Implement new "most common" regridder. #46
Conversation
|
@dcherian I don't seem to be a maintainer on this repo. Perhaps due to the repository not being linked to the team? See https://github.com/orgs/xarray-contrib/teams/xarray-regrid-devs/repositories |
|
I changed it here: https://github.com/xarray-contrib/xarray-regrid/settings/access Should work now? |
|
@dcherian the routines seem very fast and efficient now. Didn't see too many memory issues, and especially the statistical reduction method is very fast. Thanks for your help! |
|
PR is basically ready, except for the padding issue with the statistic based reduction methods. The results of the regridding and edge cases are also not fully tested yet, but that can probably be added later. |
tests/test_format.py
Outdated
| assert formatted.longitude[0] == -1.5 | ||
| assert formatted.longitude[-1] == 361.5 | ||
| # And preserve integer dtypes | ||
| assert formatted.data.dtype == np.int64 |
There was a problem hiding this comment.
I couldn't find any unwanted dtype conversion with longitude padding, because there we're doing a .pad using the existing data. The latitude padding was converting to float, because I set that up to concat the .mean() of the last latitude rows as an extra point. Added a test to check this.
I haven't tested the behavior at the poles with the statistical aggregations to know whether we actually do want some padding there in certain cases. If we do, we would need some new logic to determine what to pad with.
There was a problem hiding this comment.
Yeah not sure what went wrong. I could not run the demo notebook with the landcover data, as that was unint8. Everything now seems to run OK 😕 .
I haven't tested the behavior at the poles with the statistical aggregations to know whether we actually do want some padding there in certain cases. If we do, we would need some new logic to determine what to pad with.
I think we would not want to do any padding there. The zonal statistics methods really are all about groupby/binning reductions.
This implements the method discussed in xarray-contrib/flox#263
Also added
regrid.statto implement other aggregations from flox.Todo:
regrid.stat