Enable passing a CFTimedeltaCoder to decode_timedelta
#9966
Conversation
|
Yes, I was just playing around a bit. Please go ahead, @spencerkclark. I'll catch up tomorrow. |
There was a problem hiding this comment.
Thanks @spencerkclark! One suggestion for adding a FutureWarning and explicit mention of decode_times=False in the docs.
| .. ipython:: python | ||
|
|
||
| coder = xr.coders.CFDatetimeCoder(time_unit="s") | ||
| xr.open_dataset("test-timedeltas2.nc", decode_times=coder) |
There was a problem hiding this comment.
There is still some unresolved issue, where we should think about a way forward. At least we could document it, like this:
| xr.open_dataset("test-timedeltas2.nc", decode_times=coder) | |
| xr.open_dataset("test-timedeltas2.nc", decode_times=coder) | |
| To opt-out of timedelta decoding (see issue `Undesired decoding to timedelta64 <https://github.com/pydata/xarray/)issues/1621>`_) pass ``False`` to ``decode_timedelta``: | |
| .. ipython:: python | |
| xr.open_dataset("test-timedeltas2.nc", decode_times=False) | |
I'm not sure if it would be a good idea to change the default to decode_times=False in this cycle of changes (as discussed in #1621), but we could at least add a future warning when decode_timedelta=None. WDYT?
There was a problem hiding this comment.
Yes, I think that is a good idea—thanks for reminding me about that issue. I'll work on an update when I get a chance.
| @@ -171,7 +173,10 @@ def decode_cf_variable( | |||
| original_dtype = var.dtype | |||
|
|
|||
| if decode_timedelta is None: | |||
| decode_timedelta = True if decode_times else False | |||
| if isinstance(decode_times, CFDatetimeCoder): | |||
There was a problem hiding this comment.
Here we could add a FutureWarning, that decode_timedelta=None will default to decode_timedelta=False. See #1621.
I had some time this weekend so I worked on enabling passing a
CFTimedeltaCodertodecode_timedeltainopen_datasetet al.@kmuehlbauer I didn't realize until this morning that you had started on this too in your fork (branch)—I was able to make a simplification to the
dtypeinference code I originally had based on looking at your initial work. I'm posting this here, since it seems a little further along than your branch, but feel free to supersede this PR with one of your own if you have un-pushed progress. I listed this PR in the same what's new entry as #9618, since it seemed natural, and gives us both credit.It looks like we were fairly aligned though in terms of inheriting the
time_unitfromdecode_timesif nothing was explicitly passed todecode_timedelta, which is good.whats-new.rst