rero / rero-mef / 34334196048

87%

staging: 87%

push

github

feat(concepts): link on the normalised BNF number

IdRef and GND write the same BNF number in incompatible ways, with or
without the trailing check character, and the identifier was compared as
stored. Several records can also claim one number, which the old rules
took as candidates to choose from rather than the related concepts they
are: sixteen camera models close-match the IdRef concept
`Canon (appareils-photo)`.

* reduce every BNF number to `FRBNF` plus its eight digits, and settle a
  contested number with the single exactMatch record
* identify on `bf:Nbn` only, never on a BNF ark uri, and warn when the
  two name different BNF records: GND `Sieg` close-matches "Victoire"
  with the ark of `Victoire` and the number of `Épée (sport)`, and the
  number is what decides. Fixing those belongs in the source
* let an IdRef heading state every number it holds and match a GND
  record whose number is among them: a RAMEAU heading that absorbed
  several BNF records keeps all of theirs, and demanding exactly one
* keep every `033` ark and every `035 $z` number in `identifiedBy`
  instead of the latest of each: two arks sharing one `$d` overwrote
  each other and the `$z` fields carry no date, so the surviving number
  depended on field order
* keep closeMatch as the fallback when no exactMatch carries a BNF
  number, and take it only when the closeMatches name exactly one
  number. That replaces the exactMatch/closeMatch count configuration,
  which counted occurrences rather than distinct numbers and refused a
  number spelled twice, and the 13 character truncation that stripped
  the check character on the GND side alone
* give every concept and place source one `association` property
  returning the identifiers and the match level together, and read one
  side's candidates in a single scan
* keep GND's `075 $b saf` in `bnf_type`: under `$2 gndspec` it marks a
  record usable as a form designation, the counterpart to IdRef's
  `genre/forme Rameau`
* au... (continued)

215 of 228 new or added lines in 17 files covered. (94.3%)

5594 of 6387 relevant lines covered (87.58%)

0.88 hits per line