Andrew Robbins
Andrew Robbins
that's all of them- the rest were either duplicates or garbage
Looks like MCDS.open doesn't like the outputs of these (fails on obs_dim lookup with a KeyError) so...bear with me i guess.
Siloing out the zarr v3 stuff into a PR for a later date, was more broken than I thought.
Okay, I can confirm it actually works now.
Well, it works in the single chunk, small dataset case. I don't know whether increasing the chunk size broke it or if large datasets don't work. not great either way-testing...
Judging by my script actually taking the time to process my data this go-around, i think its reasonable to say that this PR works on chunk-size 1 and no other....
If I'm remembering correctly, the difference in the use case is that our region definition is ~400k tiny sites, which causes the output of write_single_zarr to be much larger and...
I think-although I could be incorrect-that the data we're working with is a combination of SBP and very small regions of several BP, so if I can get MCDS to...
A new update on this-it appears that the no-output issue I'm having is not due to chunk size *or* the dictionary of regiongroups. It seems that above some number of...
....oh. malformed allc_table. bear with