series.str.extract does not work for timeseries

[andrewkittredge]

series.str.extract does not work for time-series because core.strings.str_extract does not preserve the index. I am submitting a unittest and patch that demonstrates and hopefully fixes the issue.

[andrewkittredge]

apparently duplicated by my pull request.

[jreback]

when I merge your PR this will close automatically

[jreback]

closed via 8efa4ac

[patricksurry]

Got bitten by this - it also silently gives the wrong answer if you've made a selection on a default index (example below). Work around is to explicitly reset_index(drop=True) before using str.extract.

print pd.__version__
df = pd.DataFrame([
    {'a': 'quick', 'b': 0}, 
    {'a': 'brown', 'b': 1}, 
    {'a': 'fox', 'b': 0},
    {'a': 'jumped', 'b': 1}
])
print df
df2 = df[df.b == 1]
df2['b'] = df2.a.str.extract('(p)')
print df2

output:

0.13.1
        a  b
0   quick  0
1   brown  1
2     fox  0
3  jumped  1

[4 rows x 2 columns]
        a    b
1   brown    p
3  jumped  NaN

[2 rows x 2 columns]

[jreback]

this was fixed for 0.14.0

0.14.0rc1 was just released, see here: https://groups.google.com/forum/m/#!topic/pydata/Bx8oiSAZwOc

[patricksurry]

Yup, thanks, I saw it was closed already, but I just wanted to leave a trail in case anyone else ran into this.