Calling pandas.cut with series of timedelta and timedelta bins raises TypeError, but should succeed

[nmusolino]

Problem description

Attempting to call pandas.cut with a Series that contains timedelta values and bins that are timedelta values fails with TypeError.

Code Sample

In [1]: import datetime

In [2]: import numpy

In [3]: import pandas

In [4]: s = pandas.date_range('1/1/2018', periods=5, freq='s').to_series() - datetime.datetime(2018, 1, 1, 0, 0, 0)

In [5]: s
Out[5]:
2018-01-01 00:00:00   00:00:00
2018-01-01 00:00:01   00:00:01
2018-01-01 00:00:02   00:00:02
2018-01-01 00:00:03   00:00:03
2018-01-01 00:00:04   00:00:04
Freq: S, dtype: timedelta64[ns]

In [6]: pandas.cut(s, bins=[numpy.timedelta64(t, 's') for t in [0, 2, 5]])
---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-6-42f313379671> in <module>()
----> 1 pandas.cut(s, bins=[numpy.timedelta64(t, 's') for t in [0, 2, 5]])
      2
      3

C:\...\lib\site-packages\pandas\tools\tile.py in cut(x, bins, right, labels, retbins, precision, include_lowest)
    112     else:
    113         bins = np.asarray(bins)
--> 114         if (np.diff(bins) < 0).any():
    115             raise ValueError('bins must increase monotonically.')
    116

TypeError: Cannot cast ufunc less input from dtype('<m8[s]') to dtype('<m8') with casting rule 'same_kind'

This also occurs when using nanosecond timedeltas ([11]), and with pandas.Timedelta values ([14])

In [9]: bins = numpy.array([numpy.timedelta64(t * 1000000000, 'ns') for t in [0, 2, 5]])

In [10]: bins
Out[10]: array([         0, 2000000000, 5000000000], dtype='timedelta64[ns]')

In [11]: pandas.cut(s, bins=bins)
---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-11-83e991bf4051> in <module>()
----> 1 pandas.cut(s, bins=bins)

C:\...\lib\site-packages\pandas\tools\tile.py in cut(x, bins, right, labels, retbins, precision, include_lowest)
    112     else:
    113         bins = np.asarray(bins)
--> 114         if (np.diff(bins) < 0).any():
    115             raise ValueError('bins must increase monotonically.')
    116

TypeError: Cannot cast ufunc less input from dtype('<m8[ns]') to dtype('<m8') with casting rule 'same_kind'

In [14]: pandas.cut(s, bins=[pandas.to_timedelta(t, unit='s') for t in [0, 2, 5]])
---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-14-8f423565c547> in <module>()
----> 1 pandas.cut(s, bins=[pandas.to_timedelta(t, unit='s') for t in [0, 2, 5]])

C:\...\lib\site-packages\pandas\tools\tile.py in cut(x, bins, right, labels, retbins, precision, include_lowest)
    112     else:
    113         bins = np.asarray(bins)
--> 114         if (np.diff(bins) < 0).any():
    115             raise ValueError('bins must increase monotonically.')
    116

pandas\tslib.pyx in pandas.tslib._Timedelta.__richcmp__ (pandas\tslib.c:46569)()

TypeError: Cannot compare type 'Timedelta' with type 'int'

On line 114 of tile.py in the exception traceback above:

if (np.diff(bins) < 0).any():

the comparison to zero should really be a comparison to zero in the datatype of the series.

Expected Output

The call should succeed, returning a series of bins that correspond to each element.

Output of pd.show_versions()

INSTALLED VERSIONS

commit: None
python: 3.4.5.final.0
python-bits: 64
OS: Windows
OS-release: 7
machine: AMD64
processor: Intel64 Family 6 Model 79 Stepping 1, GenuineIntel
byteorder: little
LC_ALL: None
LANG: en_US.UTF-8
LOCALE: None.None

pandas: 0.19.1
nose: 1.3.7
pip: 9.0.1
setuptools: 27.2.0
Cython: 0.24.1
numpy: 1.11.2
scipy: 0.18.1
statsmodels: 0.6.1
xarray: 0.8.2
IPython: 5.1.0
sphinx: 1.4.8
patsy: 0.4.1
dateutil: 2.6.0
pytz: 2016.7
blosc: 1.5.0
bottleneck: 1.2.0
tables: 3.2.2
numexpr: 2.6.1
matplotlib: 2.0.0
openpyxl: 2.4.0
xlrd: 1.0.0
xlwt: 1.1.2
xlsxwriter: 0.9.3
lxml: 3.6.4
bs4: 4.5.3
html5lib: 0.999
httplib2: 0.9.2
apiclient: None
sqlalchemy: 1.1.3
pymysql: None
psycopg2: 2.6.2 (dt dec pq3 ext lo64)
jinja2: 2.8
boto: 2.43.0
pandas_datareader: None

[nmusolino]

See also issue #19891.

[gfyoung]

the comparison to zero should really be a comparison to zero in the datatype of the series.

Does that patch work for you BTW?

[mroeschke]

Looks like this was fixed back in v0.20

pandas/doc/source/whatsnew/v0.20.0.txt

Line 479 in 24ab22f

- ``pd.cut`` and ``pd.qcut`` now support datetime64 and timedelta64 dtypes (:issue:`14714`, :issue:`14798`)

And we have tests for testing timedelta bins with cut.