[FEA] Support accepting compiled regular expression objects from re.compile() as a pattern

[Nicholas-7]

I’d like for cuDF to support accepting compiled regular expression objects from re.compile() as a pattern similarly to how Pandas functions.

Example:

import re

cudf_regex_pat = re.compile(r"^.a|dog", flags=re.IGNORECASE)

cudfSeries4.str.replace(cudf_regex_pat, "XX-XX ", regex=True)

Result:

---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-50-65e18a60f111> in <module>
      3 cudf_regex_pat = re.compile(r"^.a|dog", flags=re.IGNORECASE)
      4 # TypeError: object of type 're.Pattern' has no len()
----> 5 cudfSeries4.str.(cudf_regex_pat, "XX-XX ", regex=True)

/opt/conda/envs/rapids/lib/python3.7/site-packages/cudf/core/column/string.py in replace(self, pat, repl, n, case, flags, regex)
    776         return self._return_or_inplace(
    777             cpp_replace_re(self._column, pat, cudf.Scalar(repl, "str"), n)
--> 778             if regex is True and len(pat) > 1
    779             else cpp_replace(
    780                 self._column,

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

Pandas does this by utilizing the replace method accepting a compiled regular expression object from re.compile() as a pattern. All flags should be included in the compiled regular expression object.:

import re
pandas_regex_pat = re.compile(r"^.a|dog", flags=re.IGNORECASE)
pandasSeries4.str.replace(pandas_regex_pat, "XX-XX ", regex=True)

Output

0 A
1 B
2 C
3 XX-XX ba
4 XX-XX ca
5
6
7 XX-XX BA
8 XX-XX
9 XX-XX t
dtype: string

[kkraus14]

Hmm, we can extract the pattern and the flags from the compiled regular expression here, but I don't see a clear way to translate that into something libcudf can use.

[github-actions]

This issue has been labeled inactive-30d due to no recent activity in the past 30 days. Please close this issue if no further response or action is needed. Otherwise, please respond with a comment indicating any updates or changes to the original issue and/or confirm this issue still needs to be addressed. This issue will be labeled inactive-90d if there is no activity in the next 60 days.

[davidwendt]

It looks like we could extract the pattern from the compiled regex object.

import re
p = re.compile(r"^.a|dog")
print(p.pattern)

^.a|dog

The p.pattern string could be passed to libcudf regex functions.

[beckernick]

For additional information: https://docs.python.org/3/library/re.html#re.Pattern.pattern