TST: fixed cython doctests

[debnathshoham]

• closes TST: Add doctests on cython files #43757

Skipped some of the tests which were failing

[MarcoGorelli]

Nice work, good to see this happening!

Couple of questions:

• In Python docstrings, you can add lines with descriptions between examples (e.g. in https://pandas.pydata.org/pandas-docs/stable/development/contributing_docstring.html#section-7-examples) - is that not possible for Cython docstrings?
• Is it possible to not skip the tz_standardize tests, for example by constructing tz?

[debnathshoham]

thanks @MarcoGorelli for reviewing this!

• yes I think that's valid for cython docstrings as well. I have removed the # doctest: +SKIP for the line pandas supports PEP 3141 numbers: and it's running fine.

• I have tried my best and was able to do the first two examples, but not able to replicate the third one (I don't have much experience working with tz).

So currently, only two tests are being skipped, i.e. last example in tz and Int64Factorizer.factorize in hashtable.pyx

Edit - Another point worth highlighting, is the below is failing on my local (as the output is different than from the CI).

??? >>> pd.Timestamp.fromtimestamp(1584199972)
Expected:
    Timestamp('2020-03-14 15:32:52')
Got:
    Timestamp('2020-03-14 21:02:52')

[jreback]

why does this need skipping?

[debnathshoham]

I am not able to replicate this example

In [33]: from dateutil.tz.tz import tzutc as tz

In [34]: tz
Out[34]: dateutil.tz.tz.tzutc

In [35]: from pandas._libs.tslibs.timezones import tz_standardize

In [36]: tz_standardize(tz)
---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-36-94d2ff882b92> in <module>
----> 1 tz_standardize(tz)

TypeError: Argument 'tz' has incorrect type (expected datetime.tzinfo, got _TzSingleton)

[MarcoGorelli]

@jreback sorry to unresolve this, but can this example be removed if it can't be run?

[jbrockmendel]

Edit - Another point worth highlighting, is the below is failing on my local (as the output is different than from the CI). [...] ??? >>> pd.Timestamp.fromtimestamp(1584199972)

This is going to depend on what timezone you are in locally.

[debnathshoham]

Edit - Another point worth highlighting, is the below is failing on my local (as the output is different than from the CI). [...] ??? >>> pd.Timestamp.fromtimestamp(1584199972)

This is going to depend on what timezone you are in locally.

would be a good idea to skip this no? (or maybe make it timezone independent if possible, to ensure it doesn't fail in anyones local)

[MarcoGorelli]

IMO the

In [33]: from dateutil.tz.tz import tzutc as tz

In [34]: tz
Out[34]: dateutil.tz.tz.tzutc

In [35]: from pandas._libs.tslibs.timezones import tz_standardize

In [36]: tz_standardize(tz)
---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-36-94d2ff882b92> in <module>
----> 1 tz_standardize(tz)

TypeError: Argument 'tz' has incorrect type (expected datetime.tzinfo, got _TzSingleton)

test can be removed, if it doesn't run.

Other than that, the rest looks good to me

[MarcoGorelli]

Looks good to me

[jreback]

thanks @debnathshoham

i suppose that if we had more doc-tests in cython this would pick them up. if someone wants to create an issue about adding more doc-strings wouldn't object (for example there is an older PR that adds a ton to the timestamp functions).