Decide where the name starts once, for title-particle heads (#624) - #633

Summary

freiherr and st are both titles and particles (TITLES ∩ particles_ambiguous). Such a word in the leading titles is the particle chain's leading position (rules.md#P4), so the particle behind it is inside a name (Freiherr von Berg → family 'von Berg'). Three places each decided where the name starts, and they disagreed when two such words stood together:

• Group's chain used a scan of its own that stopped at the first title-particle, and chained the second as a surname particle.
• Assign read both as titles.
• The trailing read's unit count (_chain_units, Should S2's trailing read count the particle chain's units instead of merging a copy of them? #620) opened a unit inside the titles, to follow the chain.

Decided by Derek: consistency and one shared answer, no special case for garbage.

• One answer: _pieces.chain_lead gives the leading position once. It's the last title in assign's own title run that is also a particle, else the first piece past the run. Group's chain and _chain_units both take it.
• Titles as written: the run is read before H3's give-back. A title handed back to the name because only suffix words follow it is still a title to the chain, so the particle behind it leads and chains nothing (P4). leading_titles stops at a leading title only where it gave that piece back, so chain_lead recognises the give-back from n alone, behind the same inline tag test.
• Less machinery: no particle stands between the lead and the run's end, so the chain opens no unit inside the titles. _chain_units loses its at answer and its q > k + 1 gate, and group's chain drops its own scan.
• Group's chain skip: it now skips every piece up to the lead (k <= leading). That's equivalent to the old k == leading wherever the lead was the first title-particle.

Not done: the issue also proposed retiring assign's re-read fallback. Measured, it still does real work: disabled, sixteen or seventeen more tests fail (depending on how it's disabled), most of them crashing with an IndexError. So it guards totality, not just a reading. Among them is a_title_read_before_the_chain_can_strand_the_name (Freiherr von Berg Dr. and Ed. Prof., a connective join that makes the name a title). It stays.

Measured (decisions.md#P2, 2026-10-10)

Against master 8e6e524, py3.11, nameparser.__file__ asserted on each side:

• Title-particle head grid: every 1–4 word sequence over Freiherr St Dr. Sir Xyz. von van de do Mc John Smith MA PhD VI Prof. née Jr. and, plus comma forms. That's 137,810 names under three orders. Result: 5,964 role moves, and none in report text only.
  • 5,868 have two title-particles in the leading title run, now all titles. Where H3 then gives the last one back, the word behind it is the post-nominal H3 saw (Dr. Freiherr St Mc → family 'St', suffix 'Mc'; master family 'St Mc').
  • 96 are H3's give-back before dual particles, read as H3 states (Dr. Mc Mc → given 'Dr.', suffix 'Mc Mc'). Master gave given 'Dr.', middle 'Mc', family 'Mc', because the read counted the second 'Mc' as bound into a run the chain never built.
• Comma-tail grid (97,650 texts, recipe in the entry): 1,368 parses on 456 texts move. Every one is a part past the second comma led by two title-particles, now all titles as C2 reads them (,, Freiherr Freiherr Bart → title 'Freiherr Freiherr', suffix 'Bart').
• S2 stress grid (Rethink the credential-run rule (S2): read a name's trailing run once, over name units, and bind it before any join #614's recipe, 324,552 parses): 0 moves.
• Corpus names and case texts: only the case rows rewritten here move.
• Five differential gates: all exit 0, with the new rules.md example explained by a fix(#624) rule in each ledger.
• Frames: every baseline moves down by 4. parse 308 → 304 on 3.11 and 289 → 285 on 3.12–3.15; facade 345 → 341 and 326 → 322. Re-recorded in _CALL_BASELINE and decisions.md#parse-cost.

Tests

• Five case rows pinned the old decision and now read the new one (fix(#624), each note giving the earlier reading): Freiherr St van Berg MA, Freiherr Freiherr Prof do, St St VI, Freiherr St MA, Freiherr St Prof..
• New rows: Freiherr St John Smith MA, the issue's example, and Dr. Mc Mc, which pins the give-back half.
• Moved row: the one holding group's particle-or-given emitter goes from St St née to St van née (parity). It's the same structure with a chained particle that is not a title, since St St née no longer reaches that emitter.
• test_pieces.py: _chain_units returns units only. test_the_chain_opens_no_unit_inside_the_titles replaces the test of the old design.
• The field sweep checks 1149 claims (was 1158): the rewritten rows lost their wrong title-or-name reports, and Dr. Mc Mc adds three. Its control was re-measured against 26cdb89 over today's rows: still 14 failures, with Freiherr St John Smith MA in place of St St née.

Docs

• rules.md#P4: one clause (the last title-particle in the leading titles leads, read before H3's give-back), plus the example pair Freiherr St John Smith → title 'Freiherr St', given 'John'. interacts: adds H3; implemented: adds _pieces.py.
• rules.md#H4: the Accepted example moves from St St née to St van Bishop.
• decisions.md:
  • the Where does Freiherr St John Smith MA start: should a title that is also a particle open the name? #624 entry under P2;
  • a parse-cost bullet;
  • a pointer from Should S2's trailing read count the particle chain's units instead of merging a copy of them? #620's "LEFT OPEN" line;
  • "superseded in part" notes where Rethink the credential-run rule (S2): read a name's trailing run once, over name units, and bind it before any join #614's and Should S2's trailing read count the particle chain's units instead of merging a copy of them? #620's entries describe the removed mechanism.
• No release-log bullet: the inputs are malformed, and no corpus name moves.

Review

• c03f61e: the change.

• 07957c5: fixes the review:

  • Option (a), Derek's call: the chain reads the titles as written, before H3's give-back. The first draft read the give-back as the lead, so Dr. Mc Mc chained its leading particle behind a plain title (title 'Dr.', family 'Mc Mc'), against P4, and a caller lexicon could make group's emitter fire behind a plain title.
  • Also: the H4 example, the Rethink the credential-run rule (S2): read a name's trailing run once, over name units, and bind it before any join #614/Should S2's trailing read count the particle chain's units instead of merging a copy of them? #620 pointers, stale comments, P4's wording ("the last"), the field-sweep docstring, and the fallback's description.

• 12ef837: the fix commit's review nits: chain_lead's docstring and the decisions entry note that leading_titles also stops at a segment's last piece, P4 says a title-particle reads as a title unless H3 gives it back, and the five ledger comments quote P4's current wording (they had kept the first draft's).

Closes #624

🤖 Generated with Claude Code