ContextQMD
Back to Plate

TipTap harvest placeholders need harness routing

docs/solutions/best-practices/2026-05-10-tiptap-harvest-placeholders-need-harness-routing.md

53.0.61.8 KB
Original Source

TipTap harvest placeholders need harness routing

Problem

TipTap demo specs can be syntactically runnable while containing no behavior: they visit a route, then leave TODO: Write tests. Counting those as portable-mixed makes the harvest look deeper than it is.

Symptoms

  • Inventory shows runnable Cypress files with zero extracted test names.
  • The file body contains route setup only.
  • The report has more apparent behavior rows than actual assertions.

What Didn't Work

  • Routing by path alone was too generous. Files under useful folders like GuideContent/GenerateText or Nodes/TaskItem still had no assertions.

Solution

After test-name extraction, re-open every runnable file with zero names. If the body is only setup plus TODO, classify it as harness with an explicit reason.

In the TipTap harvest, 14 files moved this way, including:

  • ../tiptap/demos/src/GuideContent/GenerateText/React/index.spec.js
  • ../tiptap/demos/src/GuideContent/GenerateText/Vue/index.spec.js
  • ../tiptap/demos/src/Nodes/TaskItem/React/index.spec.js
  • ../tiptap/demos/src/Nodes/TaskItem/Vue/index.spec.js

Why This Works

The harvester should count behavior, not file shells. A no-assertion Cypress file can still be useful as evidence that a demo route exists, but it cannot justify a Slate or Plate test row.

Prevention

  • Treat runnable && testNames.length === 0 as a pressure check, not as a row.
  • Read zero-name files before closure.
  • Route assertion-free TODO files to harness.
  • Keep behavior counts separate from route/demo existence counts.
  • docs/editor-test-harvester/tiptap/report.md
  • docs/editor-test-harvester/tiptap/inventory.md
  • docs/solutions/best-practices/2026-05-09-lexical-normalization-harvest-rows-need-selection-query-boundaries.md