Write full GPL citation styles and locales from the style manuals

[dknauss]

Summary

What changed?

The nine bundled CSL styles and three locales are rewritten in full, each from its own manual:

Style	Source
Chicago notes-bibliography	Chicago Manual of Style, 17th ed.
Chicago author-date	Chicago Manual of Style, 17th ed.
APA 7	APA Publication Manual, 7th ed.
MLA 9	MLA Handbook, 9th ed.
Harvard	Cite Them Right
IEEE	IEEE Reference Guide (2023)
Vancouver	NLM Citing Medicine
OSCOLA	OSCOLA, 4th ed.
ABNT	NBR 6023:2018

Before and after, for one journal article:

Style	Before	After
Chicago NB	Green, S. Metadata Drift in Institutional Repositories. Journal of Library Systems. 2020. https://doi.org/…	Green, Samuel. “Metadata Drift in Institutional Repositories.” Journal of Library Systems 27, no. 2 (April 2020): 97–111. https://doi.org/…
APA 7	Green, S. 2020. …	Green, S. (2020). Metadata Drift in Institutional Repositories. Journal of Library Systems, 27(2), 97–111. https://doi.org/…
OSCOLA	(same shape as Chicago)	Green S, ‘Metadata Drift in Institutional Repositories’ (2020) 27 Journal of Library Systems 97
ABNT	(English terms)	GREEN, Samuel. Metadata Drift in Institutional Repositories. Journal of Library Systems, v. 27, n. 2, p. 97-111, abr. 2020. DOI: https://doi.org/…

The full reviewed output for all 20 corpus items in every style is in tests/fixtures/csl-styles/*.txt.

Why were the styles stand-ins? The official CSL styles and locales are CC BY-SA 3.0. That isn't GPL-compatible, and SPEC.md and THIRD-PARTY-NOTICES.txt forbid bundling them. So these are written for Borges from each manual's rules, not copied from the CSL files, and stay GPL-2.0-or-later.

• Before: about 2 KB stand-ins. Every given name was initialized. Volume, issue, pages, editors, translators, editions, and access dates were dropped. The locales had no month names or quotation marks, and pt-BR was an English copy, so ABNT printed "and" and "accessed".
• Now: each manual's author-list rules apply ("and" or "&", "et al." thresholds, APA's 21+ ellipsis). So do its forms for journals, chapters, edited and translated books, magazines and newspapers, webpages, theses, reports, conference papers, preprints, cases, and statutes. Harvard now uses en-GB.

Formatter changes. These work around citeproc-php 2.7 behavior; each quirk is in docs/external-eccentricities.md.

• One CiteProc per entry. Name state leaked between entries: once one entry was cut to "et al.", every later entry lost its "and". This bug was live before this PR.
• bibliography_builder_prepare_csl_for_formatter():
  • maps literal names to family-only names. Organization authors rendered empty, which was also live before this PR;
  • derives page-first for OSCOLA.
• Cleanup in bibliography_builder_normalize_formatted_text():
  • removes the " ," citeproc-php puts before labels;
  • initializes hyphenated names fully ("J.-W.", not "J.- woo");
  • moves , and . inside closing quotes for en-US styles.

Deliberate deviations, listed in docs/csl-styles.md:

• Titles keep their stored capitalization, because the block re-applies italics by matching the title text.
• URLs and DOIs stay plain https:// links, so OSCOLA drops its angle brackets.
• There is no Chicago 3-em dash for repeated authors.
• Journal titles are not abbreviated.

Existing posts keep their saved text until an entry is added or edited or the style is changed. The editor never reformats on load.

Heads-up for the planned year suffixes (2020a/2020b, REQ-S4). Formatting each entry on its own means citeproc can't see the whole list. When that feature lands, it will need to assign suffixes itself; the plan already coordinates suffixes with the JS sorter. SortCoordinationTest still passes, because the styles keep <sort> blocks, now in separate sort-* macros.

Validation

• composer test:php passes (287). It includes:
  • the new CslStyleGoldenTest, which pins 9 styles × 20 items and fails if the vendor/ copy is stale;
  • the new FormatterAdaptationTest (7 tests).
• composer lint:php and composer analyze:php pass.
• npm run lint:js and npm run test (872) pass. No JS changed.
• verify-metrics.sh passes. The LOC rows are updated.
• Tested in the block editor. I haven't loaded a live editor yet. save()'s italics and links rely on the title and URL text, which the styles keep as stored.

Checklist

• Linked issue or explained why none was needed: requested directly (replace the stand-in styles).
• No sensitive details disclosed publicly

🤖 Generated with Claude Code

https://claude.ai/code/session_01KbVuMPN7YTcV27vwsG3LZ2

---

Generated by Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 86.75%. Comparing base (cd4a381) to head (1522f1e).
⚠️ Report is 2 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #103      +/-   ##
==========================================
+ Coverage   86.65%   86.75%   +0.09%     
==========================================
  Files          52       52              
  Lines        5367     5406      +39     
  Branches      597      597              
==========================================
+ Hits         4651     4690      +39     
  Misses        235      235              
  Partials      481      481              

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.