Thomas Zilio

I'm not sure to fully understand your proposal and its implications but here is what I understand/expect. I don't know what the original intent was when designing the "context" attribute...

This is correct and looks cleaner than the way I'm doing it. I think (not 100% sure) this would allow me to not use contexts this way anymore and not...

You're right when you say > Note that it's very easy to later convert from float32 to float64, e.g., by writing ds.astype(np.float64). You'll have a float64 in the end but...

To explain the full context and why it became some kind of a problem to us : We're experimenting with the parquet format (via pyarrow) and we first did something...

So, a more complete example showing this problem. NetCDF file used in the example : [test.nc.zip](https://github.com/pydata/xarray/files/2270125/test.nc.zip) ````python from netCDF4 import Dataset import xarray as xr import numpy as np import...

> @Thomas-Z is it possible for you to remove the defaults channel? I'm usually using the following options: `--override-channels -c $OCTANT_CHANNEL -c conda-forge ` so the defaults channel is not...

I did not have time to create a reproducible example but I have an additional element regarding the KeyError problem. When using mamba build I have the following debug message...

> The first issue is with Timedelta decoding. If you remove the units attribute the pipeline works. Not sure if it helps but keeping the unit and removing the fill_value...

My problem is more about the fact that we can no longer read these type of variables without setting `mask_and_scale` to False (and I broke some of my colleagues CryoSat...

Same here. I’ve been having this problem for months now. I tried with logs in debug verbosity but nothing found in these logs seemed useful to me. Sometimes it will...