Tracking: openwashdata cleanup day #20

The first full-organisation harvest read 96 repositories and found 64 data packages. 32 carry a Zenodo DOI. 32 do not. This issue tracks closing that gap and the smaller defects the harvest surfaced, as one focused cleanup rather than a trickle.

Why now

The website catalogue used to come from a hand-maintained Google Sheet, which called 50 packages published. The sheet meant "we consider this released"; the harvest means "it has a citable deposit". Those are different claims, and the sheet was the only thing hiding the difference.

The catalogue now shows every package with a DOI column, so a missing deposit is visible without the package disappearing from the site (openwashdata/website#153). That makes this list actionable instead of invisible.

The 32 without a DOI

Each has an issue on its own repository titled "Deposit on Zenodo so the package is citable", listing the deposit steps and whatever else that package is missing.

19 the site currently calls published. These are the visible ones, worth doing first:

basisghana, boreholelabdata, chckapmalawi, cleanteamghana, dowalatrinesurvey, dowaodfsurvey, drillingdata, dryingfaecal, floodresponsemw, grdwtrsmpkwale, handpumpstatusdata, mwefloodresponse, mwgroundwaterdata, nchalohhwatersurvey, portawaterperu, thyolocbcc, universityrankingafrica, waterconsumersurvey, wsabrazil

13 never claimed published, so no public statement changes when they get a DOI:

climatevulnerabilitykhm, faecalsludgekampala, glaas, gwinfraassess, jmpdata, rainwaterhsurvey, washdataportalsl, wasteburningblantyre, wastecollectionblantyre, wastedata, waterpointstatus, whatawaste, whdxdata

Other findings

• waterpointfunctionality is private. It is a real data package with a data/*.rda, the site lists it as published, and the harvest cannot see it. Decide whether it becomes public or leaves the catalogue.
• No package carries X-schema.org-spatialCoverage. The sheet held 14 hand-curated locations for 56 packages; that information now exists nowhere machine readable, and the landing page lost its "countries and regions" figure because of it. Recovering it means writing the location into each DESCRIPTION.
• No package carries the 1.1.0 JSON-LD (pkgdown/templates/in-header.html from washr::update_metadata()), so none of them is described to search engines.
• 24 of the 32 have no R CMD check workflow. washr::setup_ci() writes it, but only from washr 1.1.1: 1.1.0 wrote a file whose expressions were corrupted, setup_ci() writes a broken R-CMD-check workflow: whisker eats the ${{ }} expressions washr#130.
• Version drift between DESCRIPTION and CITATION.cff in grdwtrsmpkwale and portawaterperu.
• hydrogeologicalghana has a DESCRIPTION URL that does not name its own repository.
• waterpointstatus writes its version as v1.0.0; the harvest normalises it, but the file should be fixed.

Suggested order

1. The 19 the site calls published, since the claim is public.
2. waterpointfunctionality, which is a decision rather than a task.
3. The remaining 13 deposits.
4. Spatial coverage and JSON-LD as a sweep, once washr::update_metadata() is run over the fleet.
5. The four one-off defects above.

Steps 4 and 5 get cheaper after washr #81 lands the Config/washr fields, since the harvest can then report which packages are behind.