Website switchover: openwashdata.org renders the harvested CSVs

[larnsce]

Goal

openwashdata.org reads the harvested catalog instead of the Google Sheet. This is the launch criterion: the catalog is not live until the website shows it.

Rewritten 2026-09-04 per dev/decisions-2026-09.md.

Why now

washr 1.1.0 (2026-09-02) deleted update_gsheet_metadata(), the only writer of the sheet, so since then no new package reaches it. washr #69 stays open for the replacement, which is owdata.

Website change

pages/gallery/data/index.qmd in the website repository reads the sheet with googlesheets4 and renders a DT table with name, location, status, description, published date and website link. Replace the sheet read with:

• https://raw.githubusercontent.com/openwashdata/owdata/main/inst/extdata/owd_packages.csv for the package table: name (pkg_name linked to url_docs), location (spatial_coverage), status (published), description, date, maintainer, DOI, keywords
• https://raw.githubusercontent.com/openwashdata/owdata/main/inst/extdata/owd_variables.csv for a variable search table on the same page or a sibling page. About 2,700 rows is fine client-side in DT.

Tasks:

• The website PR, after Weekly harvest workflow that opens a pull request #9 has produced its first merged harvest
• Freeze the Google Sheet the same day and note it in the website repo
• Close washr #69

washr documentation (follow-up)

One sentence in the washr vignette after "Cite and release" and in the publishing guide: a package with a Zenodo DOI appears in the openwashdata catalog within a week.

Acceptance criteria

• The website data page renders from the CSVs with no googlesheets4 dependency and at least the columns it shows today
• Variables are searchable on openwashdata.org
• The sheet has a recorded freeze date and washr #69 is closed

[larnsce]

Premortem findings for this issue (advisory, non-blocking)

From the 2026-07-18 premortem (report in docs/ on branch claude/implementation-premortem-jd5udn):

The premortem rated the website switchover the single highest-leverage item in the plan, and its current framing ("nothing here blocks the metapackage", no owner, no date, lives in a repo outside this issue list) is exactly how it fails to happen. If openwashdata.org keeps rendering the Google Sheet after the owdata site launches, the org runs two catalogs that disagree (56 published versus about 32 catalogued with legacy flags), maintainers keep updating the sheet because that is what the website shows, and the variable search never finds its audience because nothing links to it.

Suggestions:

• Give the switchover an owner and a target date, and treat "openwashdata.org reads or prominently links the harvested artifact" as part of the owdata launch rather than a follow-up.
• Agree a freeze date for the Google Sheet at the same time; every sheet edit after launch is a maintainer voting for the old catalog.
• Reconcile the public vocabulary before launch: a package labeled published on openwashdata.org should not appear as unreviewed candidate on owdata without a sentence somewhere explaining the relationship.

---

Generated by Claude Code