rero / sonar / 32370721682

95%

master: 95%

push

github

fix(documents): correct context, identifiers and cite keys in RIS/BibTeX export

Follow-up to #1130, from an audit of the export against the source JSON
of a full test corpus.

* Place and publisher now follow what BibTeX, RIS and CSL expect for a
  hosted item: the host publication statement, split from its ISBD form
  ("Paris : PUF, 2016"), takes that slot when present, otherwise the
  record's own provisionActivity keeps it. No exporter read the statement
  until now, leaving @incollection without its required publisher.

* A thesis fills the same slot with its granting institution (BibTeX
  school, RIS PB) and states its degree in BibTeX type. Only a doctoral
  thesis maps to phdthesis; every other level rides on mastersthesis,
  whose printed label type replaces. Thesis detection keys off
  documentType, not off the entry type, which cannot tell the levels
  apart.

* extract_identifiers() dropped every non-DOI/ISBN/ISSN identifier (ark,
  URN, Local, ReportNumber...), about half of the corpus. They are now
  exported as RIS N1 notes and as BibTeX fields named after the type,
  numbered when a type repeats.

* BibTeX cite keys ("{Surname}{Year}") now include the record's pid,
  which makes them unique by construction. The letter-suffix
  disambiguation they relied on could only see the keys of a single
  serialize_search() call: an export assembled from independent calls --
  a paginated search, or repeated single-record exports, which had no
  disambiguation at all -- kept producing collisions, and BibTeX
  consumers either reject or silently merge colliding entries. The
  suffix machinery is dropped with the need for it.

Co-Authored-By: Miguel Moreira <miguel.moreira@rero.ch>

59 of 60 new or added lines in 3 files covered. (98.33%)

7667 of 8084 relevant lines covered (94.84%)

0.95 hits per line