[derek73]

Closes #959. This PR covers what #961 left Open in decisions/O6 (two family names): the hyphenated-surname gap and the both-switches example.

Hyphenated surname taken

Parser(policy=Policy(two_family_names=True)).parse("Andrés Sánchez-Cid Torres") gives family 'Sánchez-Cid Torres'. 2.6.0 through 2.8.0 kept 'Sánchez-Cid' a middle name, because rules/O6 counted the hyphen's two words against the word it takes. As decided:

• A hyphenated word counts as one name word when it is the word O6 takes.
• A hyphen in the family still counts its words, so Javier Olmedo Pérez-Montaut is unchanged.
• The veto reads each part. A hyphenated word every part of which is a listed given name or an initial stays where it is:
  • Ana María-José López under the Spanish list keeps middle 'María-José';
  • Ana M.-J. López stays under any lexicon.
  • Without a list, Ana María-José López reads family 'María-José López'. That is recorded as the switch's usual cost.

Measured on the cached ORCID researchers. Countries ES, MX, CO, PE, CL, VE and AR: 135,735 records, of which 10,794 distinct records hold a hyphen. Under locales.ES:

reading	right	fixed	broken
master	10,281	—	—
hyphen counts as one everywhere	10,038	277	520
taken word only, with the veto (this PR)	10,464	186	3

• The "everywhere" variant loses because Latin-American family boxes write two surnames with a hyphen.
• This PR fixes ES 125, MX 24, CL 20, PE 6, CO 5, VE 5 and AR 1.
• With the switch alone, the same records move.

Accepted: a hyphenated bound pair (Ana Abdul-Rahman López) is taken like any hyphenated word. A veto for it was built and backed out: it cost 2 ORCID records whose hyphenated word is a surname (Abdul-Hadi Pru) and kept none.

Both switches

rules/O7's precedence over O6 now shows María del Carmen de la Vega:

• family 'del Carmen de la Vega' with both switches on;
• given 'María del Carmen' with O7 alone.

No behavior moves.

Docs and interaction

• Updated: rules/O6 and O7, decisions/O6 (an entry, the Standing line, Open items removed), Policy.two_family_names' docstring, and a release-log bullet.
• With Cut a family-first particle chain where P1's fold cuts it under two_family_names (#1024) #1049 (Under a family-first order, should Pérez de la Vega Juan read given 'Juan'? The particle chain takes the given name, and two_family_names can't take it back #1024) merged too, Pérez de la Vega-Santos Juan under family-first will read given 'Juan'.

Review

An independent review found:

• the stale docstring;
• a wrong prototype figure (10,040 should be 10,038);
• a missing dedup step in the recipe;
• the release range;
• the bound-pair and initial-plus-name spellings.

All are addressed in the second commit.

Checks

• uv run pytest: 36,086 passed. Ruff and mypy clean.

🤖 Generated with Claude Code