issue_comments

• this issue was submitted because of https://github.com/pydata/xarray/pull/7085#discussion_r981443796
• I also tend to agree that this behavior is fine for the non-specific netcdf attributes.
• For specific attributes as unit and _FillValue, is it also fine ? I would expect that Dataset.to_netcdf check the type (string for unit and the type of the variable for _FillValue and raise at least a warning. It is currently possible to save a number for the unit... Is the resulting netcdf CF-compliant ?
• if any kind of types for attributes is allowed, this requires to chase bugs of the kind solved in #7085.

yes, no problem, take your time. This is still a wanted feature in any case, I've recently needed it again. I've found a solution by converting the real index to integer after adequate scaling, but an integrated solution would be much better.

Interesting discussion #2217. As far as I understand, it solves a wider problem than using a tolerance on and a member by member comparison... but more complex to implement.

Here is the simple solution that works for me (combine.py:70):

        if not (all(index.equals(indexes[0]) for index in indexes[1:])
                or (tolerance > 0 and \
                    all(index.is_numeric() for index in indexes) and \
                    all(np.allclose(index, indexes[0], atol=tolerance, rtol=0) for index in indexes[1:]))):

I've not tested in depth, I can make a PR, tests, doc if you agree with this solution.

Regarding the name of the arg. "tolerance" is nice, but allclose has atol and rtol. My solution above only set atol, but both may be useful. Should we use: 2args: atol, rtol 2 args: atolerance, rtolerance 1 args: tolerance could be a number (->atol) or a tuple interpreted as atol, rtol = tolerance or a dict .... ?

The proposed PR completely rewrite how the Cartesian product is computed, MultiIndex.from_product is unable to deal with MultiIndex which was written for any iterables.

thanks for the very clear response. The behaviro make sense.

In fact, I should have explained what I'm trying to achieve, as this is kind of "take". I've a dict like this: python {'label1' : dict(coord1=1, coord2=4), 'label2' : dict(coord1=5, coord2=6), 'label3' : dict(coord1=4, coord2=2), } and I want to build an xarray (and then a dataframe) with coord1 and coord2 replaced by a new dims with values 'label1', 'label2', 'label3'.

I've done that by iterating over the dict, selecting with sel using the dict values, convert to dataframe and then concat the dataframes. pd.concat([x.sel(**d[k]).to_dataframe() or k in d]

A better option would be to do this "sel" or "take" with xarray only. Do you have an idea how to do it with existing xarray methods?