issues

What is your issue?

I'm trying to produce a facet plot which contains maps with different overlaid layers (e.g. a pcolormesh and streamplot). At the moment I'm creating the plot and then iterating over the axes to add the plots manuallay

```python p = dss['LH'].plot.pcolormesh( x='lon', y='lat', col="exp", )

for i, ax in enumerate(p.axes.flat): ax.coastlines() ax.streamplot( dss.isel(exp=i).lon.values, dss.isel(exp=i).lat.values, dss.isel(exp=i)['u_10m_gr'].values, dss.isel(exp=i)['v_10m_gr'].values, ) ```

This is far from optimal and doesn't really look clean to me. Also, I'm not entirely sure the order of p.axes.flat correspond to the one of the exp dimension I'm using to facet.

All examples in the doc (https://docs.xarray.dev/en/stable/user-guide/plotting.html) refer to the plot method of DataArray, so it seems that, once created the p object, no other variable from the dataset can be accessed.

However, on the doc it is mentioned

TODO: add an example of using the map method to plot dataset variables (e.g., with plt.quiver).

It is not clear to me whether the xarray.plot.FacetGrid.map method can indeed be used to plot another dataset variable or not. If that's not the case, is there any way to achieve what I'm doing without manually looping through the axes?

What happened?

I have a script structured like this

```python def main(): global ds ds = xr.open_dataset(file) for point in points: compute(point)

def compute(point): ds_point = ds.sel(lat=point['latitude'], lon=point['longitude'], method='nearest') print(ds_point.var.mean()) # do something with ds_point and other data...

if name == "main": main() ```

This works as expected. However, if I try to parallelize compute by calling it with

python process_map(compute, points, max_workers=5, chunksize=1)

The results of the print are completely different from the serial example and they change every time that I run the script. it seems that the sel is giving back a different part of the dataset when there are multiple processes running in parallel.

If I move the open_dataset statement inside compute then everything works also in the parallel case in the same way as in the serial one. Also, if I load the dataset at the beginning, i.e. ds = xr.open_dataset(file).load(), I also have reproducible results.

Is this supposed to happen? I really don't understand how.

What did you expect to happen?

The behaviour of sel should be the same in parallel or serial execution.

Minimal Complete Verifiable Example

No response

MVCE confirmation

• [X] Minimal example — the example is as focused as reasonably possible to demonstrate the underlying issue in xarray.
• [ ] Complete example — the example is self-contained, including all data and the text of any traceback.
• [ ] Verifiable example — the example copy & pastes into an IPython prompt or Binder notebook, returning the result.
• [ ] New issue — a search of GitHub Issues suggests this is not a duplicate.

Relevant log output

No response

Anything else we need to know?

No response

Environment

What happened: When trying to write a dataset to netcdf file using the netcdf4 engine time is not saved correctly.

What you expected to happen: Time to be saved correctly as in the original dataset.

Minimal Complete Verifiable Example:

python ds.to_netcdf(filename, encoding={product_type: {'zlib': True, 'complevel': 9}}, engine='netcdf4')

is giving me the warning SerializationWarning: saving variable time with floating point data as an integer dtype without any _FillValue to use for NaNs

xarray Dataset

saved on disk (notice time values)

I cannot see anything special in the time array...is there a limitation because of the compression?

Environment: