infer_freq broken in 0.13.1

[dhirschfeld]

In [3]: dates = pd.date_range('01-Jan-2015', '01-Jan-2016', freq='MS')
   ...: s = pd.TimeSeries(1, dates)
   ...: pd.infer_freq(s)
   ...: 
Traceback (most recent call last):

  File "<ipython-input-3-a7de7a7e9245>", line 3, in <module>
    pd.infer_freq(s)

  File "C:\dev\bin\Anaconda\lib\site-packages\pandas\tseries\frequencies.py", line 656, in infer_freq
    inferer = _FrequencyInferer(index, warn=warn)

  File "C:\dev\bin\Anaconda\lib\site-packages\pandas\tseries\frequencies.py", line 680, in __init__
    self.is_monotonic = self.index.is_monotonic

  File "C:\dev\bin\Anaconda\lib\site-packages\pandas\core\generic.py", line 1815, in __getattr__
    (type(self).__name__, name))

AttributeError: 'Series' object has no attribute 'is_monotonic'


In [4]: pd.__version__
Out[4]: '0.13.1'

[jreback]

This is actually an enhancement it didn't work in 0.12 (just gave a wrong answer!)

takes a DatetimeIndex only

In [10]: pd.infer_freq?
Type:       function
String Form:<function infer_freq at 0x3b207d0>
File:       /mnt/home/jreback/pandas/pandas/tseries/frequencies.py
Definition: pd.infer_freq(index, warn=True)
Docstring:
Infer the most likely frequency given the input index. If the frequency is
uncertain, a warning will be printed

Parameters
----------
index : DatetimeIndex
warn : boolean, default True

Returns
-------
freq : string or None
    None if no discernible frequency

works correctly on the index

In [12]: pd.infer_freq(s.index)
Out[12]: 'MS'

out of curiosity, what is your usecase for this?

[dhirschfeld]

Right - UserError!

I want to create a contiguous date range at the same frequency as another series where the other series may have its freqstr equal to None. The "same frequency" is defined as the lowest frequency such that a contiguous date range will include every date in the other series - e.g. in the trivial case
microsecond frequency will interpolate any series.

[jreback]

ok..thanks

your example will raise (as its using the VALUES of the Series, otherwise this is very confusing).
Rather than return None (which IMHO is wrong). FYI their is no real distinction between TimeSeries and Series since 0.13