Using series.str with compiled regular expressions. Problem with is_re

[florentguymartin]

I am using pandas 0.23.1 with python 3.6.4

import pandas as pd
import regex as re
s = pd.Series(['foo', 'bar'])
pat = re.compile('a')
s.str.replace(pat, ' ')

Raises the error

---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-5-4fc8cf4e83c0> in <module>()
----> 1 s.str.replace(pat, ' ')

~/anaconda3/lib/python3.6/site-packages/pandas/core/strings.py in replace(self, pat, repl, n, case, flags, regex)
   2427     def replace(self, pat, repl, n=-1, case=None, flags=0, regex=True):
   2428         result = str_replace(self._data, pat, repl, n=n, case=case,
-> 2429                              flags=flags, regex=regex)
   2430         return self._wrap_result(result)
   2431 

~/anaconda3/lib/python3.6/site-packages/pandas/core/strings.py in str_replace(arr, pat, repl, n, case, flags, regex)
    652             if case is False:
    653                 flags |= re.IGNORECASE
--> 654         if is_compiled_re or len(pat) > 1 or flags or callable(repl):
    655             n = n if n >= 0 else 0
    656             compiled = re.compile(pat, flags=flags)

TypeError: object of type '_regex.Pattern' has no len()

I think that the problem comes from the call to the function is_re, namely

from pandas.core.dtypes.common import is_re
is_re(pat)

returns False while I think it should return True.

[jreback]

this was fixed in #21098

@TomAugspurger this should have been backported?

[florentguymartin]

I just realized that the reason is the wrong import statement

which should have just been
import re
Sorry for that.

[TomAugspurger]

It'd be nice if we could support the 3rd party regex module, but I have no idea how much work that would be.