qcut raising ValueError if NaT present

[gulp21]

Code Sample, a copy-pastable example if possible

from io import StringIO
import pandas as pd

csv = 'Index,Date\n1,2013-01-01 23:00:00\n2,\n3,2013-01-01 23:00:01'
df = pd.read_csv(StringIO(csv), index_col=0, parse_dates=[1])
pd.qcut(df["Date"], 2)

Problem description

qcut raises a ValueError:

Traceback (most recent call last):
  File "mve.py", line 26, in <module>
    pd.qcut(df["Date"], 2)
  File "/tmp/test/env/lib/python3.5/site-packages/pandas/core/reshape/tile.py", line 208, in qcut
    dtype=dtype, duplicates=duplicates)
  File "/tmp/test/env/lib/python3.5/site-packages/pandas/core/reshape/tile.py", line 251, in _bins_to_cuts
    dtype=dtype)
  File "/tmp/test/env/lib/python3.5/site-packages/pandas/core/reshape/tile.py", line 344, in _format_labels
    labels = IntervalIndex.from_breaks(breaks, closed=closed)
  File "/tmp/test/env/lib/python3.5/site-packages/pandas/core/indexes/interval.py", line 370, in from_breaks
    name=name, copy=copy)
  File "/tmp/test/env/lib/python3.5/site-packages/pandas/core/indexes/interval.py", line 411, in from_arrays
    copy=copy, verify_integrity=True)
  File "/tmp/test/env/lib/python3.5/site-packages/pandas/core/indexes/interval.py", line 225, in _simple_new
    result._validate()
  File "/tmp/test/env/lib/python3.5/site-packages/pandas/core/indexes/interval.py", line 265, in _validate
    raise ValueError('missing values must be missing in the same '
ValueError: missing values must be missing in the same location both left and right sides

Expected Output

qcut returning something like

Index
1    (2013-01-01 22:59:59.999999999, 2013-01-01 23:00:01.0
2    NaT
3    (2013-01-01 22:59:59.999999999, 2013-01-01 23:00:01.0

Output of pd.show_versions()

INSTALLED VERSIONS

commit: None
python: 3.5.2.final.0
python-bits: 64
OS: Linux
OS-release: 4.13.0-32-generic
machine: x86_64
processor: x86_64
byteorder: little
LC_ALL: None
LANG: de_DE.UTF-8
LOCALE: de_DE.UTF-8

pandas: 0.22.0
pytest: None
pip: 9.0.1
setuptools: 38.5.1
Cython: None
numpy: 1.14.0
scipy: None
pyarrow: None
xarray: None
IPython: None
sphinx: None
patsy: None
dateutil: 2.6.1
pytz: 2018.3
blosc: None
bottleneck: None
tables: None
numexpr: None
feather: None
matplotlib: None
openpyxl: None
xlrd: None
xlwt: None
xlsxwriter: None
lxml: None
bs4: None
html5lib: None
sqlalchemy: None
pymysql: None
psycopg2: None
jinja2: None
s3fs: None
fastparquet: None
pandas_gbq: None
pandas_datareader: None

[gfyoung]

1. Could you rewrite your example to not use read_csv (i.e. just construct the DataFrame from scratch).

2. Yeah...that does look weird indeed. Even the error message is confusing. PR to patch is welcome!

[jschendel]

Issue looks to occur in the integer conversion here:

pandas/pandas/core/reshape/tile.py

Lines 290 to 291 in 80241e6

	elif is_datetime64_dtype(x):
	x = to_datetime(x).view(np.int64)

This converts NaT to a negative integer value:

In [2]: dti = pd.DatetimeIndex(['20180101', pd.NaT, '20180103'])

In [3]: dti
Out[3]: DatetimeIndex(['2018-01-01', 'NaT', '2018-01-03'], dtype='datetime64[ns]', freq=None)

In [4]: dti.view(np.int64)
Out[4]: array([ 1514764800000000000, -9223372036854775808,  1514937600000000000], dtype=int64)

So NaT is getting passed through as numeric, then looks to be assigned as the end point of an interval in the output after getting converted back to NaT, which is an invalid way to construct intervals.