Slate v2 Example Parity Recovery Plan

Purpose

Make the example story honest for the broader no-regression claim.

This lane is not closed by:

• same-path current example files
• same-path current Playwright example tests
• replacement compatibility rows that only cover a subset of examples

It closes only when legacy/current example parity is explicit across the full example surface.

Core Rule

Maximum parity with legacy is now required for all examples that still matter to the blanket replacement claim.

Legacy examples may be added or extended when the rewrite unlocked a fairer comparison surface, but those rows must be labeled explicitly as extended instead of being smuggled in under vague “better cut” language.

For same-path legacy examples, the default recovery target is the closest legacy source structure and contributor-facing chrome that the current runtime can honestly support.

“Behavior passes somewhere” is not enough. A scratch rewrite is debt unless the current-only divergence is recorded explicitly.

Git diff wins ties. If the same-path file still reads like a rewrite in source, the row stays open even when a browser row is green.

Owners

• true-slate-rc-proof-ledger.md
• release-readiness-decision.md
• master-roadmap.md
• 2026-04-13-slate-v2-full-no-regression-story-plan.md

Supporting inputs:

• legacy-playwright-example-tests.md
• example-parity-matrix.md
• release-file-review-ledger.md
• /Users/zbeyens/git/slate/site/examples/**
• /Users/zbeyens/git/slate-v2/site/examples/**
• /Users/zbeyens/git/slate-v2/playwright/integration/examples/**

Required Classification

Every example row must land in exactly one bucket:

• mirrored
• recovered
• extended
• mixed
• explicit cut
• open

Workstreams

1. Legacy Example Inventory

• inventory legacy TS example surfaces
• inventory legacy Playwright example tests
• map them to current example and proof owners

2. Mirror vs Extend Decisions

• default to mirror
• extend only when the rewrite unlocked a clearly better comparison
• record the reason for every extension

3. Proof Ownership

• direct browser proof where possible
• replacement compatibility where honest
• benchmark owner where the example is really benchmark-owned
• explicit-cut rationale where the example no longer belongs to the kept claim

4. Control-Doc Sync

• roadmap
• verdict
• proof ledger
• release file review ledger
• maintainer drift register

5. Source/UI Recovery Default

• audit same-path examples for legacy-source closeness, not only behavior proof
• reopen rows where proof exists but the contributor-facing example drifted
• recover from legacy example source first; keep rewrites only with explicit justification

Current Progress

• exact example drift inventory now exists in example-parity-matrix.md
• first recovery batch is already landed in replacement-compatibility.test.ts:
  • replacement search-highlighting parity row
  • replacement code-highlighting parity row
• targeted cross-repo proof for that batch is green:
  • legacy + replacement search-highlighting
  • legacy + replacement code-highlighting
• second recovery batch is now landed in replacement-compatibility.test.ts:
  • replacement inlines parity row
  • replacement richtext parity row
  • replacement custom-placeholder parity row
• targeted cross-repo proof for that batch is green:
  • legacy + replacement inlines
  • legacy + replacement richtext
  • legacy + replacement custom-placeholder
• third batch results:
  • android-tests is now recovered as an explicit extended row: the current page carries the legacy manual case picker plus the newer IME hub links
  • check-lists is now recovered with direct legacy/current parity rows
  • huge-document is now treated as an explicit benchmark-owned parity row instead of an ownerless gap
  • scroll-into-view is still open: the current row is green, but the mirrored legacy/current parity action is still red on the legacy inner-container scroll behavior even after a fair legacy extension attempt
  • targeted proof is green for:
  • legacy + replacement android-tests
  • legacy + replacement check-lists
• 2026-04-15 source + DOM audit result:
  • the softer runtime/DOM audit was still too generous for a source-first recovery standard
  • the git-diff rebaseline invalidated more rows than the earlier audit:
    • custom-placeholder
    • forced-layout
    • markdown-preview
    • markdown-shortcuts
    • plaintext
    • read-only
    • shadow-dom
    • styling
    • tables
  • code-highlighting remains the clearest hard-open row: source similarity is only 0.062
  • some rows are now legitimately source-close after the actual ports:
    • check-lists
    • embeds
    • hovering-toolbar
    • iframe
    • inlines
    • mentions
  • check-lists is no longer one of the invalidated source-rewrite rows: the file now reads close to legacy again; the earlier as CustomEditor and test-only id drift were removed, and the remaining source-level gap is that the current EditableBlocks surface still does not expose legacy spellCheck / autoFocus and still carries current type/module syntax; the red standalone checkbox test is explicitly deferred to the later proof-fix pass
  • some rows are source-close but still not fully closed in proof:
    • editable-voids
    • images
    • paste-html
    • richtext
  • continuation batch:
    • the slate-react exported-type audit found no new legacy barrel misses beyond the already-explicit cuts:
      • RenderChunkProps
      • NODE_TO_INDEX
      • NODE_TO_PARENT
    • the audit did expose one real current public type regression: EditableTextBlocksProps was still intersecting the native textarea placeholder?: string with the current custom placeholder?: ReactNode, so JSX placeholder hosts in custom-placeholder / placeholder-surface typed as garbage until the inherited native placeholder key was omitted again
    • forced-layout is now source-close again after restoring the in-file legacy layout normalizer, legacy copy, and direct title/paragraph shell
    • targeted current proof is green for forced-layout:
      • initial title + paragraph shell
      • full-delete restoration
    • search-highlighting is now source-close again after restoring the legacy search/decorate flow, Slate wrapper, and contributor-facing input shell
    • remaining forced drift for search-highlighting is explicit: current projectionStore + renderSegment wiring instead of legacy decorate + renderLeaf
    • targeted current proof is green for search-highlighting:
      • typing text yields two highlighted matches

Exit

• every legacy example row is classified
• every extension is explicit and justified
• same-path current files are no longer treated as implicit parity closure
• control docs stop pretending deletion closure equals example parity
• same-path recoveries either stay close to legacy source or record the reason they do not