fix(ci): let the vendored snapshot stay compact instead of losing to the formatter - #290
Merged
Conversation
…the formatter CI has been red on main since #289. The vendor bot serializes the dataset with a bare `JSON.stringify` — compact by an explicit decision recorded on `serialize`, since pretty-printing triples the committed size of a 1961-point array for a diff git shows either way — and biome's formatter rejects the single line that produces. The bot opened its PR, the PR merged, and CI went red on a diff no human wrote. Two rules disagreeing, so one has to give. The formatter yields: a vendored, machine-generated artifact is not source, and `dataset-*.json` leaves `files.includes`. The alternative — indenting the generator — would have overturned a decision the script documents, and reformatted 2700 lines to do it. Nothing about the data moves. `dataset-2026-07.json` is pretty-printed because a human reformatted it by hand once; both spellings are now legal and the bot's own output no longer needs that step. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
…cale (#291) PR #288 described this review round in its body but merged without it. This lands the code the description already promised. SPRAWL's basemap chip named `absolute` before `TOUCH_TARGET_HEIGHT`, and the constant opens with `relative` to anchor its overlay. `cn` is tailwind-merge, which resolves a position conflict in favour of the last name — so the chip lost its corner and fell back into the flow with every target class present and correct, which is exactly why the className assertion sailed past it. The constants now document the ordering, and a test pins the position rather than the classes. The scale guard gains the `sp-*` spellings, step by step rather than by a prefix rule: it is a second scale under the same utilities and its ends do not match the base one, so `gap-3xl` is real while `p-sp-3xl` renders nothing, and only the step tells the two apart. The completeness test holds the new entries to the preset the same way it holds the old ones. `TOUCH_TARGET_ICON` joins the public surface — the height overlay plus the real width it tells you to pair with — after both canvases composed the pairing identically. Two callers, which is ADR 0014's bar. The target tests now assert against the constants instead of the classes they happen to expand to, so respelling a target cannot red them without a regression behind it. Also recorded rather than silently done: the byte-identical range-input stylesheets, and the `sp-*` row in CLAUDE.md's scale table. Co-authored-by: Claude Opus 5 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
CI has been red on
mainsince #289 — including on #288, which merged through it. Neither PR broke anything: the failure is a diff no human wrote.What happens
vendor-dataset.mjsserializes with a bareJSON.stringify, and that is deliberate — the rationale sits on the function itself:Biome's formatter disagrees, and a 1961-point array on one line is exactly what it rejects. So the monthly bot (ADR 0022) opens its PR, a human reviews the per-point capacity diff and merges, and
npm run checkgoes red onmain.dataset-2026-07.jsonlooks fine only because someone reformatted it by hand after generation. That is the workaround this removes, not the convention.Which rule yields
The formatter. A vendored, machine-generated artifact is not source, so
dataset-*.jsonleavesfiles.includes, and the script's comment now records where the exclusion lives and why it has to.The alternative was indenting the generator. That overturns a decision the script explicitly documents and reformats ~2,700 lines to do it — so it needed a stronger reason than "the formatter said so."
No data moves in this PR. Both spellings are now legal, and the bot's output stops needing a human formatting pass to be mergeable.
Verification
npm run check— green repo-wide, 291 files, against the existing compactdataset-2026-08.json, which is what proves the exclusion matchesnpm run test:run --workspace @cyberdeck/sprawl— 91 passednpm run vendor:datasetto confirm freshly generated output is legal under the new config, then reverted the snapshot: live drift belongs in the bot's own PR with a human on the per-point diff, not smuggled in hereNo changeset — tooling and CI only, no app behavior.
🤖 Generated with Claude Code