API: pandas.Series.replace has different method signature to pandas.DataFrame.replace

[roblevy]

Using Pandas 0.12.0

It's hard to think of why a DataFrame would allow a regex=True replace method, and a Series wouldn't. This must just be an oversight, right?

>>> test = pd.DataFrame(['hello','goodbye','thankyou'],columns=['to_replace'])
>>> test
  to_replace
0      hello
1    goodbye
2   thankyou
>>> test.replace("he", "je")
  to_replace
0      hello
1    goodbye
2   thankyou
>>> test.replace("he", "je", regex=True)
  to_replace
0      jello
1    goodbye
2   thankyou
>>> test['to_replace'].replace("he", "je", regex=True)
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
TypeError: replace() got an unexpected keyword argument 'regex'

[cpcloud]

I think this will be fixed in 0.13 by the Series -> NDFrame refactor.

[jreback]

yep...#3482

[jreback]

@roblevy this is why the refactor is happening in the first place! (to provide this kind of consistency)

[roblevy]

Wow. I've just been pulled briefly into the whirlwind of development at the heart of the pandas project. So much going on that I (as a mere mortal user of the library) had no idea of. Look forward to 0.13.0. Keep up the amazing work!

[jreback]

related to #4118 / closed by #3482