Reimplement GroupBy.argmax

[Hoeze]

Please implement

Is your feature request related to a problem? Please describe.
Observed:

da.groupby("g").argmax(dim="t")

---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
<ipython-input-84-15c199b0f7d4> in <module>
----> 1 da.groupby("g").argmax(dim="t")

AttributeError: 'DataArrayGroupBy' object has no attribute 'argmax'

Describe the solution you'd like
Expected: Vector of length len(unique(g)) containing the indices of da["t"] where the value was maximum.

Workaround:

da.groupby("g").apply(lambda c: c.argmax(dim="t"))

<xarray.DataArray 'da' (st: 11, g: 1)>
array([[ 7],
       [ 0],
       [14],
       [14],
       [ 0],
       [ 0],
       [ 7],
       [ 0],
       [14],
       [ 0],
       [ 7]])
Coordinates:
  * st      (st) object 'a' ... 'z'
  * g       (g) object 'E'

[markusritschel]

This was definitely possible in the last versions (at least in v.0.15.1). Can we make this available again?

[zxdawn]

@markusritschel I tested v0.17.0 and it doesn't work.

data = np.random.rand(4, 3)
locs = ["IA", "IL", "IN"]
times = pd.date_range("2000-01-01", periods=4)
foo = xr.DataArray(data, coords=[times, locs], dims=["time", "space"])
foo.groupby('space').argmax('time')

Error: AttributeError: 'DataArrayGroupBy' object has no attribute 'argmax'

[markusritschel]

@zxdawn I am pretty sure it was possible in v0.15.1. But I would need to check that again.

edit: I just checked the old documentation from that version, and it confirms my memory:
http://xarray.pydata.org/en/accessor-documentation/generated/xarray.core.groupby.DataArrayGroupBy.argmax.html

[sjvrijn]

Using git bisect and @zxdawn's example, I've narrowed it down to commit bdcfab5.

I'm guessing the exact culprit is the removal of argmin and argmax from NAN_REDUCE_METHODS in xarray/core/ops.py because dedicated implementations were added for Variable/DataSet/DataArray.

[emanuelegent]

Hi I would just like to add that having the argmax functionality would make possible to extract timestamp of maximum values, such as daily maximum values. Current resample in xarray doesn't allow that

[max-sixty]

We have fast versions of these in Numbagg — e.g. group_nanargmax — so these should be possible to add without too much trouble...