Turn the pack-shaped exclusions into readings off by default (#753, 3c)

docs/design/decisions/generational-words.md

@@ -10,7 +10,7 @@ Closes #661. 31 spelled-out generational words join SUFFIX_WORDS, and one predic
 - A refused word carries `excluded = { generation = … }` with its reason in the file of the language whose reading is refused (`uv run python tools/vocab/query.py --excluded generation`) — 2026-10-14 #661; 2026-10-14 #679.
 - The words ship spaced only, `二世` excepted, so a hyphen-glued Jr. is no suffix: `Теодор Рузвелт-младши` reads family 'Рузвелт-младши' — 2026-10-14 #661.
 - `二世` is a generation word that peels spaced and glued, the one exception to spaced-only shipping: `陳怡君二世` reads family '陳怡君', suffix '二世', and its `ja.toml` and `zh.toml` entries carry `issue = 671` and `attached` with `issue = 707` — 2026-10-21 #762.
-- The generational words whose collisions are with OTHER languages written in the same script (fr `père`; the Latin abbreviations `ml.` and `st.`) ship in no default, that being the case a locale pack exists for: `Jiří Dienstbier ml.` reads family 'ml.', and no cs pack exists; cs `st.` is recorded as a reading off by default, for that pack — 2026-10-14 #679; 2026-10-25 #752.
+- The generational words whose collisions are with OTHER languages written in the same script (fr `père`; the Latin abbreviations `ml.` and `st.` in cs, hr, sk and sl) ship in no default, that being the case a language choice exists for: each is a reading written `default = false` in its language's file, so `Jiří Dienstbier ml.` reads family 'ml.' by default and suffix 'ml.' with `cs` chosen — 2026-10-14 #679; 2026-10-25 #752; 2026-10-29 #753, phase 3c.
 - fr `fils` is a generation reading gated by its family reading, its case and the count (rules/S11): `Alexandre Dumas fils` reads suffix 'fils' and `Marie Fils` family 'Fils', both reported — 2026-10-25 #752.
 
 Elsewhere: the run start that lets any unambiguous suffix word of two or more letters begin the credential run, whatever its script (`Antonio Pavlović mlađi Jones` reads suffix 'mlađi Jones', as does `Pieter Bruegel stariji Jones` with 'stariji Jones'), is decisions/S8's 2026-10-14 (Derek), #679 entry, which retracted this file's #602 run-start bullet; the glued peel of 二世 is decisions/W2's 2026-10-17 #707 entry.
@@ -38,3 +38,5 @@ Open: #684 (locale packs for `Jiří Dienstbier ml.` and `st.`, the generational
 - **2026-10-23 #843 (Derek) — JÚNIOR SHIPS; THE AGNOMENS WAIT.** "Suffix, Júnior only for now." `júnior` joins `pt.toml` as a generation reading with `same_as = "jr"`: the standard Portuguese spelling, which `José Silva Júnior` read as family 'Júnior', middle 'Silva' in 1.4.0 and 2.5.0 (checked against both wheels) while `José Silva Junior` read suffix 'Junior'. Its collisions are `junior`'s, already accepted: a given name (Wikidata 37+ as Júnior, 43,182+ US SSA births as Junior), where the suffix claim does not act (`Júnior Silva` keeps given 'Júnior'), and family bearers (Wikidata 49+, US Census 2,146 as JUNIOR), which a suffix word costs its surname wherever it ends a name (`Pedro Júnior` reads given 'Pedro', suffix 'Júnior', as `Pedro Junior` does), counts from `tools/name_usage.py` on 2026-10-23. In the local corpus `tse-donors` (fetched 2026-10-22), 1,071 of its 854,531 records write it anywhere in the name (`tools/name_usage.py`), 991 distinct filed names end in it, and 983 records end in it, in either case, where the tax registry's name for the same donor ends `JUNIOR` (last words compared after `str.upper()`). The accented spelling reads as the unaccented one in all 120 pairs of a grid of 20 shapes (comma shapes, an empty head, a title, a particle, a trailing title, a nickname, a maiden clause, a credential and a name word behind it, both one-case spellings) under all three orders, composed and decomposed, comparing fields, ambiguity kinds and `initials()`; with `júnior` removed from the lexicon (`Lexicon.default().remove(suffix_words={'júnior'})`), 108 of the 120 pairs differ, and the 12 that do not hold Júnior where no suffix reading reaches it (`Júnior Silva`, and `Júnior, José`, before a family comma). Measured 2026-10-23 on `tse-donors` against the tree before this change (`snapshot.py --local tse-donors`, baseline recorded with `--tree`): 993 of 848,351 filed names move, 990 ending in Júnior and gaining it as the suffix out of the family name (3 of them two-word names, `RONALDO JÚNIOR`, left with no family as `RONALDO JUNIOR` is), and 3 with Júnior inside the name, where it starts a credential run (decisions/S8) that takes the surnames behind it, `JORGE ANTÔNIO JÚNIOR BARBOSA SILVA` reading family 'ANTÔNIO', suffix 'JÚNIOR BARBOSA SILVA' as the JUNIOR spelling already did; and `local_corpora.py pairs tse-donors --agreement` stops listing JÚNIOR / JUNIOR (866 accents-only pairs read with two surnames before, none after). The agnomens (distinct filed names ending in them: `Filho` 10,979, `Neto` 8,439, `Sobrinho` 955, `Netto` 432) stay family names for now. They are legally part of the surname, and `Neto` is a borne surname and given name alone (Wikidata 276+ family, 12+ given), which a suffix reading would leave with no family, as `Pedro Junior` has none. No exclusion records them, because nothing was refused: the question is open on #843.
 
