issue_comments

If we think can improve an error message by adding additional context, the right solution is to use raise Exception(...) from original_error: https://stackoverflow.com/a/16414892/809705

On the other hand, if xarray doesn't have anything more to add on top of the original error message, it is best not to add any wrapper at all. Users will just see the original error from dask.

I also thought that, after the dask error message it's pretty easy to then look at the dataset and check what the problem dimension is.

In general though, is that the type of layout you'd suggest for catching and re-raising errors? Using raise Exception() ?

Nice, that's neater! Would this work, in the maybe_chunk() call? Sorry about the basic questions!

python def maybe_chunk(name, var, chunks): chunks = selkeys(chunks, var.dims) if not chunks: chunks = None if var.ndim > 0: # when rechunking by different amounts, make sure dask names change # by provinding chunks as an input to tokenize. # subtle bugs result otherwise. see GH3350 token2 = tokenize(name, token if token else var._data, chunks) name2 = f"{name_prefix}{name}-{token2}" try: return var.chunk(chunks, name=name2, lock=lock) except NotImplementedError as err: raise Exception("Automatic chunking fails for object arrays." + "These include cftime DataArrays.") else: return var

The error message from dask is already pretty descriptive: NotImplementedError: Can not use auto rechunking with object dtype. We are unable to estimate the size in bytes of object data

I don't think we have much to add on top of that?

is_scalar(chunks) might be the appropriate condition. is_scalar is already imported from .utils in dataset.py

This seems to work fine in a lot of cases, except automatic chunking isn't implemented for object dtypes at the moment, so it fails if you pass a cftime coordinate, for example.

Can we catch this error and re-raise specifying "automatic chunking fails for object arrays. These include cftime DataArrays" or something like that?

Also, the contributing docs have been super clear so far! Thanks! :)

Cheers! Just had a look, is it as simple as just changing this line to the following, @dcherian ?

python if isinstance(chunks, Number) or chunks=='auto': chunks = dict.fromkeys(self.dims, chunks)

This seems to work fine in a lot of cases, except automatic chunking isn't implemented for object dtypes at the moment, so it fails if you pass a cftime coordinate, for example.

One option is to automatically use self=xr.decode_cf(self) if the input dataset is cftime? Or could just throw an error.

Agreed, this would be very welcome!

chunks='auto' isn't supported only because xarray support for dask predates it :)

Awesome! Please see https://xarray.pydata.org/en/stable/contributing.html for docs on contributing

Oh ok I didn't know about this, I'll take a look and read the contribution docs tomorrow ! It'll be my first PR so may need a bit of hand-holding when it comes to tests. Willing to try though!

so da.chunk({dim_name: "auto"}) works but da.chunk("auto") does not. The latter is a relatively easy fix. We just need to update the condition here: https://github.com/pydata/xarray/blob/bd84186acbd84bd386134a5b60111596cee2d8ec/xarray/core/dataset.py#L1736-L1737

A PR would be very welcome if you have the time, @AndrewWilliams3142