * When reading from an attribute, we generally don't care what domain the attribute is stored on. So we want to abstract away the interpolation that that adapts attributes from one domain to another domain (this is not actually implemented yet). Other possible improvements for later iterations include: * Actually implement interpolation between domains. * Don't use inheritance for the different attribute types. A single class for read access and one for write access might be enough, because we know all the ways in which attributes are stored internally. We don't want more different internal structures in the future. On the contrary, ideally we can consolidate the different storage formats in the future to reduce the need for this indirection. * Remove the need for heap allocations when creating attribute accessors. It includes commits from: * Dalai Felinto * Hans Goudey * Jacques Lucke * Léo Depoix
2020-12-02 13:25:25 +01:00
{
}
inline DOutputSocket::DOutputSocket(const DSocket &base_socket) : DSocket(base_socket)
{
BLI_assert(base_socket->is_output());
}
2021-10-05 11:10:25 +11:00
/** \} */
2021-10-05 11:10:25 +11:00
/* -------------------------------------------------------------------- */
/** \name #DerivedNodeTree Inline Methods
* \{ */
inline const DTreeContext &DerivedNodeTree::root_context() const
{
return *root_context_;
}
inline Span<const bNodeTree *> DerivedNodeTree::used_btrees() const
Geometry Nodes: initial scattering and geometry nodes This is the initial merge from the geometry-nodes branch. Nodes: * Attribute Math * Boolean * Edge Split * Float Compare * Object Info * Point Distribute * Point Instance * Random Attribute * Random Float * Subdivision Surface * Transform * Triangulate It includes the initial evaluation of geometry node groups in the Geometry Nodes modifier. Notes on the Generic attribute access API The API adds an indirection for attribute access. That has the following benefits: * Most code does not have to care about how an attribute is stored internally. This is mainly necessary, because we have to deal with "legacy" attributes such as vertex weights and attributes that are embedded into other structs such as vertex positions. * When reading from an attribute, we generally don't care what domain the attribute is stored on. So we want to abstract away the interpolation that that adapts attributes from one domain to another domain (this is not actually implemented yet). Other possible improvements for later iterations include: * Actually implement interpolation between domains. * Don't use inheritance for the different attribute types. A single class for read access and one for write access might be enough, because we know all the ways in which attributes are stored internally. We don't want more different internal structures in the future. On the contrary, ideally we can consolidate the different storage formats in the future to reduce the need for this indirection. * Remove the need for heap allocations when creating attribute accessors. It includes commits from: * Dalai Felinto * Hans Goudey * Jacques Lucke * Léo Depoix
2020-12-02 13:25:25 +01:00
{
return used_btrees_;
Geometry Nodes: initial scattering and geometry nodes This is the initial merge from the geometry-nodes branch. Nodes: * Attribute Math * Boolean * Edge Split * Float Compare * Object Info * Point Distribute * Point Instance * Random Attribute * Random Float * Subdivision Surface * Transform * Triangulate It includes the initial evaluation of geometry node groups in the Geometry Nodes modifier. Notes on the Generic attribute access API The API adds an indirection for attribute access. That has the following benefits: * Most code does not have to care about how an attribute is stored internally. This is mainly necessary, because we have to deal with "legacy" attributes such as vertex weights and attributes that are embedded into other structs such as vertex positions. * When reading from an attribute, we generally don't care what domain the attribute is stored on. So we want to abstract away the interpolation that that adapts attributes from one domain to another domain (this is not actually implemented yet). Other possible improvements for later iterations include: * Actually implement interpolation between domains. * Don't use inheritance for the different attribute types. A single class for read access and one for write access might be enough, because we know all the ways in which attributes are stored internally. We don't want more different internal structures in the future. On the contrary, ideally we can consolidate the different storage formats in the future to reduce the need for this indirection. * Remove the need for heap allocations when creating attribute accessors. It includes commits from: * Dalai Felinto * Hans Goudey * Jacques Lucke * Léo Depoix
2020-12-02 13:25:25 +01:00
}
2021-10-05 11:10:25 +11:00
/** \} */
} // namespace blender::nodes