Cross-renderer verification against independent HTML sources - #71
Merged
Conversation
A new harness renders a shared corpus through the gem and through the prominent independent implementations: a live headless Tiptap editor (byte parity, the gem's contract), @tiptap/html's static generateHTML, tiptap-php in docker when available, and Lexical's $generateHtmlFromNodes on a headless editor bound over @lexical/yjs. Comparisons beyond the byte-parity contract are DOM-normalized, with each source's documented spelling quirks translated or allowlisted so anything new still fails. Runs in CI after the render-parity e2e. It found a real bug: adjacent text runs sharing marks rendered as sibling wrappers instead of keeping the shared marks open, the way ProseMirror's serializer does. run_stack now builds each run's ordered open/close stack and render_text_runs diffs consecutive stacks, closing and opening only the tails. A mark whose attributes differ between runs never merges, since sharing is by identical opening tag. The prosemirror-yjs-html crate bumps to 0.1.2 with the fix; a crate test pins the merge.
The merging emitter replaced render_run's production call site, leaving it test-only; clippy under -D warnings rejects the dead public path.
Add @payloadcms/richtext-lexical's convertLexicalToHTML as a fifth cross-renderer source, over the same headless editor's serialized state. Link nodes are re-shaped into Payload's fields dialect first. A custom converter covers the code-block node, which Payload's default set does not. Payload's own spellings — styled spans for strike and underline, em over strong for bold italic, list classes, empty li attributes — are translated before the DOM-normalized compare, and anything else fails. Peer auto-install is now off for the frontend (bunfig.toml). Payload's admin-UI peers (react, next) are dead weight for the converter. The peers the demo really uses are explicit devDependencies instead: payload, @tiptap/y-tiptap, @tiptap/suggestion. @lexical/table and @lexical/text are pinned at 0.44 so Payload's nested 0.41 copies do not win the hoist.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
A new harness (examples/actioncable-demo/frontend/cross_render.mjs,
wired into CI) renders a shared corpus through the gem and through the
prominent independent implementations:
existing contract (the table case strips the documented editor-view
chrome: colgroup and min-width styles)
verified quirks allowlisted (drops textStyle color, writes
data-checked="" for checked items)
configured and its spellings translated (
for, explicitstart="1"); its dropped textStyle attributes are allowlisted
@lexical/yjs, with stock export idioms stripped (white-space spans,
and doubling, spellcheck attributes)
Every comparison converges to zero differences, so any new divergence
from any source fails the run.
The harness immediately found a real bug: adjacent text runs sharing
marks rendered as sibling wrappers (ab
) instead of keeping shared marks open
(ab) the way ProseMirror's serializer does.
The fix builds each run's ordered open/close stack and diffs
consecutive stacks; sharing is by identical opening tag, so a mark
with different attributes never merges. Pinned by a crate test and by
the mark-matrix corpus case against all sources. The captured fixtures
never contained adjacent shared-prefix runs, which is how it hid.
prosemirror-yjs-html bumps to 0.1.2. render_check.rb gains
prosemirror-core and lexical-core kinds so the harness can compare the
core renderers.