Demo: repo-wide style enforcement (.editorconfig, Prettier, .gitattributes)

[joewiz]

Note: This PR is opened on the joewiz fork (not eXist-db/eXide) to facilitate discussion among the eXide devs before deciding whether/how to land this upstream.

Goals

After noticing eXist-db#796 silently flipped a few lines from spaces to tabs in src/util.js, we wanted to look at what it would take to enforce a consistent style across the repo automatically — so reviewers don't have to spot indent flips by eye and so contributors get correct formatting from their editor with zero local setup.

This PR explores three lightweight, complementary mechanisms plus a demonstration of the effect:

1. .editorconfig — picked up automatically by VS Code, JetBrains, Vim plugins, etc. Establishes baseline indentation and line-ending rules without requiring any tooling.
2. .gitattributes — pins eol=lf so Windows contributors can't introduce CRLF churn.
3. Prettier config — eXide already depends on prettier, @prettier/plugin-xml, and prettier-plugin-xquery (used internally by the editor's Format command). This PR exposes that same formatter as a repo-level check.

What's in here

Commit 1 — config files only (no source changes)

File	Purpose
.editorconfig	2-space indent for JS/TS/CSS/JSON/YAML/Markdown (modern community convention); 4-space for XQuery/XML/HTML/template (existing project style). LF, trim trailing whitespace, final newline.
.gitattributes	text=auto eol=lf + binary file list.
.prettierrc.json	Global tabWidth 2; XML/XQuery overrides bump back to 4. xmlSelfClosingSpace: false so <br/> and <?xml ... ?> emit without the (long-obsolete) XHTML-compat leading space.
.prettierignore	Skips generated bundles, REx-generated parsers, vendored support tree, build output, lockfiles.
package.json	Adds npm run format and npm run format:check.
.gitignore	Allowlists the four new dotfiles past the existing .* rule.

Landing just commit 1 is essentially free — no source changes, just future safety.

Commit 2 — demo: npm run format applied across the repo

This is the "what would happen if we hit the button" commit. 143 files changed, ~31.6k inserts / ~27.2k deletes.

File type	Files	Lines inserted	Indent
.js	95	20,966	2
.css	5	3,502	2
.svg	3	2,266	2
.xqm	15	1,593	4
.json	4	1,557	2
.tmpl	2	840	4
.xq	7	428	4
.xml	4	240	4
.xsd	1	98	4
.html	2	63	4
.md	2	35	2
.yml	2	30	2
.xconf	1	1	4

All 179 unit tests pass after the sweep.

How to run the checks

npm run format:check  # read-only — exits non-zero if any file would change
npm run format        # apply
npx prettier --check src/util.js   # one file

A CI gate (prettier --check . in GitHub Actions) is the natural follow-up — soft warning first, then hard fail.

Options for landing this

Commits are organized so that we can land just commit 1 (config baseline, no source churn), or both. The big format sweep is optional and deliberately separable. A few plausible paths:

1. Land config only. Future PRs run prettier on touched files; everything else stays as-is. Slow but invisible.
2. Land config + sweep as a single commit on develop. One-time disruption, clean baseline going forward. Pair with .git-blame-ignore-revs so git blame skips the format-only commit on GitHub.
3. Split the sweep into per-file-type PRs to ease review.
4. Drop the sweep, just land config + CI check. New PRs are automatically formatted; old code drifts toward correctness over time.

Open questions for discussion

• 2-space vs 4-space JS. Modern norm is 2-space (Airbnb/Standard/Google/npm); eXide is currently mixed (some tabs, some 4-space). This PR proposes 2-space. The 4-space variant produces ~12k inserted lines instead of ~21k — smaller diff but locks in a less-conventional style. Genuinely open to flipping.
• XQuery formatter scope. prettier-plugin-xquery (DrRataplan) only supports XQuery 3.1. If we have any 4.0-specific syntax in .xq/.xqm files, they need an explicit ignore.
• Timing. A 143-file diff invalidates every open PR's merge base. Best landed during a quiet window.
• Hooks? Should we also add lint-staged + a pre-commit hook so contributors don't have to remember to run npm run format? Or just rely on CI to nag them?

Risk / verification

• 179 unit tests pass before and after the sweep.
• No semantic changes — pure formatting.
• Generated bundles, REx parsers, and the support/ tree are excluded so we don't fight the build pipeline.

Credits

Initiated after noticing eXist-db#796 flipped indentation in src/util.js from spaces to tabs. Cc @duncdrum, @line-o, @wolfgangmm, @dizzzz, @reinhapa for input.

🤖 [This PR description was co-authored with Claude Code. -Joe]

[duncdrum]

Leave the two commits, and add the prehook.

As for churn now is a good time. 2 spaces is fine for js.

Xq4 is not for this release, but I have faith in prettier support coming soon™

[joewiz]

[This response was co-authored with Claude Code. -Joe]

Pushed e12c2cb adding the pre-commit hook per your feedback, @duncdrum. Summary of how I'm reading your comment:

• ✅ "Leave the two commits" — kept (the new commit is a third one for the hook itself, separate from the config baseline + format sweep).
• ✅ "Add the prehook" — done. husky + lint-staged, hook runs npx lint-staged on staged files. Smoke-tested locally: deliberately-misformatted JS gets caught, valid files get auto-prettied in place. The hook installs automatically on fresh npm install via the prepare script.
• ✅ "Churn now is a good time" — agreed in principle, but I'd like to delay the upstream PR (eXist-db/eXide) until Modernize eXide: CM6 editor, REx parser, Roaster API, LSP integration eXist-db/eXide#794 lands and the freeze settles, to avoid invalidating its merge base today. This joewiz PR is fine to grow in the meantime.
• ✅ "2 spaces is fine for js" — confirmed by the .prettierrc.json already in commit 1.
• ✅ XQ4 deferred to future prettier-plugin-xquery support — noted, no action.

The diff is unfortunately noisy because the npm install for husky/lint-staged also re-indented package-lock.json from tabs to 2 spaces (npm now respects .editorconfig). The substantive changes are in package.json, .husky/pre-commit, and .gitignore.