Set one-dimensional data variable as dimension coordinate?

[nedclimaterisk]

Code Sample

I have this dataset, and I'd like to make it indexable by time:

<xarray.Dataset>
Dimensions:                (station_observations: 46862)
Dimensions without coordinates: station_observations
Data variables:
    time                   (station_observations) datetime64[ns] ...
    SNOW_ON_THE_GROUND     (station_observations) float64 ...
    ONE_DAY_SNOW           (station_observations) float64 ...
    ONE_DAY_RAIN           (station_observations) float64 ...
    ONE_DAY_PRECIPITATION  (station_observations) float64 ...
    MIN_TEMP               (station_observations) float64 ...
    MAX_TEMP               (station_observations) float64 ...
Attributes:
    elevation:     15.0

Problem description

I expected to be able to use ds.set_coords to make the time variable an indexable coordinate. The variable IS converted to a coordinate, but it is not a dimension coordinate, so I can't index with it. I can use assign_coords(station_observations=ds.time) to make station_observations indexable by time, but then the name in semantically wrong, and the time variable still exists, which makes the code harder to maintain.

Expected Output

ds.set_coords('time', inplace=True)
<xarray.Dataset>
Dimensions:                (station_observations: 46862)
Coordinates:
    time                   (station_observations) datetime64[ns] ...
Dimensions without coordinates: station_observations
Data variables:
    SNOW_ON_THE_GROUND     (station_observations) float64 ...
    ONE_DAY_SNOW           (station_observations) float64 ...
    ONE_DAY_RAIN           (station_observations) float64 ...
    ONE_DAY_PRECIPITATION  (station_observations) float64 ...
    MIN_TEMP               (station_observations) float64 ...
    MAX_TEMP               (station_observations) float64 ...
Attributes:
    elevation:     15.0

In [95]: ds.sel(time='1896')
ValueError: dimensions or multi-index levels ['time'] do not exist

with assign_coords:

In [97]: ds=ds.assign_coords(station_observations=ds.time)

In [98]: ds.sel(station_observations='1896')
Out[98]: 
<xarray.Dataset>
Dimensions:                (station_observations: 366)
Coordinates:
  * station_observations   (station_observations) datetime64[ns] 1896-01-01 ...
Data variables:
    time                   (station_observations) datetime64[ns] ...
    SNOW_ON_THE_GROUND     (station_observations) float64 ...
    ONE_DAY_SNOW           (station_observations) float64 ...
    ONE_DAY_RAIN           (station_observations) float64 ...
    ONE_DAY_PRECIPITATION  (station_observations) float64 ...
    MIN_TEMP               (station_observations) float64 ...
    MAX_TEMP               (station_observations) float64 ...
Attributes:
    elevation:     15.0

works correctly, but looks ugly. It would be nice if the time variable could be assigned as a dimension directly. I can drop the time variable and rename the station_observations, but it's a little annoying to do so.

Output of xr.show_versions()

INSTALLED VERSIONS

commit: None
python: 3.6.6.final.0
python-bits: 64
OS: Linux
OS-release: 4.16.0-041600-generic
machine: x86_64
processor: x86_64
byteorder: little
LC_ALL: None
LANG: en_AU.UTF-8
LOCALE: en_AU.UTF-8

xarray: 0.10.2
pandas: 0.22.0
numpy: 1.13.3
scipy: 0.19.1
netCDF4: 1.3.1
h5netcdf: None
h5py: None
Nio: None
zarr: None
bottleneck: 1.2.0
cyordereddict: None
dask: 0.16.0
distributed: None
matplotlib: 2.1.1
cartopy: None
seaborn: None
setuptools: 39.0.1
pip: 9.0.1
conda: None
pytest: None
IPython: 5.5.0
sphinx: None

[fujiisoup]

Hi @nedclimaterisk.
Thanks for the raising an issue.

