washopenresearch 0.4.0

Two new datasets, and the author email addresses removed from all of them.

New datasets

ws (Water Supply, 4,884 articles, 2001 to 2026) and jwh (Journal of Water and Health, 2,013 articles, 2003 to 2026) were scraped alongside washdev but had never been exported, so they were reachable only from a local clone of the source repository. With them in place the corpus reads 8,403 screened articles across four journals.

Journal	Articles	Reaching an open repository
Water Supply	4,884	1%
J. Water and Health	2,013	2%
J. WASH for Development	1,173	4%
PLOS Water (research articles)	333	54%

Across the three journals without an enforced data policy, one in 54 articles reaches an open repository. PLOS Water, which enforces one, sits at 54%.

Data availability statements were mapped to the shared das_type levels for 1,596 of 1,623 statements in ws and 717 of 733 in jwh. The unmapped tail keeps its full statement text and is listed in per-journal review files rather than being silently reclassified.

Breaking changes

Author email addresses are removed from every dataset. first_author_email and correspondence_author_email no longer exist in washdev, ploswater, uncnewsletter, ws or jwh, and the 52 addresses authors had written into the statements themselves are masked in place, keeping the domain so the statement still reads. Code reading either column needs updating.

These are personal data, and they earn nothing analytically: the research questions use author country, das_type, keyword frequency and supplementary counts. A CC BY table of corresponding author addresses is a ready-made mailing list, which is the concrete harm. The addresses remain in data-raw/ for provenance.

This does not undo the past. The addresses were published up to and including v0.3.0 and remain in those deposits.

inst/extdata is no longer built into the package. The flat files are still generated and still live in the repository, but shipping a CSV and an XLSX for six datasets would push the built tarball past the 5 MB CRAN guidance. system.file("extdata", ...) no longer resolves; read the datasets directly, for example data(ws). inst/CITATION is unaffected, so citation("washopenresearch") is unchanged.

Zenodo pre-signed access tokens are stripped from repository URLs. One statement carried a link granting access to an otherwise restricted record; the token goes and the record URL stays.

Also in this release

• washdev, ws and jwh come off one scraper with one schema, so the shared cleaning moved into process_iwa_journal() and washdev was retrofitted onto it.
• The package's first tests: row counts, absence of email addresses in structured and free-text fields, absence of access tokens, schema agreement across the IWA datasets, and the no-list-columns guarantee.
• Yash Dubey added as an author, for the Selenium scraper and washdev data update contributed in December 2024.
• The README now documents all six datasets, and its method section describes the R scrapers rather than the Python tooling removed several releases ago.
• AQUA was scraped in the same run and is deliberately not exported: 539 of its 1,819 rows carry has_das TRUE while the statement text is empty, so its central variable would ship blank. It needs a re-scrape first.

Full changelog: v0.3.0...v0.4.0