Series.str.split() behavior on multi-character patterns (pandas 0.9.1, py2.7)

[gdraps]

Not sure if this is an issue in my setup or the intended behavior, but str.split behavior with multi-character patterns changed between 0.9.0 and 0.9.1, most likely due to #2119. Here is the original behavior going back to 0.8.1 as best I can tell.

In [1]: pd.__version__
Out[1]: '0.9.1.dev-c252129'

In [2]: s = pd.Series(["D0->D2"])

In [3]: s.str.split("->")
Out[3]: 0    [D0, D2]

In 0.9.1:

In [1]: pd.__version__
Out[1]: '0.9.1'

In [2]: s = pd.Series(["D0->D2"])

In [3]: s.str.split("->")
Out[3]: 0    [D0->D2]

Setting n=0 restores the behavior, on my Python install at least.

In [5]: s.str.split("->", n=0)
Out[5]: 0    [D0, D2]

Reproducible in 0.10.0b1.

In [1]: pd.__version__
Out[1]: '0.10.0b1'

In [2]: s = pd.Series(["D0->D2"])

In [3]: s.str.split("->")
Out[3]: 0    [D0->D2]

Python 2.7.2+ (default, Jul 20 2012, 22:12:53)
[GCC 4.6.1] on linux2

Any thoughts on changing the default n for str.split back to 0?

Many thanks!

[wesm]

thanks for the report, look like a bug. will have a look

[gdraps]

Dug a little and found that str.split and re.split treat maxsplit differently: str.split returns all splits when maxsplit is -1, while re.split does the same when maxsplit is 0.

[changhiskhan]

n is None by default now and {None, 0, -1} will all have the same behavior