fix(#544): a credential run keeps its ambiguous members

[derek73]

Closes #544.

What changes

A credential run keeps its ambiguous members (#544). Readers that met an ambiguous acronym (ma, ba, ed, do, jd, meng, lac, …) took its name reading without looking at the credential in front of it, and C1 read only a one-word part after a comma as a credential. #540's meng/lac marking and #289's case lean turned both into ordinary misreadings.

	2.3.0	master	now
John Smith, PhD MEng	suffix PhD MEng	given PhD, middle MEng, family John Smith	suffix PhD MEng, reported
John Smith, Ed Ma	given Ed, middle Ma, family John Smith	the same	suffix Ed Ma, reported
Jane Doe, MS LAc	suffix MS LAc	title MS, given LAc, silent	suffix MS LAc, reported
John Smith PhD MEng	suffix PhD MEng	middle Smith PhD, family MEng	suffix PhD MEng, reported
abdul Smith Jr Ma	suffix Jr Ma	middle Jr, family Ma	suffix Jr Ma
Smith, PhD Ma	given PhD, middle Ma	the same	suffix PhD Ma, reported
doe, jane v phd do	family do doe	the same	suffix v phd do, reported
Wang M.Eng.	suffix M.Eng.	family M.Eng.	suffix M.Eng.

• C1 reads a run. With two or more name words before the comma, a part of two or more words, each suffix vocabulary or a member of the ambiguous credential class, with no one-letter roman numeral, is the credential run, and the flip reports once. A run whose every listed member is written in capitals in a mixed-case name is left to the listing-form path, which already reads it whole (John Smith, PhD MA, silent as before). 1.4.0 read all of these as suffixes.
• A credential in front speaks for the member at every trailing slot: the no-comma peel (the maiden walk and the P5 reserve inherit it), a comma part read wholly as credentials, and the given slot, where it outranks P6's particle attachment. Only a credential IN FRONT speaks (Wang Ma PhD keeps family Ma); a connective, a particle (vd, mc: Smith vd Ma, John keeps family Smith vd Ma), a one-letter numeral, a dual in the given part's leading title run, and the walk's own leading word (Om Ma, PhD Ma keep family Ma) speak for nothing, and it does not cross a maiden clause. This reverses decisions.md#S2's Use letter case to disambiguate suffix vs. surname: "Jack MA" vs "Jack Ma" #289 item (ii).
• Reports: an anchored pick is reported where the member's own writing declined it; a member whose capitals already say credential is not.
• M.Eng. passes S2's period gate as M.A. does (two or more period-closed chunks spelling a listed member).
• A part read wholly as suffixes raises no name-reading report (rules.md#C2). Found on the way: since 2.0.0, John Smith, Jr., Freiherr von Richthofen reported particle-or-given for von, a word the same parse put in the suffix; this branch's run rule also made the particle chain's report reachable in John Smith, PhD Do Ma. Report-only; no field moves.

Decisions taken along the way (Derek, 2026-09-27/28)

• The run rule takes all-member runs and a leading member (Ed Ma, MEng PhD), not only a trailing one.
• One rule at every trailing slot, anchored in front only.
• A title/suffix dual counts as suffix vocabulary in the C1 run (John Smith, MD MEng → suffix), but a dual in the given part's leading title run is a title: Smith, Ms Ma keeps given Ma, and Smith, Ms MD Ma keeps master's title Ms MD, given Ma.
• A one-letter roman numeral anchors nothing and voids a run (single-letter shape, as S3); Jr, III anchor.
• A by-shape member never settles a run: John Smith, X.Y.Z. MA reads suffix X.Y.Z. MA, reported.
• An anchored pick in a comma part read wholly as credentials reports (Smith, PhD Ma, John Smith, PhD Ma Prof.).

Recorded boundaries and costs

• Accepted: John Smith, Ms Ma → suffix Ms Ma, reported (as John Smith, Ms alone reads); Om Jr Ma → suffix Jr Ma, no family; a capitalised MA opening a one-word-family given part with a degree behind it (Smith, MA PhD Ma) reads wholly as credentials, reported; doe, jane v phd do now reports suffix-or-name rather than particle-or-given.
• Limits kept a name word (rules.md#S2 Accepted): John Smith Ph. D. MEng, John Smith PhD Prof. Ma, John Smith PhD Do Do, Smith, PhD Do Ma (the P2 chain takes Do Ma).
• Jane Doe Jr. nee Smith Ma keeps maiden Smith Ma (rules.md#M2 Accepted).

