Closed ocefpaf closed 1 year ago
The failures here seems to be unrelated to this PR and hopefully will be fixed in #70
I'm on board but let's hold this until the other PRs are merged, I will add some corner case tests and try to break things.
PS:
h5netcdf
is used by many Earth Sciences communities, includingxarray
, so it will be properly maintained and the switch should not bring any headaches.
h5netcdf
maintainer here. Please report any issues you might discover on the way. Trying to keep any headaches as short as possible :wink:
@kmuehlbauer I've postponed a lot the move to h5netcdf, mostly b/c I was lazy. Lately I'm using it pretty much everywhere without any issues. The main reason that forced me to move is the ability to use it with pyscript, hdf5 and h5py have larger communities than libnetcdf and netcdf4-python, so they are readily available everywhere., That makes h5netcdf the easiest way to read/write netcdf files with pyscript.
Merged manually after a rebase, this had some leftover changes from the xarray
change branch: https://github.com/ioos/ioos_qc/commit/06f401d3cd44886fd639bf6f1bf3bea623c43a00
This is the first step to reduce the dependency on complex binaries. The
h5netcdf
library usesh5py
to mimicnetCDF4
's API and, b/c wheels for it are more readily available we can increase the options where we can installioos_qc
. See https://github.com/h5netcdf/h5netcdf for more information on the advantages not depending on netcdf-c.PS:
h5netcdf
is used by many Earth Sciences communities, includingxarray
, so it will be properly maintained and the switch should not bring any headaches.