[v3] support for datetime and timedelta dtypes

[jhamman]

Zarr-Python 2 supported datetime and timedelta dtypes. This functionality has not made it into Zarr-Python 3 yet (in part because the v3 spec does not have established extension dtypes for these yet).

Example of the behavior for Zarr 2:

z = zarr.array(['2007-07-13', '2006-01-13', '2010-08-13'], dtype='M8[D]')
z
<zarr.core.Array (3,) datetime64[D]>
z[:]
array(['2007-07-13', '2006-01-13', '2010-08-13'], dtype='datetime64[D]')
z[0]
numpy.datetime64('2007-07-13')
z[0] = '1999-12-31'
z[:]
array(['1999-12-31', '2006-01-13', '2010-08-13'], dtype='datetime64[D]')

This issue tracks the development of datetime and timedelta support in Zarr-Python 3.

[christine-e-smit]

Hey Joe! All our zarr stores use the datetime type for our time dimension variable. It sounds like we'll have to transform all our zarr stores or help you make the change to support this data. What kind of effort do you think would be required for relative newbies to add datetime data type support to zarr v3?

[jhamman]

#2874 should be merged in the next week or so and will include support for datetime dtypes.

[christine-e-smit]

Oh I'm sorry. My eyes went right over the links in this ticket for some reason. Thank you.

[christine-e-smit]

We are now actively in the process of updating our code. It looks as though #2874 has not been merged. Any idea on the ETA? Do we need to delay our migration?