test(website): compare docs and editor screenshots before and after deployment

[benfdking]

Follow-up to the docs theme asset fix in #5034, which is now merged. This PR targets main and contains the regression tests and their build/deployment wiring.

The documentation build check only verifies HTML and the search index, so missing theme assets can reach GitHub Pages. Compare the docs overview, installation page, and playground editor with committed screenshot baselines. Keep the checks for successful theme CSS/JavaScript responses and editing SQL with updated formatted output.

• //playground:website_test tests the combined deployment artifact and is included in bazel test //... and the pre-deployment checks.
• //playground:deployed_website_test runs the same suite against https://playground.quary.dev, without rebuilding the website. A separate verify-deployment job depends on successful deployment and runs this target with --nocache_test_results. Its manual tag excludes it from wildcard test runs.
• Commit separate macOS and Linux baselines, use a fixed 1280×720 viewport and light theme, and bundle licensed Liberation fonts for consistent Linux rendering. Resolve Bazel's fontconfig path to an absolute path so Chromium loads these fonts on CI. Disable animations, hide the editor caret, normalize native scrollbar gutters before capturing docs, and block external docs fonts. The editor comparison covers its panels below the header.
• Allow at most 100 differing pixels. Missing baselines and mismatches fail without updating reference images. PR and deployment jobs upload expected/actual/diff images and traces on failure.

Validation:

• All three Playwright cases pass through Bazel on macOS and against read-only Linux baselines in the pinned Playwright 1.62.1 container, including Bazel-style font symlinks and its relative environment path.
• The deployed target passes all three cases against https://playground.quary.dev after the docs fix deployment.
• A temporary copy with only the docs header color changed fails both screenshot comparisons (60,272 differing pixels each), while its assets and text still load. Expected, actual, and diff images are retained in Bazel test outputs.
• The original regression cases fail without fix(docs): include theme assets in the website build #5034.
• Workflow action pinning, Prettier, and git diff --check pass.

For an intentional visual update, regenerate on each supported platform, review the changed PNGs, and commit them:

bazel test //playground:website_test \
  --test_arg=--update-snapshots=all \
  --test_env=WEBSITE_SNAPSHOT_DIR="$PWD/playground/website-tests/snapshots" \
  --strategy=TestRunner=local --nocache_test_results

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-10-07T08:57:50.231266Z	3be9f12	PR opened

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[github-actions]

Benchmark for 8addc3e

Click to view benchmark

Test	Base	PR	%
DepthMap::from_parent	130.0±2.89µs	128.9±3.79µs	-0.85%
LT11/union_all/100	2.2±0.02ms	2.2±0.02ms	0.00%
LT11/union_all/200	4.4±0.12ms	4.7±0.21ms	+6.82%
LT11/union_all/400	10.7±1.00ms	10.2±0.69ms	-4.67%
LT11/union_all/800	27.7±2.33ms	25.9±3.10ms	-6.50%
fix_complex_query	25.4±0.35ms	24.5±0.30ms	-3.54%
fix_superlong	77.3±4.87ms	76.1±4.96ms	-1.55%
parse_complex_query	4.5±0.08µs	4.5±0.23µs	0.00%
parse_expression_recursion	7.0±0.06µs	7.1±0.06µs	+1.43%
parse_simple_query	1297.7±31.31ns	1312.4±30.35ns	+1.13%
tpc-ds/lex_tpc-ds_99	12.5±0.04ms	12.5±0.07ms	0.00%
tpc-ds/parse_tpc-ds_99	509.9±5.18ms	503.3±4.29ms	-1.29%
tpc-h/lex_tpc-h_22	1341.4±3.80µs	1319.9±10.34µs	-1.60%
tpc-h/parse_tpc-h_22	43.1±0.19ms	43.1±0.25ms	0.00%

[github-actions]

Benchmark for 6501bb3

Click to view benchmark

Test	Base	PR	%
DepthMap::from_parent	128.7±4.00µs	128.9±3.61µs	+0.16%
LT11/union_all/100	2.2±0.01ms	2.2±0.01ms	0.00%
LT11/union_all/200	4.4±0.09ms	4.4±0.03ms	0.00%
LT11/union_all/400	9.4±0.26ms	9.2±0.13ms	-2.13%
LT11/union_all/800	23.0±1.30ms	22.4±1.14ms	-2.61%
fix_complex_query	25.3±0.12ms	25.3±0.17ms	0.00%
fix_superlong	70.1±5.58ms	72.3±6.09ms	+3.14%
parse_complex_query	4.4±0.07µs	4.5±0.12µs	+2.27%
parse_expression_recursion	7.1±0.07µs	7.2±0.11µs	+1.41%
parse_simple_query	1283.9±56.63ns	1332.2±22.02ns	+3.76%
tpc-ds/lex_tpc-ds_99	12.4±0.06ms	12.2±0.03ms	-1.61%
tpc-ds/parse_tpc-ds_99	491.3±0.91ms	491.9±2.41ms	+0.12%
tpc-h/lex_tpc-h_22	1321.8±4.89µs	1339.7±11.95µs	+1.35%
tpc-h/parse_tpc-h_22	42.8±0.10ms	42.7±0.12ms	-0.23%