Because we use virtual classes (and for other reasons), we had to do a
small allocation when simply retrieving the data type and domain of an
existing attribute. This happened quite a lot actually-- to determine
these values for result attributes.
This patch adds a simple function to retrieve this meta data without
building the virtual array. This should lower the overhead of every
attribute node, though the difference probably won't be noticible
unless a tree has very many nodes.
Differential Revision: https://developer.blender.org/D11047
A virtual array is a data structure that is similar to a normal array
in that its elements can be accessed by an index. However, a virtual
array does not have to be a contiguous array internally. Instead, its
elements can be layed out arbitrarily while element access happens
through a virtual function call. However, the virtual array data
structures are designed so that the virtual function call can be avoided
in cases where it could become a bottleneck.
Most commonly, a virtual array is backed by an actual array/span or
is a single value internally, that is the same for every index.
Besides those, there are many more specialized virtual arrays like the
ones that provides vertex positions based on the `MVert` struct or
vertex group weights.
Not all attributes used by geometry nodes are stored in simple contiguous
arrays. To provide uniform access to all kinds of attributes, the attribute
API has to provide virtual array functionality that hides the implementation
details of attributes.
Before this refactor, the attribute API provided its own virtual array
implementation as part of the `ReadAttribute` and `WriteAttribute` types.
That resulted in unnecessary code duplication with the virtual array system.
Even worse, it bound many algorithms used by geometry nodes to the specifics
of the attribute API, even though they could also use different data sources
(such as data from sockets, default values, later results of expressions, ...).
This refactor removes the `ReadAttribute` and `WriteAttribute` types and
replaces them with `GVArray` and `GVMutableArray` respectively. The `GV`
stands for "generic virtual". The "generic" means that the data type contained
in those virtual arrays is only known at run-time. There are the corresponding
statically typed types `VArray<T>` and `VMutableArray<T>` as well.
No regressions are expected from this refactor. It does come with one
improvement for users. The attribute API can convert the data type
on write now. This is especially useful when writing to builtin attributes
like `material_index` with e.g. the Attribute Math node (which usually
just writes to float attributes, while `material_index` is an integer attribute).
Differential Revision: https://developer.blender.org/D10994
Previously only attributes of "real" geometry were displayed in
attribute search. This commit adds code to look through attributes
on instances and add those to the search drop-down too.
This required implementing the same sort of recursive traversal as
the realize instances code. The situation is a bit different though,
this can return early and doesn't need to keep track of transforms.
I added a limit so that it doesn't look through the attributes of
too many instanced geometry sets. I think this is important, since
this isn't a trivial operation and it could potentially happen for
every node in a large node tree. Currently the limit is set at 8
geometry sets, which I expect will be enough, since the set of
attributes is mostly not very unique anyway.
Fixes T86282
Diffrential Revision: https://developer.blender.org/D10919
Now that we have `ATTR_DOMAIN_AUTO`, it makes sense to use it to skip
automatic domain interpolation. This can make code that depends on the
non-interpolated domain of the attribute a bit simpler.
This patch renames two domains:
* `Polygon` -> `Face`
* `Corner` -> `Face Corner`
For the change from `polygon` to `face` I did a "deep rename" where I updated
all (most?) cases where we refere to the attribute domain in code as well.
The change from `corner` to `face corner` is only a ui change. I did not see
a real need to update all code the code for that. It does not seem to improve
the code, more on the contrary.
Ref T86818.
Differential Revision: https://developer.blender.org/D10803
After further thought, the implementation of the "normal" attribute
from D10541 is not the best approach to expose this data, mainly
because it blindly copied existing design rather than using the
best method in the context of the generalized attribute system.
In Blender, vertex normals are simply a cache of the average normals
from the surrounding / connected faces. Because we have automatic
interpolation between domains already, we don't need a special
`vertex_normal` attribute for this case, we can just let the
generalized interpolation do the hard work where necessary,
simplifying the set of built-in attributes to only include the
`normal` attribute from faces.
The fact that vertex normals are just a cache also raised another
issue, because the cache could be dirty, so mutex locks were
necessary to calculate normals. That isn't necessarily a problem,
but it's nice to avoid where possible.
Another downside of the current attribute naming is that after the
point distribute node there would be two normal attributes.
This commit reverts the `vertex_normal` attribute so that
it can be replaced by the implementation in D10677.
Differential Revision: https://developer.blender.org/D10676
This was meant to be part of rB9ce950daabbf, but the change dropped from
the set at some point in the process of updating and committing.
Sorry for the noise.
This attribute exposes mesh vertex normals as a `vertex_normal`
attribute for use with nodes. Since the normal vector stored in
vertices is only a cache of data computable from the surrounding faces,
the attribute is read-only. A proper error message for attempting to
write this attribute is part of T85749. A write-only normal attribute
will likely come later, most likely called `corner_normal`.
The normals are recomputed before reading if they are marked dirty.
This involves const write-access to the mesh, protected by the mutex
stored in `Mesh_Runtime`. This is essential for correct behavior after
nodes like "Edge Split" or nodes that adjust the position attribute.
Ref T84297, T85880, T86206
Differential Revision: https://developer.blender.org/D10541
The crash happened when the density in the Point Distribute node was
above zero but so small, that no point was generated. In this case, there
was a point cloud component, but the point cloud was empty, making some
attributes unavailable.
One could also make more attributes available in this case, but that can
be done separately if necessary.
This allows accessing attribute meta data like domain and data type
without having to create a `ReadAttribute`. I kept the `attribute_names`
method for now to keep the patch more self contained.
