Spencer Clark

Apologies for taking a while to look into this. I have not been able to set up an environment to reproduce these test failures, which makes it tricky. It seems...

Thanks for trying that. Maybe it has to do with casting to a `pd.Series`. Could you maybe also try: ``` >>> import numpy as np; import pandas as pd >>>...

Thanks @amckinstry. I guess my last try to produce a pandas minimal example might be: ``` >>> import numpy as np; import pandas as pd >>> pd.Series(np.array([np.int64(1000000).astype("

Interesting. Thanks for checking that #6988 indeed solves this. I went ahead and merged it, but when I get a chance I’ll keep trying to track down the root cause...

I was able to reproduce this issue in a Docker container using the [s390x Debian image](https://hub.docker.com/r/s390x/debian/). After a little experimentation I narrowed it down to the following minimal example: ```...

Very interesting. I agree it seems there is significant room for improvement. I *think* we may be able to follow a similar line of thought as to what led to...

Oops `_dayofwk` and `_dayofyr` are the cached values so they do not trigger computation. Therefore I suspect my proposed strategy will not make a difference.

A fairer comparison might be to look at the performance of pickling an array of `datetime.datetime` objects. There we find pickling an array of cftime objects is ~3x slower --...

I'm somewhat surprised that encoding a cftime array is faster than pickling it (encoding requires repeated timedelta arithmetic, which is not needed for pickling). Have you done timing experiments to...

Ah...that makes perfect sense now, thanks. Indeed it does seem like the optimization might best take place before cftime is involved. If you can put together a simple example that...