Publish: Remove dataclasses from publish plugins

[iLLiCiTiT]

Changelog Description

Use classic classes instead of dataclasses in publish plugins.

Additional info

Dataclasses are broken in case they are in publish plugin files which are discovered using pyblish discovery logic (it works if AYON's discovery is used...).

Replaced dataclasses with just classes -> added __init__ and similar "work-like" logic.

Testing notes:

1. Collect addons, Extract Thumbnail, Extract USD Layer plugins should work if pyblish publish logic is triggered instead of AYON's logic.

Resolves ynput/ayon-nuke#366

[moonyuet]

publish-report-260915-17-43.json

Tested successfully with maya USD publish and review.

[BigRoy]

Looks ok to me - note that the issue only occurs if the file has from __future__ import annotations at the top. So some of these fixes aren't entirely needed, but it's maybe a safe preparation.

[Copilot]

🟡 Changes recommended

Two moderate compatibility and maintainability issues remain unresolved.

Get a fresh assessment by requesting another Copilot review.

Pull request overview

This pull request replaces dataclasses in publish plugin helpers with classic classes for Pyblish compatibility.

Changes:

• Added explicit constructors for contribution and addon models.
• Reimplemented thumbnail initialization and dictionary conversion.
• Updated related type annotations.

File summaries

File	Review summary
client/ayon_core/plugins/publish/extract_usd_layer_contributions.py	Classic contribution classes added; moderate issue with preserving positional constructor compatibility (2 votes).
client/ayon_core/plugins/publish/extract_thumbnail.py	Manual thumbnail initialization added; moderate issue with duplicated constructor-field whitelist (1 vote).
client/ayon_core/plugins/publish/collect_addons.py	AddonInfo converted to a classic class; no final comment.

Review details

Suppressed comments (1)

client/ayon_core/plugins/publish/extract_thumbnail.py:93

• This whitelist duplicates every constructor field, so adding a setting to __init__ without also updating this set will silently drop that setting from from_dict; the previous dataclasses.fields(cls) implementation could not drift this way. Use a single source of truth for the accepted constructor fields (for example, derive the names from the constructor) instead of maintaining this second list.

        field_names = {
            "integrate_thumbnail",
            "target_size",
            "duration_split",
            "oiiotool_defaults",

• Files reviewed: 3/3 changed files
• Comments generated: 1
• Review effort level: Lite

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[BigRoy]

@iLLiCiTiT can we merge this and get to a release sometime soon?