issue_comments: 115933651

OK, I understand now. One of these files looks like:

<xray.Dataset> Dimensions: (lat: 39, lon: 59, mean_height_agl: 50, time: 1) Coordinates: * time (time) datetime64[ns] 2011-05-21T13:00:00 * lon (lon) float64 -29.0 -28.0 -27.0 -26.0 -25.0 -24.0 ... * lat (lat) float64 32.0 33.0 34.0 35.0 36.0 37.0 38.0 39.0 ... * mean_height_agl (mean_height_agl) float64 28.28 97.21 191.1 310.7 ... Data variables: ash_concentration (mean_height_agl, lat, lon) float64 9.583e-16 ... ash_mass_loading (lat, lon) float64 1.091e-11 1.091e-11 1.091e-11 ... ash_drydep (lat, lon) float64 4.086e-10 4.084e-10 4.08e-10 ... ash_wetdep (lat, lon) float64 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 ... so2_concentration (mean_height_agl, lat, lon) float64 3.199e-13 ... so2_mass_loading (lat, lon) float64 2.602e-09 2.602e-09 2.602e-09 ...

The problem is that the mean_height_agl coordinate changes between each file.

Another interesting aspect of this file, which relates to how I as hoping to fix https://github.com/xray/xray/issues/438, is that it includes a time coordinate with length 1, but none of the other dataset variables use that coordinate.

This suggests to me that we need some sort of hook that can allow you to transform a single dataset before they are joined with open_mfdataset. Perhaps a preprocess argument? Then you could write, e.g.,:

``` python def fix_my_data(ds): return (ds.assign_coords( agl=('mean_height_agl', range(ds.dims['mean_height_agl']))) .swap_dims({'mean_height_agl': 'agl'}) .squeeze('time'))

ds = xray.open_mfdataset('*.nc', preprocess=fix_my_data) ```