Line plots of 3D data

[ZedThree]

I'm trying to recreate an existing plot of a 3D array:

for i in range(array.shape[1]):
    for j in range(array.shape[2]):
        plt.plot(time, array[:, i, j])

but the naive translation array.plot.line(x="t") gives me

ValueError: Line plots are for 1- or 2-dimensional DataArrays. Passed DataArray has 3 dimensions

Sort of dropping that restriction from xarray itself, what's the most idiomatic way to do this with xarray? Is there a nice way to iterate over a particular coordinate? Obviously the original method still works, but it doesn't feel very idiomatic!

[kmuehlbauer]

Maybe just select the wanted entries, something along the line (dimension names have to be changed according to your array):

array.isel(dim1=i, dim2=j).plot()

[ZedThree]

Possibly a stupid question, but is there any difference between array[:, i, j] and array.isel(dim1=i, dim2=j)?

[kmuehlbauer]

They are interchangeable AFAIK, but maybe I misunderstood your question.

[TomNicholas]

The only differences are that:

• The isel call will work regardless of the dimensionality of your array, whereas the array indexing expects that your array is 3D,
• The isel call requires that your array have dimensions named 'dim1' and 'dim2', whereas the array indexing doesn't care what the names of your dimensions are, because its just looking up nameless numpy axes,
• The isel call is insensitive to the transposition of your array (and so array.isel(dim1=i, dim2=j) is equivalent too).

So as long as array is 3D, has a dim1 and a dim2, which are the 2nd and 3rd dimensions, then the calls should be equivalent.

[TomNicholas]

You could maybe also try using stack to stack the two dimensions you want to loop over into one dimension, before plotting?