resample function gives 0s instead of NaNs

[xzenggit]

What happened:
When I use resample(time='1d').sum(dim='time') to resample a time series with NaNs, the resampled result gives me 0s instead of NaNs, while NaNs should be the correct answer.

What you expected to happen:

NaNs should be the correct answer.

Minimal Complete Verifiable Example:

import xarray as xr

dates =  pd.date_range('20200101', '20200601', freq='h')
data = np.linspace(0, 10, num=len(dates))
data[0:30*24] = np.nan

da = xr.DataArray(data, coords=[dates], dims='time')
da.plot()

# Instead of NaNs, the resampled time series in January 20202 give us 0s, which not right.
da.resample(time='1d', skipna=True).sum(dim='time', skipna=True).plot()

Anything else we need to know?:

Did I misunderstand something here? Thanks!

Environment:
xarray - '0.15.1'

Output of xr.show_versions()

xarray - '0.15.1'

[max-sixty]

I agree this isn't that clear. I think min_count will solve this.

In [20]: da.resample(time='1d').sum(min_count=1)
Out[20]:
<xarray.DataArray (time: 153)>
array([         nan,          nan,          nan,          nan,
                nan,          nan,          nan,          nan,
                nan,          nan,          nan,          nan,

IIRC this mirrors pandas, would be keen to know if that's not the case

[xzenggit]

Thanks. It'll make more sense if the default value of min_count is 1 instead of 0.

[max-sixty]

Yes, IIRC there have been extensive discussion about this over at pandas.

To the the extent there's anything that doesn't pandas' convention, please let us know, that wouldn't be intentional.

I'll close this for now but please reopen with any other questions. Thanks for raising @xzenggit .