numpy datetime conversion with DataArray is not working

[jules-ch]

What happened?

I have a simple DataArray with datetime[ns] inside, but when I try to convert this to a supported numpy datetime dtype, the results is still a datetime[ns].

Using the values attribute & converting it, returns the correct results

What did you expect to happen?

I excpected the DataArray to be converted in the correct dtype

Minimal Complete Verifiable Example

>>> time = pd.date_range('2015-01-01 00:00:00', '2015-12-31 23:59:00', inclusive='left', freq='15min')
>>> da = xr.DataArray(time)
>>> da
<xarray.DataArray (dim_0: 35040)>
array(['2015-01-01T00:00:00.000000000', '2015-01-01T00:15:00.000000000',
       '2015-01-01T00:30:00.000000000', ..., '2015-12-31T23:15:00.000000000',
       '2015-12-31T23:30:00.000000000', '2015-12-31T23:45:00.000000000'],
      dtype='datetime64[ns]')
Coordinates:
  * dim_0    (dim_0) datetime64[ns] 2015-01-01 ... 2015-12-31T23:45:00

>>> da.astype("datetime64[Y]")
<xarray.DataArray (dim_0: 35040)>
array(['2015-01-01T00:00:00.000000000', '2015-01-01T00:00:00.000000000',
       '2015-01-01T00:00:00.000000000', ...,
       '2015-01-01T00:00:00.000000000', '2015-01-01T00:00:00.000000000',
       '2015-01-01T00:00:00.000000000'], dtype='datetime64[ns]')
Coordinates:
  * dim_0    (dim_0) datetime64[ns] 2015-01-01 ... 2015-12-31T23:45:00
>>> da.values.astype("datetime64[Y]")
array(['2015', '2015', '2015', ..., '2015', '2015', '2015'],
      dtype='datetime64[Y]')

Relevant log output

No response

Anything else we need to know?

No response

Environment

INSTALLED VERSIONS

commit: None
python: 3.8.10 (default, Nov 26 2021, 20:14:08)
[GCC 9.3.0]
python-bits: 64
OS: Linux
OS-release: 5.13.0-37-generic
machine: x86_64
processor: x86_64
byteorder: little
LC_ALL: None
LANG: en_US.UTF-8
LOCALE: ('en_US', 'UTF-8')
libhdf5: 1.12.1
libnetcdf: None

xarray: 0.20.2
pandas: 1.4.1
numpy: 1.22.3
scipy: 1.8.0
netCDF4: None
pydap: None
h5netcdf: None
h5py: 3.6.0
Nio: None
zarr: None
cftime: None
nc_time_axis: None
PseudoNetCDF: None
rasterio: None
cfgrib: None
iris: None
bottleneck: None
dask: 2022.03.0
distributed: 2022.3.0
matplotlib: 3.5.1
cartopy: None
seaborn: None
numbagg: None
fsspec: 2022.02.0
cupy: None
pint: 0.18
sparse: None
setuptools: 60.6.0
pip: 22.0.3
conda: None
pytest: 6.2.5
IPython: 8.1.1
sphinx: None

[spencerkclark]

This is a common point of confusion, but is in fact expected. Xarray intentionally converts any np.datetime64 type to np.datetime64[ns]. The primary motivation is compatibility with pandas, which xarray relies on for time indexing and other time-related operations through things like pandas.DatetimeIndex or the pandas.Series.dt accessor (see, e.g., discussion in #789).

May I ask what your reason is for requiring a lower-precision datetime type? In xarray we have tried to provide alternatives (e.g. through cftime) for use-cases like longer date ranges.

[jules-ch]

I'm using xarray on a Dataset & it's convenient for me to make calculation using DataArray. Here when I want to retrieve the year of datetime, instead of casting back to an array of object & using datetime.year, it's handy to use built-in numpy datetime64 conversion.

It's really confusing astype is not working like numpy does. If you want to keep this behavious maybe add a warning in the docs and log a warning aswell.

[spencerkclark]

I agree this should be better documented.

Here when I want to retrieve the year of datetime, instead of casting back to an array of object & using datetime.year, it's handy to use built-in numpy datetime64 conversion.

The recommended way to extract datetime components, like the year, is to use the DatetimeAccessor. For example if you have a DataArray of datetime-like values (whether they are of type np.datetime64[ns] or cftime.datetime) you can do something like this:

da.dt.year

and it will return a DataArray containing the year of each datetime (for more information see the "Datetime components" section of the documentation).