Skip to contents

Changelog

Source: NEWS.md

openalexConvert 0.0.4

Require openalexPro (>= 0.11.0)

.normalize_doi() delegates to openalexPro::extract_doi(normalize = TRUE) (R/corpus_to_csljson.R:430), and before openalexPro 0.11.0 that function silently truncated DOIs containing <, >, [ or ] – returning a shorter string that still looked like a valid DOI, with no warning and no NA. Roughly 0.38% of real DOIs are affected (SICI-style ones, common in older journal content).

The dependency carried no version floor, so installing against an older openalexPro produced quietly corrupted DOIs in the CSL-JSON output.

Fixtures regenerated: they encoded the truncation bug

Once the floor was in place the suite failed, because every comparison fixture had been recorded while extract_doi() was still truncating. The fixtures expected the wrong answer:

generated: "10.1659/0276-4741(2005)025[0206:pfbcs]2.0.co;2"   <- correct
fixture:   "10.1659/0276-4741(2005)025"                       <- truncated

The BibTeX and BibLaTeX fixtures were stale in a second way: entries whose DOI had been lost now carry a doi = {...} field they previously lacked.

Every difference was verified to be DOI-related before regenerating – 6 differing CSL fields and 187 differing .bib lines, all of them a DOI value or a doi = line (plus the trailing comma that appears on the line above a newly added field). Nothing else moved.

data-raw/regenerate_fixtures.R now does this reproducibly, and refuses to run outside a UTF-8 locale: in a C locale R renders non-ASCII author names as literal <U+00F3> escapes and pandoc writes Bay\textless U+00F3\textgreater n into the .bib files, which is exactly how a broken fixture gets committed and then becomes the expectation.

The suite now passes: 119 pass / 0 fail, from 105 pass / 14 fail.

openalexConvert 0.0.3

Bug fixes

  • corpus_export_via_pandoc(): now produces a single output file (e.g. corpus.bib) as documented. Previously it passed the chunked CSL JSON directory to csljson_convert_pandoc(), which created a directory of per-chunk files at the output path, and it returned the wrong path when an extension had to be appended.
  • csljson_convert_pandoc(): the pdf_engine parameter is now respected. Previously --pdf-engine=xelatex was unconditionally appended, overriding any user-specified engine (e.g. "lualatex" or "pdflatex").
  • DOI extraction now delegates to openalexPro::extract_doi(), which fixes a regex that missed lowercase characters in DOI suffixes (e.g. 10.1234/abc-def).

Improvements

  • openalexPro declared as a formal dependency in DESCRIPTION (Imports).
  • rmarkdown moved from Suggests to Imports (it is used in exported functions).
  • Additional_repositories added to DESCRIPTION so CI and users can resolve openalexPro from r-universe automatically.
  • CI aligned with the rest of the openalexPro ecosystem: standard R-CMD-check.yaml (macOS, Windows, Ubuntu × devel/release/oldrel-1) and test-coverage.yaml added.
  • Test coverage raised to ~86%, with new tests for corpus_export_via_pandoc(), Pandoc rendering paths, and csljson_to_zotero_upload().
  • Leftover _problems/ test artefacts from a previous failed run removed.