Verification

• Suite: 10,487 passed. mypy and uv run ruff check clean; sphinx doctest and html -W clean; _assign.py, _group.py, _pieces.py, _segment.py, _vocab.py at 100%.
• Differential gate: exit 0 at all five baselines, 0 unexplained, radar unclassified unchanged (0/5/6/7/7). At 2.3.0 five fix(#544) rules classify 17 names. Every new or moved corpus name is attributed against the baseline wheel, master and HEAD; Wang M.Eng. belongs to feat(#449) / fix(suffix-routing), not Should a credential run after a comma keep its comma when the last word is an ambiguous acronym? John Smith, PhD MEng reads first PhD, last John Smith #544, and two labels are joint with parse("John Smith MD PhD").suffix returns "MD, PhD" — a run without a comma renders with one #436/Should the suffix separator be derived from the comma the writer typed, rather than from segment shape? #437 where the suffix diff is only the spacing change.
• Grid (254,496 parses, recipe in decisions.md's Should a credential run after a comma keep its comma when the last word is an ambiguous acronym? John Smith, PhD MEng reads first PhD, last John Smith #544 entry, comparator e10e83b): 89,825 move; 0 new violations of either invariant (anchored member in a name field 29,199 → 1,530; unambiguous credential in a name field 36,675 → 4,734); clean-to-clean moves are only M.Eng.'s dropped report.
• Corpus sweep (master's 1,418 names): 4 names move, all intended.
• Frames per parse (py3.11, master → now): unchanged for ordinary names (Smith, John 182, John Smith 151, John Q. Public 226, Doe, John MA 283, Smith, Ed 213); a comma name with two or more words each side pays one (Doe Smith, Jane Q. 317 → 318); a name whose ambiguous member declines on its writing pays a cheap reach test and, only where a suffix piece stands in front, the anchor pass (John Smith Ma 244 → 246, Smith, John Ma 275 → 279, Doe, John Q. Ma 316 → 320); anchored names pay the pass (John Smith PhD MEng 286 → 329); reference band 370/407 unmoved.
• Negative controls: anchor off → the company property fails on 48 parses; the walk's leading word allowed to anchor (both exclusions removed) → the converse fails on 42; anchor_in_reach reading a suffix word as none → its sweep fails at 878 member positions; per-member look-behind → credential_run 15.8× for 4× the input.
• Review: spec and quality review per commit, re-review of every fix. Findings fixed on the branch: an inline fold that duplicated a predicate (now _vocab.run_word_fold, with an agreement test); the walk's own leading word anchoring (Om Ma lost its family); a property oracle blind to the no-comma half; ledger attribution (Wang M.Eng., three 1.4.0 maiden names, joint labels); two silent readings from the design-docs review (commit 6, decided by Derek); the stray tail report (commit 7); a frame on double-surname comma names (commit 8). A whole-branch review found no critical or important issue. A final four-agent pass (code, tests by mutation, comments, silent failures) found a particle anchoring the acronym behind it (Smith vd Ma, John lost the family's contiguity, silently), a duplicated anchor walk, two unpinned branches and a set of stale comments: commits 9-10, each reviewed again. A /simplify pass (reuse, simplification, efficiency, altitude) folded the three hand-built given-slot anchor queries into one _pieces.given_slot_anchors (commit 11); readings byte-identical over 27,662 × 3 and 76,128 × 3 generated names plus every case row, plain names at master's frames.

Commits

1. 076a28a fix(Should a credential run after a comma keep its comma when the last word is an ambiguous acronym? John Smith, PhD MEng reads first PhD, last John Smith #544): a listed member written in period-closed chunks passes S2's period gate
2. 460e2aa fix(Should a credential run after a comma keep its comma when the last word is an ambiguous acronym? John Smith, PhD MEng reads first PhD, last John Smith #544): C1's name-word count reads a credential run as it reads one word
3. 4b708ca fix(Should a credential run after a comma keep its comma when the last word is an ambiguous acronym? John Smith, PhD MEng reads first PhD, last John Smith #544): an unambiguous credential in front of an ambiguous member speaks for it
4. 9e37aad test(Should a credential run after a comma keep its comma when the last word is an ambiguous acronym? John Smith, PhD MEng reads first PhD, last John Smith #544): case rows, rules.md examples, corpora and ledgers
5. 0c8d333 docs(Should a credential run after a comma keep its comma when the last word is an ambiguous acronym? John Smith, PhD MEng reads first PhD, last John Smith #544): decision records, release notes and prose sweep
6. 16d835f fix(Should a credential run after a comma keep its comma when the last word is an ambiguous acronym? John Smith, PhD MEng reads first PhD, last John Smith #544): an anchored pick after a family comma reports, and a dual title run keeps its reading
7. a83cb81 fix(Should a credential run after a comma keep its comma when the last word is an ambiguous acronym? John Smith, PhD MEng reads first PhD, last John Smith #544): a part read wholly as suffixes reports no name reading of its words
8. 364cc33 perf(Should a credential run after a comma keep its comma when the last word is an ambiguous acronym? John Smith, PhD MEng reads first PhD, last John Smith #544): the C1 run test ends on a name word before the numeral test
9. 080810c fix(Should a credential run after a comma keep its comma when the last word is an ambiguous acronym? John Smith, PhD MEng reads first PhD, last John Smith #544): a particle anchors nothing, and the anchor pass runs only where a suffix piece is in reach
10. 5c8e8ac perf(Should a credential run after a comma keep its comma when the last word is an ambiguous acronym? John Smith, PhD MEng reads first PhD, last John Smith #544): a member opening a comma part asks no anchor pass
11. a075420 refactor(Should a credential run after a comma keep its comma when the last word is an ambiguous acronym? John Smith, PhD MEng reads first PhD, last John Smith #544): one given-slot anchor query, and a flatter anchor branch

🤖 Generated with Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 98.85%. Comparing base (e10e83b) to head (a075420).

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #552      +/-   ##
==========================================
+ Coverage   98.81%   98.85%   +0.03%     
==========================================
  Files          45       45              
  Lines        3812     3934     +122     
==========================================
+ Hits         3767     3889     +122     
  Misses         45       45              

☔ 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.