Read a saint particle glued to a name as one name word (rules/T8) - #1309

Closes #1071.

A saint particle glued to its name was read by its chunks. period_joined_vocab's any-title rule read St.John Price as title St.John with no given name. rules/S3's dotted shape read Mark Jhon Sta.Ana as an unlisted credential, suffix Sta.Ana. #1304 (rules/T8) deliberately left this shape out of its division, and this PR settles it.

The rule: rules/T8, a saint particle glued to a name

A saint particle glued to a capitalized name is one name word: neither a title, nor a credential by its dotted shape. It divides at its stop only when all of these hold:

• it is the last name word of the part;
• the word right before it is a name, meaning no title of any kind and no connective, with no connective earlier in the part;
• the family ends the part: before a family comma under every order, or otherwise under a given-first order.

Then rules/P10's saint run opens the surname, as it does when written apart. str() keeps the word as written.

Why not divide everywhere, as T8's other division does: written apart, Jeremy St. John Middleton reads family St. John Middleton. That would move 16 British names where St John is a middle name into the family name; the glued writing reads them right today.

Why divide as the last word: in the Filipino writing, dela Cruz before the surname is the mother's surname, and it reads as the middle name only when the particle opens the surname.

A maiden clause of one word or two (née, z domu) now ends a part's own words, both for this division and for T8's end-of-name report.

Measured against master 3744aa0

• Population: 79 glued saint words in the 13 local corpora. Before a comma: 37 family parts alone, 20 last words, 16 middle words, 2 behind titles only and 2 first words. After a comma: 2.
• Moves (snapshot.py --local <corpus> --base <master>):
  • psc-individuals 14, including 9 initials-only moves matching the spaced writing (St.Hill → K. C. H.);
  • orcid-researchers 8;
  • fec-individuals 2;
  • patentsview-inventors 2;
  • ctgov-officials, tse-donors and ukec-individuals none.
• Committed snapshot: no corpus name moves.
• Accepted costs (recorded in decisions/T8):
  • Juan Sta.Maria Reyes reads middle Sta.Maria, where Juan Sta. Maria Reyes reads family Sta. Maria Reyes. That's the price of keeping the British middle names whole.
  • A name written wholly in lower case keeps its chunks' readings (st.john price).

Review

A design-docs review of code and docs found the first build dividing in two places P10 doesn't read:

• behind a title only: Mr. St.Clair became title Mr. St., family Clair;
• under the family-first orders: Price Mark St.John got two initials where Price, Mark St.John has one.

It also found a suffix word after the stop splitting (St.MD), the two-word maiden marker missed, and a population count that didn't add up. All are fixed, each with a test row.

A focused review of those fixes then found three more, all fixed with test rows:

• a title phrase's words (Mr and Mrs, The Hon., Dean of Students) passed as a name before the word;
• the order gate applied before a family comma;
• a maiden marker ending the text counted as a clause.

The branch merged master after #1307 landed (one conflict in decisions/T8's Standing block, both lines kept). The full suite, mypy and ruff pass, and the corpus figures hold against the new master.

🤖 Generated with Claude Code