perf: drop runLayoutPipeline from header/footer effect deps

[jedrazb]

Summary

• The header/footer re-layout useEffect listed runLayoutPipeline in its deps array. That callback's identity churns on every keystroke because the document prop changes whenever the parent pushes a new doc into history, which re-runs this effect and fires a redundant full layout pass on top of the one already scheduled by the transaction handler.
• Fix: replace runLayoutPipeline in the dep array with the concrete pagination inputs that should trigger re-layout — HF content, margins, pageSize, columns, pageGap, zoom, sectionProperties. Each of these is reference-stable across keystrokes (margins/pageSize/columns come from useMemo([sectionProperties]); sectionProperties itself is preserved by fromProseDoc's baseDocument path).

Why both commits

The first commit simply dropped runLayoutPipeline from the dep array. That fixed the keystroke duplicate but silently broke another path: under the old code, changing margins / pageSize / sectionProperties also churned runLayoutPipeline's identity (those values are in its useCallback deps) and piggy-backed a full re-layout through this effect. Dropping the dep without replacement meant ruler-drag and Page Setup apply no longer reflow pages until the next keystroke. The second commit addresses the review finding by listing the concrete inputs instead of relying on callback identity.

Performance impact

Local benchmark, 310-page document, 10 keystrokes at document start (tests/performance-large-docs.spec.ts), n=3 runs each = 30 measurements per branch:

Branch	start-of-doc avg	start-of-doc peak
main	81.7 ms	158 ms
first commit of this PR (regression)	64.3 ms	109 ms
this PR HEAD	70.3 ms	121 ms

~14% faster than main while correctly re-laying out on margin/pageSize/sectionProperties changes.

Why now

Found while auditing #263 — a much larger incremental-layout PR. The duplicate-pipeline issue was masking that PR's perf claims (every "fast incremental" pass was being followed by a full pass anyway). Even without the rest of #263, this dep-fix gives most of the keystroke-latency improvement #263 was reaching for, without the architectural cost.

Audit details and recommendation for #263: see the comment on that PR.

Test plan

• bun run typecheck clean
• tests/performance-large-docs.spec.ts start-of-doc avg drops from ~82ms to ~70ms locally (n=30)
• Static review: margins / pageSize / sectionProperties changes still trigger re-layout (they're now explicit deps)
• CI green
• Manual: ruler-drag reflows pages; Page Setup apply reflows pages

🤖 Generated with Claude Code

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
docx-editor	Ready	Preview, Comment	Apr 16, 2026 8:34am