Set pandas.tseries.index.DatetimeIndex.freq with inferred_freq

Question

consider the DatetimeIndex tidx

tidx = pd.to_datetime([\'2016-07-29\', \'2016-08-31\', \'2016-09-30\'])
print(tidx.freq)
print(tidx         


        

Answer 1

It's unclear why the docs state you can set the freq attribute but then it doesn't persist but if you reconstruct the datetimeindex again but pass a freq param then it works:

In [56]:
tidx = pd.DatetimeIndex(tidx.values, freq = tidx.inferred_freq)
tidx

Out[56]:
DatetimeIndex(['2016-07-29', '2016-08-31', '2016-09-30'], dtype='datetime64[ns]', freq='BM')

Answer 2

https://stackoverflow.com/a/40223868/2336654

I asked another question to help with this. @root identified a function for converting frequency strings. So this should work

tidx.freq = pd.tseries.frequencies.to_offset(tidx.inferred_freq)

DatetimeIndex(['2016-07-29', '2016-08-31', '2016-09-30'],
              dtype='datetime64[ns]', freq='BM')

Answer 3

You can directly use the DatetimeIndex constructor with your list of strings and pass 'infer' as the freq:

In [2]: tidx = pd.DatetimeIndex(['2016-07-29', '2016-08-31', '2016-09-30'], freq='infer')

In [3]: tidx
Out[3]: DatetimeIndex(['2016-07-29', '2016-08-31', '2016-09-30'], dtype='datetime64[ns]', freq='BM')