Took a stab at implementing these functions.

Just to clarify, load_dataset would be equivalent to:

def load_dataset(args, kwargs): with xarray.open_dataset(args, **kwargs) as ds: return ds.load()

Yes, that is actually in my mind.

I added a tag good first issue for this. I love to implement this by myself but I don't think I have enough free time near future...

Just to clarify, load_dataset would be equivalent to: python def load_dataset(*args, **kwargs): with xarray.open_dataset(*args, **kwargs) as ds: return ds.load() This also seems pretty reasonable to me. I’ve written a version of this utility function a handful of times, so I at least would find it useful.

Would we also want load_dataarray?

I used to use the autoclose=True argument to open_dataset() all the time to avoid this issue. Then it was deprecated with the LRU introduction, and I now do a two-line ds = open_dataset().load() and then ds.close(). Just adding my use-case and what I found to be a regression in terms of usability. Something like load_dataset() would be helpful but I wonder if adding an argument to the existing open_dataset() would be better.

I didn't notice that with statement can be used for open_dataset. Thanks for the comments.

I personally want to a simpler function for the daily (not so big data) analysis without caring open / close stuff. (and probably also for new users) Is it too much if we add load_dataset function that works similar to np.load_txt in terms of the file handling?

I think this is more of a limitation of netCDF-C / HDF5 than xarray. For example, this example works if you use SciPy's netCDF reader/writer: ``` import xarray as xr ds = xr.Dataset({'var': ('x', [0, 1, 2])}) ds.to_netcdf('test.nc', engine='scipy')

ds_read = xr.open_dataset('test.nc', engine='scipy') ds.to_netcdf('test.nc', engine='scipy') ```

OK. But what about the usual scientist workflow where you work in multiple cells

``` ds.load()

do computation

```

next cell: do more computation ...

ds.close() # is this right? ds.to_netcdf('test.nc')

I wonder if we should add autoclose as a kwarg for to_netcdf. If autoclose=True, remove that file from the LRU cache if present, and then write to file.

This pattern should work: python with xr.open_dataset('test.nc') as ds: ds.load() ds.to_netcdf('test.nc')

what is the recommended code pattern for reading a file; adding a few variables; and then writing back to the same file?