Repository navigation
Should a suffix delimiter inside a maiden clause count as a name word beside a link? Smith, John, PhD née Puig Mr. - i Soler keeps i Soler in the maiden name #538
Description
Activity
- added a commit that references this issue
on Sep 27, 2026 Fixed by #550 (merged as 81cd6f7), with the skip reading. A delimiter core is stepped over by the link's neighbour search the way a connective is, so the word on a link's side is the one past the core.
With
Policy(extra_suffix_delimiters=(" - ",)):Smith, John, PhD née Puig Mr. - i Solernow keeps maidenPuig Mr., as the same text without the dash does. It had readPuig Mr. i Soler.Smith, John, PhD née Puig - i Solerstill keeps maidenPuig i Soler.- In a credential tail with no clause, a link beside a core with a credential or nothing beyond it stays a lone suffix word, and the core is dropped (Apply suffix_delimiter only at suffix-consumption sites #206).
Smith, John, PhD - i Solergives suffixPhD, i Soler.
The boundary reading (a core ends the search on its side) was declined. It fixes the same shape, but it would push the birth-name link in
… Puig - i Solerinto the credentials. rules.md#P3 gains a separator sentence scoped to connectives that are also generational vocabulary. rules.md#M2 carries the two examples, and decisions.md's #538 entry records the choice.tests/v2/test_properties.py::test_a_delimiter_core_reads_as_if_it_were_not_writtenpins the invariant over the maiden field. Its recorded control: 6 of 81 texts disagreed before the fix and 0 after.Still open: #549. A core that a connective join merges into its piece survives in the suffix text. That covers a core inside a run (
… Puig Dr. i - y Soler), a core the link joins across (Smith, John, Puig - i Soler), and a core beside an ordinary connective (PhD - and MD). All of these need a declared delimiter, so none is reachable at the default policy.- added a commit that references this issue
on Oct 4, 2026 - added 5 commits that reference this issue
on Oct 8, 2026
With a custom suffix delimiter configured, a delimiter core standing inside a maiden clause can pass as the name word beside a Catalan link, so the link stays in the birth name where the same clause without the core ends before it.
rules.md#M2says a link inside the birth name does not end it where a name word of the clause stands on each side. A delimiter core is not a name word, so the first reading contradicts the rule. A core between the marker and the clause's first word is refused correctly; only one past the first word passes.Reach. Unreachable at the default policy, where
extra_suffix_delimitersis empty, and it needs a title inside the clause beside the link as well. No corpus or case-row name has the shape. Measured over 802,282 parses under 13 core-bearing policies: 288 generated texts movemaiden.The fix threads the segment's cores into the neighbour lookup so a core is skipped like a connective. It is small but touches three call sites. Pinned as it reads today in
tests/v2/pipeline/test_group.py; that test is the one to update.Found in #537, the follow-up to #536.