eval mathematical expression using labels from an index

[robintw]

I have a DataArray that has three dimensions: x, y and bands. The labels in the bands array are names like red, blue, nir etc.

I would like to evaluate a mathematical expression that uses those labels by doing something like:

da.eval("red-nir")

or

da.eval("(red+green)/(2*(blue+nir)")

These would return a DataArray with the resulting values.

The idea is that the user would be passing these expressions to my code, so I can't just hard code them.

Obviously the da.eval method doesn't exist at the moment - but that's representing the ideal interface.

Is there any nice way to do this at the moment?

I could probably put together some rather horrible code that iterated through the labels on the band axis, extracted each band into its own variable using python's eval function with a string like bandname = da.sel(band=bandname), and then use python's eval function to do the final calculation - but I'm sure there must be a better way.

Any thoughts?

[keewis]

you could use to_dataset for this:

da = xr.DataArray(
    np.ones((10, 10, 4)),
    coords={"bands": ["red", "green", "blue", "nir"]},
    dims=("x", "y", "bands"),
)

ds = da.to_dataset(dim="bands")
(ds.red + ds.green) / (2 * (ds.blue + ds.nir))

this is a bit longer than the proposed .eval, but it is much more readable than using sel