fix(group): a wholly-credential family-comma segment renders as one entry (#429)

[derek73]

Closes #429.

A space-separated credential run after a family comma rendered with a comma the name never had. The roles were already right — this is #428's remaining half.

parse("Smith, MD PhD").suffix       # 'MD, PhD'  ->  'MD PhD'
parse("Smith, CBE MC").suffix       # 'CBE, MC'  ->  'CBE MC'
parse("Smith, Dr. MD PhD").suffix   # 'MD, PhD'  ->  'MD PhD'

The disagreement

One comma segment is one suffix entry. group decided which segments those are by index — tail_start is 1 after a suffix comma, 2 after a family comma — while assign decides the same segment by content, with _segment_holds_no_name, and routes it piece by piece (#296 / #325).

Those two answers agree everywhere except the segment a family comma leaves behind. There, group saw a name slot and assign saw a credential run, so the run was routed whole and rendered split.

Scope, as the condition actually reads

Any family comma whose following segment holds no name word — not only a one-word family, which is how the first draft described it:

• John Smith, Jr. III → suffix Jr. III (also 1.4.0's reading)
• Smith, Dr. MD PhD → suffix MD PhD (title-led segments are in)

The differential cannot witness the widening: of 1069 corpus names, 303 carry a structure comma, the predicate is true for 36, only 9 of those have ≥2 pieces in the segment, and none combines a multi-word pre-comma name with two suffix pieces. 0 unexplained is evidence about the corpus, not about this class — the case table carries it instead.

The review round found a regression, and it was the inverse of the bug

The first draft let any piece open an entry, as the tail block always had. Safe there — tail means assign routes every piece to SUFFIX. Wrong off it, where a title piece routes to TITLE. Two failures, one cause:

"Smith, Rev. Dr."     title_list  ['Rev.', 'Dr.']  ->  ['Rev. Dr.']
"Smith Jr., Mr. Jr."  suffix      'Jr., Jr.'       ->  'Jr. Jr.'

The second glues a suffix backward across a comma the writer typed — exactly what this PR exists to stop. The joined tag is role-blind and the facade heals it for every role, which is why a suffix-motivated tag corrupted title_list.

Neither was visible to any gate that passed. The differential compares strings and cannot see a list-view collapse; the case table asserts the title string, which is space-joined either way.

The rule that replaces it separates two joins that had been one:

• within a piece, the tag renders a merged piece as one unit, whatever role it holds
• between pieces, it continues an entry — and only a piece that renders into the same run may do that

Sticky across a piece that is not in the entry, so an interleaved title does not split its run (Smith, MD Dr. PhD → MD PhD). A delimiter core still closes it.

Verification

Gate	Result
uv run pytest	5676 passed, 221 skipped, 10 xfailed
uv run mypy / ruff	clean
differential 1.4.0 / 2.0.0 / 2.1.0	213 / 173 / 80 intentional, 0 unexplained each

Both regression guards were checked against a mutation copy with the old condition restored — they fail, so they are not inert. test_the_joined_tag_never_reaches_a_title asserts the list views, the only surface that shows the title collapse.

Eight case rows added for classes the mutation matrix found unpinned: title-led runs (384 inputs, previously zero coverage), interleaved titles, three-entry runs, the seg_idx == 1 conjunct, the written-comma negative control, and the full-name twin.

Prose corrections from review

All measured, all in the diff:

• The round-trip claim was false and backwards. str() of a fixed parse is a no-comma string, which re-parses with the comma back; master was the str-stable one. Struck.
• #191 was the wrong provenance for the delimiter parity — that is the German/Dutch vocabulary PR. It is Apply suffix_delimiter only at suffix-consumption sites #206 (021823e). Three existing code comments carried the error and are corrected.
• The dormant-rule tell is Make a ledger rule that explains nothing say why (#372) #373's, and refactor(group): one trailing peel for assign and the bound-given reserve (#425) #426 the precedent for dropping a shadowed rule; neither John van der Berg V reads family van der Berg V — the particle chain and the maiden walk take a trailing numeral as a name word #424 entry mentions it.
• "boundary example" in the entry and both ledgers — the example fires, which is why the annotation came off.
• "filed rather than folded in" claimed an issue that does not exist.

C1 gains _group.py in implemented:, with the verbatim citation the equality guard requires.

Out of scope, measured

The no-comma path comma-joins a run too — John Smith MD PhD renders 'MD, PhD' at 1.4.0, 2.0.0, 2.1.0 and master alike. The fix requires FAMILY_COMMA and does not reach it. Recorded in decisions.md and not yet filed; it is load-bearing for the round-trip story above.

🤖 Generated with Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 98.59%. Comparing base (d6087d9) to head (11a486d).

Additional details and impacted files

@@           Coverage Diff           @@
##           master     #435   +/-   ##
=======================================
  Coverage   98.59%   98.59%           
=======================================
  Files          44       44           
  Lines        3061     3064    +3     
=======================================
+ Hits         3018     3021    +3     
  Misses         43       43           

☔ 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.