[derek73]

Part of #761.

This PR moves S1–S9 from the block format to the skeleton format, one commit per rule. S10 already was. It is the last section. After it, a final PR retires the block format, its parser branch and axis 13.

No rule changes meaning, and every example line is kept verbatim. Each rule's example multiset matches master, corpus_rules.jsonl does not move, and the suite, mypy and ruff pass. Each commit re-records its rule's row of tests/v2/rules_budget.json.

Growth worth a look

S2 (a suffix word ends a name and never opens the string), the hub, grows from 357 to 586 counted words. Almost all of it is new Precedence bullets. These collect the order among S3–S10 that their own texts already state, the way C1 did:

• S3: periods make the acronym unambiguous.
• S4 / S5 / S6 / S7: case against count, the given part's slot, and company.
• S7 with S8: the run takes the title-between and particle-chain shapes.

The bullets also hold what S2's own statement said about opening the string, the family comma, the ordering against P3/P2/P5, and H3's floor. The title chain's history moves out of the rule into a dated decisions/S2 bullet. That history is how Dr King Jr used to read, and the Dr Smith Jr contrast.

Decided

Derek answered the four questions on 2026-10-20 ("i agree with your recommendations"). None of these moves a reading.

1. S3 (credentials written with interior periods) against S8 (the credential's trailing run): S3 wins. On John Smith M.A. Jones, S8 said a member of the ambiguous class never starts the run, and S3 said a word written with its periods counts unambiguously. The parser already read suffix M.A. Jones. Derek: "writing periods makes it unambiguously not a name". S8 now has a Precedence bullet saying so, scoped to the periods. Capitals alone still do not start the run (John Smith MA Jones reads middle Smith MA). That case, and the report on John Smith MA PhD, → Should capitals in a mixed-case name settle a credential the way periods do? #785. A dated bullet in decisions/S8 records the decision.

2. S10 (the trailing roman numeral) against S8: no change here → Should John Smith VI MD read like John Smith III MD? #784. The run start asks the vocabulary, so a listed numeral starts the run and a numeral read by shape alone does not:

   • John Smith III MD reads suffix III MD.
   • John Smith VI MD reads family VI, suffix MD.

   The issue will also ask why the numerals are recognized two ways.

3. S6 (the case signal at the suffix slot): "a comma", not "a family comma". S6's first-slot Precedence lead now says what S4 says and what the parser does: John Smith, Ma reads suffix Ma. A dated bullet in decisions/S6 records the change.

4. Two clauses stay without examples, by decision. No corpus rows are added for shapes nobody types:

   • S7's maiden-clause limit. The only input that reaches it is Jane PhD nee Smith Ma.
   • S3's clause about scripts that write no abbreviations.

Old interacts: IDs

• S1: M2. Kept, in a Precedence bullet with M3. M3 states that S1 takes a marker-led clause first.
• S2: H1, H2, H3, H4, H5, H6, C1, C8, C13, S3–S10, P2, P3, P4, P5.
  • Kept: H2, H3, H4, H6, C13, P2, P3, P4, P5. Also S3–S10, in the Then map, with S5, S6 and S7 in Precedence too.
  • Dropped: H1, H5, C1 and C8, which the old text never named. H5's Precedence states its own meeting with S2.
  • New: C4, carried in with S4's case-and-count precedence.
• S3: C4, C5, C11, H2, W3, S2, S4.
  • Kept: C5 and C11 (Precedence), H2 (Then bullet), W3 (Accepted) and S4.
  • Dropped: C4 (never named; C4 cites S3) and S2 (the hub names S3).
  • New: M2, the maiden clause slot, which the old text named in prose.
• S4: C1, C2, C4, M2, S2, S3, S5, S6, S7, S9.
  • Kept: C2, C4, M2, S5, S6 and S7.
  • Dropped: C1 (never named; C4 is the comma's count), plus S2, S3 and S9, which cite S4 rather than the reverse.
• S5: C1, C11, H5, P1, P6, S4, S6, S7.
  • Kept: H5, P1, P6, S4, S6 and S7.
  • Dropped: C1 (never named) and C11, which names S5 as its other report site; S5 never named C11.
• S6: C4, C11, S4, S5, S7.
  • Kept: C4, C11, S4 and S5.
  • Dropped: S7, which states its own rank over S6.
• S7: C4, C5, C11, M2, P2, P3, P6, S4, S5, S6, S8.
  • Kept: M2, P2, P3, P6, S4, S6 and S8.
  • Dropped: C4, C5, C11 and S5, none named in the old text.
• S8: H5, M2, P2, P3, P5, P6, S2, S4, S7, C8, C13.
  • Kept: P2 and P6.
  • Dropped: H5, M2, S7 and C8, never named; P3, P5 and S4, whose subjects (connective, bound given-name vocabulary, ambiguous class) the text names in words without the ID; and S2 and C13, which cite S8.
• S9: C4, C5, C6, C11, S3, S4. All kept.

Code comments

• nameparser/_pipeline/_vocab.py: S3's chunk-rule citation now chains two quotes, since the single-character exception became an Except bullet.
• tests/v2/cases.py: two notes point at decisions/S2's new bullet.

No other code quote moved.

Review

The docs-review agent ran (axes 1–13). It found nothing high-severity and three low findings: S1's When was too narrow, an S3 Except row showed Then holding, and an S2 Accepted bullet's "this clause" had lost its referent. All three are fixed in c7c177c, and the reviewer re-checked that commit. Its fourth finding is Decided item 3. Derek's decisions landed in a later commit, which the reviewer also checked.

🤖 Generated with Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 99.05%. Comparing base (713d2b7) to head (c3d87e1).
⚠️ Report is 1 commits behind head on master.

Additional details and impacted files

@@           Coverage Diff           @@
##           master     #783   +/-   ##
=======================================
  Coverage   99.05%   99.05%           
=======================================
  Files          47       47           
  Lines        4642     4642           
=======================================
  Hits         4598     4598           
  Misses         44       44           

☔ 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.