Rank function

[0x0L]

Hi,

I think xarray is missing a rank function.
Is there any reason not to expose a wrapper to bottleneck.nanrankdata ?

See also #1635

[edit]
Although moving rank is mentioned in the whats-new for v0.9.2 I wasn't able to find that functionality nor a trace of it in the code :)

[jhamman]

Is there any reason not to expose a wrapper to bottleneck.nanrankdata

@0x0L - I don't think so and I think we'd be open to adding this function. Even better if there is a fallback numpy equivalent but I don't think that would be required.

I looked at the (my) whatsnew note from 0.9.2 and I it seems we decided to remove this option until there is a rank method for dataarray/dataset objects. See @shoyer's comment: #1278 (comment)

[max-sixty]

Great idea.

We use bottleneck.nanrankdata manually; I don't think there's a vectorized numpy fallback

[0x0L]

A few points:

• nanrankdata only support numeric types, how about wrapping rankdata with a keyword option ?
• There's also a push push method similar to pandas ffill which would be nice.
• move_rank outputs are not normalized as rankdata, this might be a bit confusing

[shoyer]

nanrankdata only support numeric types, how about wrapping rankdata with a keyword option ?

We already do dispatching to appropriate functions based on the dtype for aggregations:

xarray/xarray/core/duck_array_ops.py

Line 174 in 9d09c16

def _create_nan_agg_method(name, numeric_only=False, np_compat=False,

(Yes, this is a bit of a mess)

Since NaN has a consistent sorting position in NumPy/bottleneck (it sorts to the end), I would suggest including a skipna keyword argument, like one we use for aggregation functions. Alternatively, we could use na_option : {‘keep’, ‘top’, ‘bottom’} like pandas.

There's also a push push method similar to pandas ffill which would be nice.

@jhamman is already working on ffill/bfill in #1640