ContextQMD
Back to Plate

DOM Selection Ralplans Should Rank Editor Test Sources By Owner

docs/solutions/workflow-issues/2026-05-08-dom-selection-ralplans-should-rank-editor-test-sources-by-owner.md

53.0.62.6 KB
Original Source

DOM Selection Ralplans Should Rank Editor Test Sources By Owner

Problem

DOM selection planning gets sloppy when every editor test corpus is treated as equally authoritative. Lexical, ProseMirror, and Tiptap are useful for different reasons.

Symptoms

  • The plan says "steal ecosystem tests" without naming exact source rows.
  • Tiptap command tests start looking like Slate API proposals.
  • IME, clipboard, table, focus, and DOM bridge rows get mixed into one lane.

What Didn't Work

  • Using the Lexical harvest alone: good browser-regression style, but not the deepest DOM selection architecture evidence.
  • Treating Tiptap as an architecture source: useful focus timing pressure, but mostly ProseMirror command wrapping and product/demo tests.
  • Copying ProseMirror shape wholesale: its view-desc and integer position model are not Slate's model.

Solution

Rank sources by the owner they actually prove:

  • ProseMirror view/test/webtest-selection.ts is the primary source for DOM selection import/export, coordinates, fallback, BR-hack, and selectable-node movement invariants.
  • Lexical harvest reports are the primary source for browser-first regression shape, honest transport, explicit skip reasons, and mobile/table deferrals.
  • Tiptap focus.ts is pressure for focus timing only: mobile direct focus, Safari preventScroll, and React RAF. It does not justify copying Tiptap's command API into Slate.

Then classify each row as steal, reject, or defer, with a Slate owner: slate-dom, slate-react, slate core, or another lane.

Why This Works

It preserves the portable behavior invariant without importing another editor's framework shape. The source ranking also prevents closure lies: jsdom renderer composition does not prove mobile IME, clipboard transform tests do not prove a DOM selection bridge, and whole-table selection does not become a raw Slate requirement just because Lexical owns it.

Prevention

  • In DOM selection ralplans, read ProseMirror selection tests before Tiptap demos.
  • Use Lexical harvests to shape browser proof and skip discipline.
  • Keep Tiptap evidence narrow unless the source is a browser behavior file.
  • Every copied test row needs: source path, behavior invariant, Slate owner, action, and verification route.
  • docs/plans/2026-05-08-slate-v2-dom-selection-focus-bridge-ralplan.md
  • docs/editor-test-harvester/lexical/report.md
  • docs/solutions/test-failures/2026-04-22-slate-browser-selectionchange-proof-must-separate-traceability-from-programmatic-import.md
  • docs/solutions/logic-errors/2026-05-06-slate-react-foreign-dom-selections-must-be-ignored-before-import.md