In that case, you can use swap_dims,

In [1]: import xarray as xr
   ...: ds = xr.Dataset({'x': ('i', [0, 1, 2]), 'y': ('i', ['a', 'b', 'c'])})
   ...: ds
   ...: 
   ...: 
Out[1]: 
<xarray.Dataset>
Dimensions:  (i: 3)
Dimensions without coordinates: i
Data variables:
    x        (i) int64 0 1 2
    y        (i) <U1 'a' 'b' 'c'

In [2]: ds.swap_dims({'i': 'x'})
Out[2]: 
<xarray.Dataset>
Dimensions:  (x: 3)
Coordinates:
  * x        (x) int64 0 1 2
Data variables:
    y        (x) <U1 'a' 'b' 'c'

[fujiisoup]

I'm closing this issue, but if you have further issue, do not hesitate to reopen this.

[nedclimaterisk]

Awesome, thank you @fujiisoup.

It might be worth putting a "see also" note in the assign_coords and set_coords documentation for this. I tried searching quite a bit, but did not find this.

[fujiisoup]

It might be worth putting a "see also" note in the assign_coords and set_coords documentation for this. I tried searching quite a bit, but did not find this.

Thanks for the suggestion. It sounds a good idea. Do you mind to send a PR for this?

[M-Harrington]

@fujiisoup This method works when trying to add a single coordinate, but what about when you're trying to add multiple coordinates? Example:

import xarray as xr
ds = xr.Dataset({'x': ('i', [0, 1, 2]), 'y': ('i', ['a', 'b', 'c']),'z': ('i', ['a', 'b', 'c'])})
ds= ds.set_coords(['y','z'])
ds.swap_dims({'i':'y','i':'z'}) #doesn't work

ds = xr.Dataset({'x': ('i', [0, 1, 2]), 'y': ('i', ['a', 'b', 'c']),'z': ('i', ['a', 'b', 'c'])})
ds= ds.set_coords('y')
ds=ds.swap_dims({'i':'y'}) 
ds.set_coords('z') #doesn't work either

Is there any reason that this is the default behavior? This is a bit frustrating to work with after creating an xarray dataset from pandas.

[dcherian]

ds.rename_dims({"i": "y"})

Is this what you want?

[M-Harrington]

@dcherian not quite because I want z and y to both have the star next to them (or bolded in your screenshot) so that they're proper coordinates. I likewise thought that the answer would be as simple as:

data=pd.DataFrame({'x': [0, 1, 2], 'y': ['a', 'b', 'c'],'z': ['a', 'b', 'c']})
xr.DataArray(data.x, coords={'y': data.y, 'x':data.z})

But again, I haven't gotten anywhere this way either and getting the error ValueError: coordinate y has dimensions ('y',), but these are not a subset of the DataArray dimensions ('dim_0',)

[dcherian]

So that's not possible. You can't have both y and z be 1D coordinate variables for x since x is 1D.

What are you ultimately trying to do after the conversion to DataArray?

[M-Harrington]

This also doesn't work feeding y and z as data.y.values and data.z.values which are 1d arrays.

Ultimately merge to another dataset with the same coordinates. Seems like there's something obvious I'm missing here but I haven't been able to figure out what it is.

Ah I see in this example I need a dataset that's 3x3, let me fix the example and see if it's still relevant to my issue

[keewis]

to get your example to work, use this:

data=pd.DataFrame({'x': [0, 1, 2], 'y': ['a', 'b', 'c'],'z': ['a', 'b', 'c']})
xr.DataArray(data.x, coords={'y': data.y, 'x':data.z}, dims="y")

to get both as dimensions, use

df = pd.DataFrame({'x': [0, 1, 2], 'y': ['a', 'b', 'c'],'z': ['a', 'b', 'c']})
ds = df.set_index(["y", "z"]).to_xarray()

[M-Harrington]

This worked perfectly, thanks so much!

[keewis]

just note that in the end the result is still 2D with the missing values filled with nan

[M-Harrington]

Right that's actually desired behavior to begin with so this works out