Differential Revision: https://developer.blender.org/D10511
This allows accessing per-point attributes on the corner domain,
which can be useful e.g. when adding per-point displacement
to per-corner uv coordinates.
Also it is required to make an upcoming patch work well, that
makes the Point Distribute node use density weights per corner
instead of per point, giving the user more precise control over
the distribution.
This makes vertex colors available in geometry nodes similar to
how uvs are available. They can be used using the attribute system.
Vertex colors are stored per corner (as are uvs, but not like vertex weights).
Ref T84297.
Differential Revision: https://developer.blender.org/D10454
The `material_index` attribute can adjust which material in the list
will be applied to each face of the mesh. There are two new things
about this attribute that haven't been exposed by the attribute API yet.
Each comes with limitations:
1. Integer data type: Most attribute nodes are currently written to use
float data types. This means that they can't write to this attribute
because they can't change the type of a built-in attribute.
2. Polygon domain: This is our first attribute using the polygon domain,
meaning until some of the interpolations are implemented, some
operations may not work as expected.
Currently the two nodes that work with this attribute are Attribute Fill
and Attribute Randomize.
Differential Revision: https://developer.blender.org/D10444
This makes it so that normals are tagged dirty whenever the position
attribute is requested for writing. This seems like a good default. If the
calling code is aware of normals, it could untag normals when they are
not changed by the operation.
Differential Revision: https://developer.blender.org/D10397
Fixes an issue in a node setup with the point separate node, where muting
a node that does nothing breaks the operation, resulting in the point
separate not copying the position attrbute to either result.
The fix is straightfoward, it looks just like a typo.
Differential Revision: https://developer.blender.org/D10379
This patch adds support for accessing corner attributes on the point domain.
The immediate benefit of this is that now (interpolated) uv coordinates are
available on points without having to use the Point Distribute node.
This is also very useful for parts of T84297, because once we have vertex
colors, those will also be available on points, even though they are stored
per corner.
Differential Revision: https://developer.blender.org/D10305
Goals:
* Clarify the distinction between builtin and other attributes at the code level.
* Reduce number of places that need to be modified to add more builtin attributes.
* Reduce number of virtual methods that need to be implemented by e.g. `MeshComponent`.
To achieve these goals, this patch implements the concept of "attribute providers".
An attribute provider knows how to give access to attributes on a geometry component.
Each geometry component can have multiple attribute providers, whereby each provider
manages an different set of attributes.
The separation of builtin and other attributes is now done at the attribute provider level.
There are two types of attribute providers. One for builtin attributes and one for all others.
This refactor also helps with T84297.
Differential Revision: https://developer.blender.org/D10341
The issue was that the mesh shared its vertex weights with the
original mesh (to reduce memory consumption). The solution is
to make a local copy of the vertex weights in this case.
This commit adds the ability to provide a default value to
`attribute_try_get_for_output` and uses it for the `Point Scale` node,
which is important because the node uses multiplication.
The idea is to keep "name-specific" functionality in nodes rather than in
the attribute API, otherwise the complexity will be hard to keep track of.
So this fix doesn't apply to the Attribute Vector Math node, but hopfully
that is okay since that's now a lower level node for this purpose anyway.
Differential Revision: https://developer.blender.org/D10115
Note that uv layers still can't be accessed with nodes, because those
only access attributes on the point domain currently, while uv data
is stored per corner. Implicit domain conversion hasn't been
implemented yet.
This fixes the behavior of some nodes when the same attribute
name is used for input and output. If both attributes have a
different type, they can't exist at the same time. Therefore,
the input attribute has to be removed in order to create the
output attribute.
Previously, the input attribute was remove before it was used
in any computations. Now, the output is written to a temporary
buffer and only later saved in the geometry component. This
allows both attributes to coexist within the node.
The temporary attribute is only create when necessary. The
normal case without name collisions still works the same
as before.
Differential Revision: https://developer.blender.org/D10109
Ref T83793.
Previously, the span returned by `WriteAttribute`s might not contain the
current value of the attribute for performance reasons. To avoid some
bugs, the span now always contains the old values (they might have to
be copied over from the internal storage, dependending on how the
attribute is stored).
The old behavior is still available with the `get_span_for_write_only`
method. The span that it returns might not contain the current
attribute values. Therefore, it should only be used when you want
to overwrite an attribute without looking at the old values.
Currently, the random attribute node doesn't work well for most
workflows because for any change in the input data it outputs
completely different results.
This patch adds an implicit seed attribute input to the node, referred
to by "id". The attribute is hashed for each element using the CPPType
system's hash method, meaning the attribute can have any data type.
Supporting any data type is also important so any attribute can be
copied into the "id" attribute and used as a seed.
The "id" attribute is an example of a "reserved name" attribute,
meaning attributes with this name can be used implicitly by nodes like
the random attribute node. Although it makes it a bit more difficult
to dig deeper, using the name implicitly rather than exposing it as an
input should make the system more accessible and predictable.
Differential Revision: https://developer.blender.org/D9832
This adds a boolean attribute and custom data type, to be used in the
point separate node. It also adds it as supported data types in the
random attribute and attribute fill nodes.
There are more clever ways of storing a boolean attribute that make
more sense in certain situations-- sets, bitfields, and others, this
commit keeps it simple, saving those changes for when there is a proper
use case for them. In any case, we will still probably always want the
idea of a boolean attribute.
Differential Revision: https://developer.blender.org/D9818
This is a non-functional change. The functionality introduced in this commit
is not used in master yet. It is used by nodes that are being developed in
other branches though.
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
