[derek73]

Part of #1115, the pack-titles batch. Phase 3g, phase 3f and #1108 wrote the Spanish, Portuguese, Czech, Slovak and Vietnamese titles default = false so that phase 3 moved no default parse, not because of a collision. #804 made that default change for profesor. This PR makes it for the rest that no name or initials spell.

What changes

Read by default: 27 words, 36 readings. The nine doctorates have readings in both cs.toml and sk.toml.

• es: lic, licenciado, licenciada, lcdo, lcda, ingeniero, ingeniera, arq, mtro, mtra, doctora
• pt: doutor, doutora, professora, profa
• cs and sk: judr, mudr, phdr, rndr, mvdr, mddr, paeddr, pharmdr, thdr
• vi: ths, đbqh, dbqh

Listed, on by default: licda, profr, gral, tte (es) and rsdr, thlic, thmgr, plk, npor (cs). None has name bearers.

Kept the choice's, each read against its moves:

• maestra: a schoolteacher and a family name
• doña: a family name
• mga: Sri Lankan initials MGA
• bca: an English credential
• pan, and mgr's degree reading
• vi gs, ts, pgs: written as initials (NGO, TS would lose its given name)

Jan Novák, MUDr          given 'MUDr', family 'Jan Novák'   →  title 'MUDr', given 'Jan', family 'Novák'   (1.4.0 through 2.8.2)
Licenciado Juan Pérez    given 'Licenciado', middle 'Juan'  →  title 'Licenciado', given 'Juan'

Measured

config_diff.py diff --local all --committed against the tree without the readings, 2026-11-06 on e6a86bc9. 312 of the 8,529 matching names move:

• by source: Mexico's DOF signatures 219, ClinicalTrials.gov 48, ORCID 26, cz-candidates 11
• 287 change a name field, a title leaving it
• 25 move only from suffix to title: 14 Czech doctorates after a comma, one plk.v.v. and 10 licentiates (Teemu Elomaa, Lic.Med)

Other costs, recorded in decisions/vocabulary-model's 2026-11-06 #1115 entry:

• Doctora en Neumología leaves en as the given name
• Lic.Yilian takes the glued name word, as Dr. does
• one credential moves into the family
• a family name written first that is one of these words reads as the title (Doctora, Licenciado and Lic are each borne once as a family name in ORCID or PatentsView, none written first)

The snapshot moves one committed name (ĐBQH Nguyễn Văn Anh), and its CLDR record is re-recorded: the family and given names now agree with CLDR, and the title disagrees.

Tests and docs

• Locale rotator rows: 47 rows that proved each pack reads these words differently now form _READ_BY_DEFAULT_SINCE_1115, which asserts that the default reads each row as its pack does.
• Vacuous tests: six tests used lic or judr as a pack-only word and now use maestra and mga. Each was checked to read under the pack and the default as the old word did on master.
• Slavic pack test: it gains MgA rows in each comma shape, order and case.
• Docs and pack docstrings: docs/locales.rst (two doctests), docs/usage.rst's CLI example and the pack docstrings follow.
• Release log: the unreleased 3.0 ES/PT and CS bullets get new examples, and a Change bullet is added.

Review

The design-docs review recomputed 312, 8,529, the per-source split and the licentiate count, all exact. It read all 93 moves outside DOF and found no real name eaten, and confirmed both release examples against the 1.4.0, 2.8.0 and 2.8.2 wheels. It found eight problems, all fixed in the second commit:

• a Standing line still listing the titles as a choice's
• missing head notes on the phase 3f, 3g and Should parser_for(locales.VI) read Nguyen Van Anh as family Nguyen, middle Van, given Anh, and Le Van Anh as family Le? #1108 entries
• locales.rst overstating the bare trailing case
• the CLI doc example
• the Slavic test's lost pack-only coverage
• readings vs words
• the incomplete cost list
• "no name bears", which three family-name bearers contradict

Test plan

• uv run pytest (39,122 passed) and uv run pytest -n0 -m serial
• uv run mypy, uv run ruff check, python -m doctest README.rst
• uv run sphinx-build -b doctest docs

🤖 Generated with Claude Code