`capitalized()` lowercases a word that `family_base` and `initials()` treat as an ordinary name (`ANH DO` -> `Anh do`)

[derek73]

capitalized() lowercases the same word that family_base and initials() now treat as an ordinary name word:

r = Parser().parse("ANH DO")
r.family_base                     # 'DO'   -- an ordinary name word
r.initials()                      # 'A. D.'
str(r.capitalized(force=True))    # 'Anh do'   <- still a particle

Compare an ordinary surname: parse("anh smith").capitalized() gives family Smith.

Why

_cap_word (_render.py) keys the lowercasing on lexicon membership plus role:

if ((normalized in lex.particles and role in (Role.MIDDLE, Role.FAMILY)) or ...):
    return word.lower()

It does not consult UNJOINED_TAG, so the same word is an ordinary surname for R2 and R3 and a particle for R4 within one name.

Scope

Introduced by #404 in the sense that #404 is what declares these words ordinary name words; _cap_word itself is untouched. capitalized() already maps over tokens with t.tags in hand, so threading the mark in is mechanical.

Worth deciding rather than assuming: lowercasing a tussenvoegsel is a display convention for a particle doing particle work. A surname that happens to be particle vocabulary — Do, Del Toro, Le — is not that, and Del toro is not how anyone writes it. Either R4 consults the mark, or R4 should state that it deliberately does not.