@dcherian @keewis I don't have permission to request reviews, but this has been open for almost a week so I wanted to nudge you.

My doubts on this are because self._replace are used elsewhere in the code.

Changing from _replace to copy causes the tests on coarsen to fail, I am looking more into this now.

It seems that a condition of copy is that the data shapes match (variable.py line 947).

If I run ds = xr.tutorial.load_dataset("air_temperature") ds.air.coarsen(lat=5) there is no problem, but once I add a .mean() to the end ds = xr.tutorial.load_dataset("air_temperature") ds.air.coarsen(lat=5).mean()

The error is ValueError: Data shape (2920, 5, 53) must match shape of object (2920, 25, 53)

This PR passes **kwargs into _coarsen_reshape and changes _replace to copy in the fx return of coarsen.

@dcherian Could this be because the return of variable.py/coarsen or is it likely happening earlier in the fx?

return self._replace(data=func(reshaped, axis=axes, **kwargs))

Added those tests for for the raised Value Error.

This issue can be closed, and is addressed in pull request #3045

Checks could be more complicated, but I think this is ready to merge!

I would definitely recommend reading through our contributor's guide on how to setup a local environment for testing your code: http://xarray.pydata.org/en/stable/contributing.html

It is much easier to fix issues when you run tests locally rather than waiting on continuous integration tests (Travis-CI) to tell you that things are broken.

Thank you was just needing to look into this.

FYI tests are failing with E NameError: name '_rename_var_dims_helper' is not defined: https://travis-ci.org/pydata/xarray/jobs/550940568

Huh not sure why that is. I can remove that function and see if it works then.

I uninstalled the bot that automatically closed the pull request #3042

@jukent check out the existing rename tests here (they're a bit over-complicated at the beginning): https://github.com/pydata/xarray/blob/master/xarray/tests/test_dataset.py#L2046-L2130

You could do something as simple as this: https://github.com/pydata/xarray/blob/master/xarray/tests/test_dataset.py#L2114-L2119 - i.e. start with a dataset, call a rename method, and compare the dataset with a manually constructed version of the expected result.

Ping back with any issues

Thanks @max-sixty. I added some very simple tests. I might copy more of the test_rename functions (if I tried to rename a variable to the same name for example)

3026

There is some unexpected behavior. The new rename_dims function does not change the dimensions of the variables. Working on this now.

Similarly rename_vars will rename the variable dimensions and variables, but not the dataset dimensions. I thought I was dealing with this with the _rename_indexes. But will have to keep working.

This works now! But to do so I created three different renaming variable helper functions (var name only, var and dims, var dims only). I will clean this up since the majority of these three functions is the same.

And tbc, we should add tests here. That will help clarify the issues above re exactly which items on the dataset each method should rename vs leave.

Yes I need to add tests, but I don't have much experience with this yet. I will ask my team here to help me.

3026

There is some unexpected behavior. The new rename_dims function does not change the dimensions of the variables. Working on this now.

Similarly rename_vars will rename the variable dimensions and variables, but not the dataset dimensions. I thought I was dealing with this with the _rename_indexes. But will have to keep working.

pydata#3026

There is some unexpected behavior. The new rename_dims function does not change the dimensions of the variables. Working on this now.

Created pull request #3042

I will submit a pull request to add separate rename dims and rename coords functions in a day or two.

Should we close this issue?

import xarray as xr ds = xr.tutorial.open_dataset('air_temperature') ds.coords['y'] = ('lat',ds.lat) ds.coords['x'] = ('lon',ds.lon) ds_swappeddims = ds.swap_dims({'lat':'y','lon':'x'}) ds_swappeddims.drop(['x','y'])

Did you do something like this?/Is this what you are after but would like a cleaner solution?

I am looking into this, but first I am making sure I understand the relationship between coordinates and dimensions. Could you give me an example of when you would use this functionality?