Skip to content

repo_map_step_28.md

Latest commit

 

History

History
64 lines (46 loc) · 3.85 KB

repo_map_step_28.md

File metadata and controls

64 lines (46 loc) · 3.85 KB
Raw

Step 28 — Golden-file snapshot rollout via candy-testing

Branch: ai/golden-file-rollout Depends on: step-04 (candy-testing exists), step-15/16/17/18 (candy-forms / sugar-prompt / sugar-charts / sugar-table use candy-testing for some tests already) Blocks: —

Goal

Roll out candy-testing's assertGoldenAnsi helpers across all rendering libs that the analysis flagged: candy-forms, sugar-prompt, sugar-bits, sugar-charts, sugar-table, sugar-glow, candy-vt, candy-vcr, candy-shine. Replaces ad-hoc string comparisons and brings every lib into the "snapshot tests for ANSI output" pattern. This is the #1 ecosystem-wide weakness called out in the analysis.

Reference: §19, §149 (zero ANSI byte assertions cited as the most common weakness across 8+ libs).

Files expected to be modified

  • Each of the 9 lib's composer.json — add sugarcraft/candy-testing (dev) via path-repo-closure.
  • Each lib's tests/ — convert existing tests that build string-equal assertions into Assertions::assertGoldenAnsi(...); add NEW golden-file tests for view() output on representative scenarios.
  • tests/fixtures/ in each lib — golden files added.

Acceptance criteria

  • Each of the 9 libs has at least one assertGoldenAnsi test for a representative renderer (text input, table render, chart, markdown, etc.).
  • Existing tests preserved; new golden tests are additive.
  • UPDATE_GOLDENS=1 vendor/bin/phpunit regenerates fixture files (tested in at least one lib end-to-end).
  • Coverage on rendering paths increases measurably (track per-lib coverage delta).
  • git status clean on master.

Coder brief

  1. path-repo closure: candy-testing (dev-only require) in each of the 9 libs.
  2. For each lib: identify the most exercised render() method; build a representative input fixture; capture the output via UPDATE_GOLDENS=1 once; commit the .golden file; the test asserts against it on subsequent runs.
  3. Convert ad-hoc string-equality tests: where a test currently does $this->assertSame($expected, $rendered) with a literal string, swap to Assertions::assertGoldenAnsi('fixtures/foo.golden', $rendered). Drop the inline expected string into the fixture file.
  4. DO NOT delete any existing tests — only convert or add.
  5. Run phpunit in each lib + spot-check UPDATE_GOLDENS=1 regenerates one of them, then re-run unset to verify the regenerated golden matches.

Tester brief

  • Per lib: add 1+ assertGoldenAnsi test for the canonical view() output of the lib's main component(s).
  • Convert any tests where the expected string is embedded literally → use a golden file.
  • Verify UPDATE_GOLDENS=1 flow on candy-vt (or pick one lib that's representative).

Scribe brief

  • Each lib's README: a ## Snapshot tests paragraph pointing at the golden-file pattern and the UPDATE_GOLDENS env-var workflow.
  • Each lib's CALIBER_LEARNINGS: "Use assertGoldenAnsi for any new render() test. Re-record goldens with UPDATE_GOLDENS=1 phpunit after intentional output changes."

Ship brief

  • PR title: 9 rendering libs: golden-file snapshot tests via candy-testing
  • PR body: 
    ## Summary
- candy-forms, sugar-prompt, sugar-bits, sugar-charts, sugar-table, sugar-glow, candy-vt, candy-vcr, candy-shine all gain assertGoldenAnsi-backed golden-file tests.
- Closes the ecosystem's most-cited weakness: "no snapshot tests for ANSI rendering" (§19/§149).
- UPDATE_GOLDENS=1 phpunit workflow documented.
- Tests are additive; existing tests preserved.

## Test plan
- [x] vendor/bin/phpunit in 9 libs (existing suites + new golden tests all green)
- [x] UPDATE_GOLDENS=1 spot-check round-trip verified
- [x] php tools/check-path-repos.php

Refs: docs/repo_map_step_28.md, docs/repo_map_update.md §19, §149, §322
  • Commit subject: 9 rendering libs: golden-file snapshot test rollout via candy-testing.