time-varying bins

[miniufo]

Hi, it is a nice package. Just want to know if the bins parameter can be time-varying. Or more flexible, can the bins parameter be also a xarray.DataArray? I find that xhistogram at here cannot accept xarray.DataArray as its bin argument.

[Descanonge]

Hi, thanks!
I hadn't thought of time-varying bins, but I can see it implemented without too much trouble.
As for accepting xarray.DataArray for the bins argument, it is also feasible (I guess it would be a good idea to make it mandatory for time-varying bins).

This needs some refactoring so that the code stays clean but right now I don't see anything impossible. That would be nice since it seems from this comment that it won't be implemented in xhistogram before a while. I will work on that as soon as I have a solid base package (in a few days hopefully).

[miniufo]

That's a good news. I have to calculate PDF of a tracer field between min and max values and these values change with time (or even with depth). So I am expecting this behavior or similar:

# tracer[T, Z, Y, X] is a 4D variable
hist = xh.histogram(tracer, bins=[tracer.min(['X', 'Y']), tracer.max(['X', 'Y'])], dims=['X', 'Y'])
# output hist[T, Z, PDF] should be 3D variable

xhistogram cannot accept xarray as its bins argument because it relies on numpy. In addition, it does not support dask well. Looking forward to your solutions. I guess the bins argument accepting xarray.DataArray would be a very good point for your release.