The convex hull tests included a reference AABB-fitting function for
comparison which was used to validate the optimized implementation.
This wasn't great as it depended on matching exact return values and
didn't test the logic of AABB-fitting worked usefully.
Replace this with a more general test that creates random polygons with
known bounds, apply a random rotation & translation, then use
AABB-fitting to un-rotate the points, passing when the bounds are no
larger than the size of the generated input.
Details:
- Make BLI_convexhull_aabb_fit_hull_2d a static function again as it was
only exposed for tests. Use BLI_convexhull_aabb_fit_points_2d instead.
- Remove brute force reference implementation from tests,
moving this to an assertion within convexhull_2d
(disabled by default since it's quite slow).
The lack of these functions in the "single trivial value" and "sliced
GVArray" implementations caused some code to call fack to the base
class functions. Those are much slower since they involve a virtual
function call per element. For example, this changed the runtime of
creating a new boolean attribute set to "true" on one million faces
from 3.4 ms to 0.35 ms.
Pull Request: https://projects.blender.org/blender/blender/pulls/118161
Implements the design from #116067.
The socket type is called "Matrix" but it is often referred to as "Transform"
when that's what it is semantically. The attribute type is "4x4 Matrix" since
that's a lower level choice. Currently matrix sockets are always passed
around internally as `float4x4`, but that can be optimized in the future
when smaller types would give the same behavior.
A new "Matrix" utilities category has the following set of initial nodes"
- **Combine Transform**
- **Separate Transform**
- **Multiply Matrices**
- **Transform Direction**
- **Transform Vector**
- **Invert Matrix**
- **Transpose Matrix**
The nodes and socket type are behind an experimental flag for now,
which will give us time to make sure it's the right set of initial nodes.
The viewer node overlay doesn't support matrices-- they aren't supported
for rendering in general. They also aren't supported in the modifier interface
currently. But they are supported in the spreadsheet, where the value is
displayed in a tooltip.
Pull Request: https://projects.blender.org/blender/blender/pulls/116166
Currently most of the data stored in `wmWindowManager` is runtime
data not saved to files. It's confusing that it's declared in DNA then. That
also prevents us from using C++ features. This commit adds an initial
runtime struct. Moving data there can be done as a separate step.
Initially I wanted to look at moving the `ReportList` system to C++.
The runtime struct has to be defined in the blenkernel module because
the members are (will be) used there in a few places.
Pull Request: https://projects.blender.org/blender/blender/pulls/118157
The Translate node crashed for single inputs in the Full Frame
compositor, that's because it always assumed image inputs. So fix this
by handling the single value case.
The ID remapper code was already largely defined in a CPP struct
(IDRemapper). Make this an actual class, and remove the C API wrapper
around.
This makes the code cleaner, easier to follow, and easier to extend or
modify in the future.
Pull Request: https://projects.blender.org/blender/blender/pulls/118146
Running background mode now behaves as if the "-noaudio" was passed in.
The -setaudio command now has a "Default" option which can be used
in the rare cases audio playback is desired in background mode. e.g.
blender --background -setaudio Default
Ref !118192
Images do not update in the compositor and other places when their file
path changes or their generated options change. This is because the
compositor relies on depsgraph updates, which weren't tagged in this
case. Fix this by tagging appropriately in RNA.
Pull Request: https://projects.blender.org/blender/blender/pulls/118187
`BKE_library_id_can_use_filter_id` was missing the ShapeKey type for
geometry IDs (meshes, legacy curves, and lattices), leading to the
remapping code to fail to do its job when deleting the ShapeKey only.
This PR prepares GPv3 for calculating falloff factors when working in multi frame editing mode.
The falloff popover is added to the UI.
A `float multi_frame_falloff` is added to `MutableDrawingInfo`. This is a factor from 0.0 to 1.0f.
A `retrieve_editable_drawings_with_falloff()` is added to the utility functions, which retrieves the falloff factor for each drawing when
multi frame falloff is enabled.
To avoid a copy, the return type of `retrieve_editable_drawings()` and friends is changed from
`Array<MutableDrawingInfo>` to `Vector<MutableDrawingInfo>`.
Pull Request: https://projects.blender.org/blender/blender/pulls/118108
Previously the bounding box was calculated for each rotation,
even though the loop would early exit if the area exceeded the best
known area, this meant the areas of the bounds was checked for
each point too. This scaled poorly, close to O(n^2).
Replace this with simple logic that keeps track of the indices
at the bounds, advancing them forward for each rotation.
This is around ~140x faster for a hull with 10k points on my system.
Note that there is potential for other improvements although
the cases where this new method doesn't perform well are quite specific
and faster than checking all points, see code-comments for details.
Add new descriptions to the `PROP_HIDDEN`, `PROP_SKIP_SAVE`, and
`PROP_LIB_EXCEPTION` flags. For the former two, these are more
explanatory than the previous descriptions. For the latter it's the
first ever description.
Personally I don't think "Do not use ghost values" was particularly
helpful, as the concept of "ghost values" isn't actually that well known
(compared to generally how many people use `bpy`). Granted, it's
described at [1], but IMO if the term "ghost values" is to be kept here,
it should include a link to that page.
[1]: https://docs.blender.org/api/current/bpy.types.bpy_struct.html
Pull Request: https://projects.blender.org/blender/blender/pulls/118036
These flags (like `'HIDDEN', 'SKIP_SAVE', etc.) mean the same thing when
used on enums and non-enum properties. Now they actually use the same
description definition in the source code, ensuring that any
improvements will apply to both.
No changes to the actual messages. Just using them for both enums and
non-enum property flags.
Pull Request: https://projects.blender.org/blender/blender/pulls/118036
The new keyframing functions introduced in #113504
didn't call the functions to decompose the NLA stack.
In practice this meant that when inserting keys into strip
that is under an additive strip, it would take the result of the additive Strip and
bake it back into the base. This would double the transform.
The fix is to call `BKE_animsys_nla_remap_keyframe_values`.
Unfortunately to do so, I had to pass through a few more
arguments to the keyframing functions.
Also adds unit tests to cover the caused bug.
Pull Request: https://projects.blender.org/blender/blender/pulls/118053