 - **2026-10-25 #752 — THE AGNOMENS AND `fils` ARE GATED GENERATION READINGS (phase 2h).** SUPERSEDES: 2026-10-23 #843's agnomen half, which kept the agnomens family names, and 2026-10-14 #679's refusal of `fils` as pack-shaped. The 2026-10-23 #752 entry in decisions/vocabulary-model gave each generation reading a gate by its kind, and the record, with its measurements, is decisions/S11. `fils` collides inside French, a generation word and a French surname, so no French pack could settle it, which is the premise of #684 the design reversed: its `generation` reading now stands beside its `family` reading, and the gate reads the name around it. The agnomens are not part of the surname: an agnomen marks its bearer against a relative of the same name, as Júnior does, and its name readings (dated counts in `pt.toml` and `pt-BR.toml`, `netto` there being the spelling of Brazil's older orthography) keep a two-word name's family. cs `st.` keeps no exclusion: it is a `generation` reading written `default = false`, recorded for a Czech pack, since nothing in its writing tells it from Saint; hr, sk and sl keep theirs, as do `ml.` and `père`.
+
+- **2026-10-29 #753, phase 3c — `ml.`, `st.` AND `père` ARE READINGS OFF BY DEFAULT.** The exclusions #679 recorded as pack-shaped are `generation` readings written `default = false` (cs `ml`, hr, sk and sl `ml` and `st`, fr `père`; cs `st` since phase 2h), read when their language is chosen, and the default does not move. The record, with what each choice moves, is decisions/vocabulary-model's 2026-10-29 #753, phase 3c entry.

