Stop reporting two things nobody can act on

[chross22]

Two sources of noise, neither indicating a problem.

[rast] unknown calendar (assuming standard): Gregorian

Some Copernicus files declare their calendar as Gregorian, which is not one of the spellings CF lists. terra warns and assumes the standard calendar — and that assumption is correct, since Gregorian is the standard calendar, spelled with a capital. The warning names nothing to do about it, and one per file makes 50 on a 50-day fetch, burying the warnings that matter.

Only that message is muffled, matched on its text. Every other warning from the read still propagates, because a file that is genuinely unreadable should still say so — there's a test for both halves.

marmap narrating a local file read

fetch_bathymetry() called getNOAA.bathy() whether or not the grid was already on disk. marmap handles that case, but announces it:

Querying NOAA database ...
This may take seconds to minutes, depending on grid size
File already exists ; loading 'marmap_coord_-70;42;-68;43_res_10.csv'

An existing grid is now read with marmap::read.bathy() instead. Same object, no commentary, and no NOAA round trip. Verified identical in values and extent.

That only helps if the download survives the session, so the cache moved from tempdir() to the R_user_dir location the Copernicus files already use. A study area downloaded once now stays downloaded. Passing path still overrides it.

The cache filename is marmap's convention rather than ours, so marmap_cache_file() reproduces it and a test pins it — including that a box written corner-reversed resolves to the same file. Should marmap ever change the convention, the match fails and every fetch goes back through getNOAA.bathy(): slower and chattier, not wrong.

Verification

• R CMD check — Status: OK. Full suite green. Installed into both R 4.6.1 and 4.3.2.
• Cached-read path verified end to end: zero lines printed, values and extent identical to the downloaded grid, and getNOAA.bathy() mocked to error so a NOAA request would fail the test rather than pass quietly.

Note

I could not reproduce the calendar warning on CHL_MODEL, NO3, or satellite CHL, monthly or daily — it depends on which product wrote the file. The fix is at the read, so it applies whichever product emits it.

🤖 Generated with Claude Code