Pandas: ENH: Add `into` argument for `to_xarray` method?

[jolespin]

Is your feature request related to a problem?

I would like to able to specify the object used for creating xarray objects.

Describe the solution you'd like

Right now, the only option is xr.Dataset but I tend to prefer xr.DataArray objects.

API breaking implications

It shouldn't affect the API in any complicated ways.

Describe alternatives you've considered

Create a xr.Dataset and then convert that to xr.DataArray. However, this step could be obviated by usage similar to to_dict method.

Additional context

import pandas as pd
import xarray as xr
df = pd.read_csv("https://pastebin.com/raw/dR59vTD4", sep="\t", index_col=0)
da = df.to_xarray(into=xr.DataArray)

[TomAugspurger]

This would only work for homogenous dataframes, correct?

cc @shoyer and @max-sixty if either of you have thoughts.

[jolespin]

By homogenous, do you mean all the same dtype? If so, I am not sure but for the data I've been working with lately they are all floats (nan too).

[TomAugspurger]

Yeah, homogenous means all columns have the same dtype. A DataArray can only hold a single dtype, so if you have a mix of datatypes storing it in a DataArray wouldn't be an option (short of casting to a common dtype, like object).

[jolespin]

What about a check like this?

if into == xr.DataArray:
    assert df.dtypes.nunique() == 1
   return xr.DataArray(df, ...

[shoyer]

You can always use xr.DataArray(df) or df.pipe(xr.DataArray) (if you want method chaining) for this sort of conversion. I don't see a strong need to make to_xarray() more complicated to support this.

[TomAugspurger]

Thanks Stephan, I think that aligns with my thoughts.

Thanks for the proposal @jolespin, but we'll pass.

[jolespin]

No worries, thanks for considering it. xr.DataArray(df) should do the trick.

Are xr.Datasets typically preferred over xr.DataArray objects even for homogenous data?

[shoyer]

Whether to prefer xr.Dataset or xr.DataArray really depends on your
use-case and what the data represents. But xr.Dataset is a better fit for
pd.DataFrame organized according to the principle of "tidy data", so that's
why it's the default for to_xarray().

On Tue, Jul 14, 2020 at 11:48 AM Josh L. Espinoza [e-mail]
wrote:

No worries, thanks for considering it. xr.DataArray(df) should do the
trick.

Are xr.Datasets typically preferred over xr.DataArray objects even for
homogenous data?

—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
https://github.com/pandas-dev/pandas/issues/35265#issuecomment-658349937,
or unsubscribe
https://github.com/notifications/unsubscribe-auth/AAJJFVRP5KPVCJO7JK6GQA3R3SSAHANCNFSM4OYWS2LA
.