docs/design/decisions/vocabulary-model.md

@@ -12,6 +12,7 @@ The vocabulary was a set of base sets plus marker subsets (`particles` and `part
 - A patronymic field is to come in 3.0, as a view that names its words without emptying `family`, and the particle kinds already draw the `clan`/`patronymic` line for it: a patronymic is followed by a parent's name, a clan prefix by a family name the lineage shares — 2026-10-15/16; 2026-10-19 #749; 2026-10-28 #754 (Derek).
 - A language pack is a choice of languages on `Locale.languages` selecting which readings compile, a tag choosing its own file and its ancestors', and a reading index, loaded on its first lookup, carries the languages to run time; strict waits for 3i, and third-party language files are deferred — 2026-10-27 #753 (Derek); 2026-10-29 #753, phase 3a (Derek); 2026-10-29 #753, phase 3b.
 - A choice is matched without regard to case, a tag whose file does not exist selects its ancestors' (`pt-PT` selects `pt`), `mul` and `und` are no language to choose, and a reading turned on also prefers, so `de`'s generation `i` drops `ca`'s and `pl`'s conjunction under a `de` choice; a choice is one call's: one chained through `base=` applies on top of the base's rather than unioning with it, and a choice drops the caller's own base entries it removes — 2026-10-29 #753, phase 3b; 2026-10-29 #753, phase 3b, chained choices (Derek).
+- A reading right in its own language and wrong only as a default, because another language written in the same script collides with it, is written `default = false` in its file, never excluded, so that after the packs an exclusion means wrong everywhere: cs `st` since phase 2h, and cs `ml`, hr, sk and sl `ml` and `st` and fr `père` since 3c, the rest in their packs' PRs — 2026-10-27 #753 (Derek); 2026-10-29 #753, phase 3c.
 - A name reading written `default = false` derives no marker in the default and counts for no check guarding the default's compile; a particle whose word's only `given` readings are off is refused until a pack needs one — 2026-10-29 #753, phase 3a.
 - A caller configures vocabulary by reading, `Lexicon` deriving the sets and gates the build derives, and the compiled sets become internal in 3.0 — 2026-10-28 #754, the `Lexicon` surface (Derek).
 - A `Locale` may carry CLDR-shaped render patterns, display and sorting, a later addition whose shape is agreed; how a call reaches one is open — 2026-10-28 #754, render patterns.
@@ -312,3 +313,10 @@ Open: #913 (the title words phase 2j left, read as titles though borne as names)
   1. **Chained choices compose one after the other.** `parser_for(EN, base=parser_for(PT))` reads `Maria do Carmo` family `Carmo` and `John Smith DO` family `DO`: `en` drops `pt`'s particle and `pt` had dropped `en`'s credential. The way to choose both is one call, `parser_for(PT, EN)`, which keeps both readings (`do Carmo`, suffix `DO`).
   2. **A choice drops what it removes from the base, whoever put it there.** `parser_for(EN, base=Parser(lexicon=Lexicon.empty().add(particles={'do', 'van'})))` keeps only `van`: the delta removes `do` from `particles`, the caller's listing included, and leaves every entry it does not touch.
   Weighed against: (b) a `Parser.languages` field recording the choice, so a base's choice unions with the packs' (a new public name, safe for pickles, `Parser`'s guard honoring `ADDED_IN`), deferred rather than declined; (c) a `Lexicon` field recording it, which every older `Lexicon`, `Parser` and pack pickle would refuse, `Lexicon`'s guard not honoring `ADDED_IN`. Open: (b), with the configure-by-reading method names (the 2026-10-28 #754 `Lexicon` surface entry), and with it whether `Reading.languages` follows the choice (the phase 3b entry above).
+
+- **2026-10-29 #753, phase 3c — THE EIGHT PACK-SHAPED EXCLUSIONS ARE READINGS OFF BY DEFAULT: cs `ml`, hr, sk and sl `ml` and `st`, fr `père`; THE DEFAULT DOES NOT MOVE.** Carries out the 3c half of item 9 of the 2026-10-27 #753 entry, and 3c of its item 11.
+  - **The split.** Every exclusion whose reason opened "pack-shaped" refused a `generation` reading that is right in its language and wrong as a default only because another language written in Latin script collides with it: `ml.` against the initials `M. L.`, `st.` against Saint, `père` outside French. Each is now a `generation` reading written `default = false` in its own file, with the issue list it had and #753, an `expands` naming the full word, and a `note` giving the collision, the shape cs `st` has had since 2026-10-25 (phase 2h). The other exclusions a pack turns on (`pan`, `roz`, `nic`, `shri`, `sig`, `jaun`, `bt`, `bay`, `sb`) stay exclusions until the PR of the pack that turns each on; every other exclusion is wrong everywhere and stays. Recompute: `grep -l 'pack-shaped' nameparser/vocab/*.toml` lists no file, which tests/v2/test_locales.py's `test_no_exclusion_is_pack_shaped_any_more` holds.
+  - **The default does not move.** The build reports `_compiled.py` current, only `_index.py` gaining the eight rows (11 readings written `default = false`, from 3), and tests/v2/test_snapshot.py passes with no re-record.
+  - **What choosing each language now does,** measured 2026-10-29 on `b27dadd5` plus this change with `snapshot.py --languages TAG`, entries moved (names held only by `corpus_cldr.jsonl` among them): `cs`, `sk`, `sl` and `hr` 14 each (4), where `cs` moved 11 (2) and the other three none before; `fr` 0, no committed name holding `père`. The three new entries under each read `ml.` as the suffix: the rules corpus's `Jiří Dienstbier ml.`, and two CLDR renderings with `ml.` before a comma of credentials (`g. Janez Peter Novak ml., univ. dipl. inž.`). The other eleven are `st`, and each language's turning on its reading also prefers it (the phase 3b entry's turning-on bullet), so en's and fr's title and particle readings of `st` leave: `St Clair` reads given `St`, family `Clair`, where the default reads title `St`. That cost reached `cs` with 3b; it reaches `sk`, `sl` and `hr` here, and settling it, whether a Saint reading should outlast a choice that holds `st` as a generation, is 3f's, which turns those packs on. tests/v2/test_locales.py's `test_a_chosen_language_reads_its_off_generation_and_the_default_does_not` parses every generation reading off by default of more than one letter (`Anna Novak ml.`, with and without the period) under its language and the default; a one-letter word (es `h`, de `i`) is left out, the initial veto taking it written with its period (`Anna Novak h.` reads family `h.` under `es`, while a bare `h` reads suffix `h`).
+  - **Open for 3f: a turned-on `ml` reads a lone comma part as the suffix, silently.** `ml` has no name reading, so under a choice it is an ungated suffix word: `Novak, ML` reads suffix `ML` with nothing reported, where the default reads given `ML`, and `Novak ml., Anna` family `Novak`, suffix `ml.`; the gated `st` reports `SUFFIX_OR_NAME`, and fr `Jean Père` reports `GIVEN_OR_FAMILY` (suffix `Père`). The initial pairs `Anna M. L. Novak` and `Anna ML Novak` do not move. Pinned in tests/v2/test_locales.py; whether `ml` takes a name reading or a gate is 3f's. So is the one-letter gap: es `h` and de `i` turned on stay inert written with a period (`Juan Pérez H.` keeps family `H.` under `es`, `Wilhelm Karl I.` family `I.` under `de`), a strict expected failure there.
+  - Not touched: no `Lexicon` or `Parser` field, no code. The vocabulary README's example of an exclusion is now `duke`, wrong everywhere, where it was `père`.