Speed up round-trip without changing the wire format

[gonzalocasas]

Three small-ish fixes found by profiling large mesh/graph round-trips:

• Move plugin discovery up (_ensure_serializers) out of the per-element any_to_pb/any_from_pb to top-level.
• Gate the legacy Any-unpacking .Is() probes in _deserialize_any
• In _fill_attribute_columns drop the redundant per-column sort (indices are already ascending) and the throwaway list(range(count)) dense check; in mesh_to_pb read x/y/z straight from the vertex dict and fill packed fields with a single extend instead of one proto call per vertex/face.

Mesh serialize ~2x faster, graph round-trip ~1.5x faster. Wire output is unchanged; all 71 tests pass.

What type of change is this?

• Bug fix in a backwards-compatible manner.
• New feature in a backwards-compatible manner.
• Breaking change: bug fix or new feature that involve incompatible API changes.
• Other (e.g. doc update, configuration, etc)

Checklist

Put an x in the boxes that apply. You can also fill these out after creating the PR. If you're unsure about any of them, don't hesitate to ask. We're here to help! This is simply a reminder of what we are going to look for before merging your code.

• I added a line to the CHANGELOG.md file in the Unreleased section under the most fitting heading (e.g. Added, Changed, Removed).
• I ran all tests on my computer and it's all green (i.e. invoke test).
• I ran lint on my computer and there are no errors (i.e. invoke lint).
• I added new functions/classes and made them available on a second-level import, e.g. compas.datastructures.Mesh.
• I have added tests that prove my fix is effective or that my feature works.
• I have added necessary documentation (if appropriate)

[Copilot]

Pull request overview

This PR optimizes compas_pb mesh/graph serialization and deserialization hot paths to reduce round-trip time while aiming to keep the protobuf wire format unchanged. It does so by moving one-time setup work out of per-element loops and reducing per-element protobuf API calls, with added regression tests around mesh defaults and attribute column behavior.

Changes:

• Moves plugin discovery (_ensure_serializers) out of per-element any_to_pb/any_from_pb and into top-level serialize/deserialize entry points.
• Optimizes _deserialize_any, _fill_attribute_columns, and mesh_to_pb to reduce redundant checks, sorting, and per-element proto calls.
• Adds mesh round-trip tests covering default coordinates and sparse/dense/non-float attribute columns; documents behavioral implications in the changelog.

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 2 comments.

File	Description
tests/test_mesh.py	Adds regression tests for mesh vertex default coordinates and attribute/default preservation across round-trips.
src/compas_pb/core.py	Moves serializer/plugin discovery to top-level entry points and gates legacy Any-unpacking checks.
src/compas_pb/conversions.py	Optimizes attribute column building and mesh serialization (packed coordinates + batched extends).
CHANGELOG.md	Notes the performance work and calls out the vertex_coordinates() override implication for subclasses.

[Copilot]

Pull request overview

Copilot reviewed 4 out of 4 changed files in this pull request and generated no new comments.

Suppressed comments (3)

src/compas_pb/core.py:110

• any_to_pb only needs serializers, but it currently triggers plugin discovery when either serializers OR deserializers are missing. This can cause unnecessary plugin discovery / imports in serialization-only call paths and makes the condition stricter than required.

    if not SerializerRegistry._SERIALIZERS or not SerializerRegistry._DESERIALIZERS:
        _ensure_serializers()

src/compas_pb/core.py:142

• any_from_pb only needs deserializers, but it currently triggers plugin discovery based on serializer availability too. This adds unnecessary work and side effects when only deserialization is needed.

    if not SerializerRegistry._SERIALIZERS or not SerializerRegistry._DESERIALIZERS:
        _ensure_serializers()

src/compas_pb/conversions.py:421

• mesh_to_pb now iterates mesh.vertex.items() directly. This bypasses the public mesh.vertices() iterator used previously and can change vertex ordering/selection for subclasses or implementations where vertices() is overridden or ordered differently, which can also change the serialized output ordering.

    for index, (key, attr) in enumerate(mesh.vertex.items()):
        coords.append(attr.get("x", default_x))
        coords.append(attr.get("y", default_y))
        coords.append(attr.get("z", default_z))
        index_map[key] = index

[WeiTing1991]

Amazing! LGTM. Thanks for performance improvement. 🚀

[chenkasirer]

LGTM!