[6.x] Asset transformers

[riasvdv]

See https://github.com/craftcms/cms/blob/feature/asset-transformers/docs/asset-transformers.md

[github-actions]

📚 Storybook previews

@craftcms/ui — open Storybook

No changed components detected in this Storybook.

resources/js — open Storybook

No changed components detected in this Storybook.

[timkelty]

FYI: I only reviewed the spec doc, not the implementation.

I’m not sold on some of the terminology:

• processor reads like a new synonym for transformer. transform and transformer may carry baggage, but they’re established terms and clearly describe the relationship.

I understand why the model needs three levels, but names rooted in the existing transform terminology feel clearer:

• AssetTransform / AssetTransformRequest
• AssetTransformer
• AssetTransformDriver

rendition feels awkward. It doesn’t appear in the public API and only occurs in production code in an error message. What about transform result, output, or even variant (Commerce Variants)?

operations and operationFields feel awkward. options / optionFields, settings, config feel more natural. “Operations” also seems to suggest discrete steps applied in succession, but these appear to be parameters that are normalized and passed to the driver.

[github-actions]

This PR had conflicts with 6.x. Claude attempted an automatic resolution and pushed a merge commit. Please review the merge commit carefully before approving/merging.

Claude's response

Merge resolved, committed as 8ada12f, and pushed.

Summary for review:

There was a single conflicted file, yii2-adapter/legacy/config/GeneralConfig.php, with one conflict hunk just above the $filename property. The two sides were doing unrelated things: 6.x (d996dea, "Move @web alias support to Yii 2 adapter") added the $resourceBasePath and $resourceBaseUrl properties there, while this branch had only reflowed the adjacent docblock (@inheritdoc → {@inheritdoc}, part of the branch-wide code-style pass that also added $generateTransformsBeforePageLoad). I kept both: the two new properties from 6.x, followed by the {@inheritdoc} docblock form used consistently everywhere else in this branch's copy of the file. The corresponding resourceBasePath()/resourceBaseUrl() fluent setters from 6.x auto-merged and are intact.

Nothing here was ambiguous, so no CONFLICT-REVIEW: comments were needed. Note that php -l can't validate this file in this environment (local PHP is 8.3 and the file uses 8.4 property hooks that were already on the branch pre-merge), so verification was done by diffing the resolved file against both sides — it is exactly this branch's version plus the two 6.x properties and their two setters, with no other change.

[github-actions]

This PR had conflicts with 6.x. Claude attempted an automatic resolution and pushed a merge commit. Please review the merge commit carefully before approving/merging.

Claude's response

Merge completed and pushed.

Summary

No conflicts existed. Despite the task description, git status reported "All conflicts fixed but you are still merging," and verification confirmed there was genuinely nothing to resolve:

• HEAD (c918975, "Move immediate transform control to Craft driver") is a strict ancestor of MERGE_HEAD (02eb3c1, the 6.x merge commit for [6.x] Asset transformers #19464) — git merge-base HEAD MERGE_HEAD returns HEAD itself, and git log MERGE_HEAD..HEAD is empty.
• Every commit on this feature branch was already merged into 6.x via [6.x] Asset transformers #19464, so the merge only fast-forwards 55 commits of 6.x onto the branch. The merge was set up with --no-ff, which is why a merge commit was still pending rather than a fast-forward.
• An independent recomputation with git merge-tree --write-tree HEAD MERGE_HEAD produced a clean tree with no conflict report, and that tree hash is byte-identical to both the staged index tree and MERGE_HEAD^{tree}. No conflict markers exist in the index or working tree, and there were no dirty or untracked files.

I made no edits — there was nothing ambiguous to resolve and therefore no CONFLICT-REVIEW: comments to leave. I ran git add -A (a no-op), git commit --no-edit (creating merge commit 4e24bdb), and pushed.

One thing worth a human's attention: the push reported * [new branch] HEAD -> feature/asset-transformers. The remote branch did not actually exist — it appears to have been deleted after #19464 was merged, and the local checkout was tracking a stale remote-tracking ref. So this push recreated the remote branch rather than updating it. The branch now points at a merge commit whose tree is identical to 6.x's. If that branch was intentionally deleted post-merge, someone may want to delete it again.

[linear-code]

CMS-2352