rolling with datetimeIndex is not selecting anything when the row value is NaN itself

[staftermath]

Code Sample, a copy-pastable example if possible

import numpy as np
import pandas as pd
import datetime

df = pd.DataFrame({'a': [1, np.nan, 5],
                   'datetime':pd.to_datetime(['20170403', '20170404', '20170405'])})
df.set_index('datetime', inplace=True)
df.rolling('2d', 1).apply(np.nanmin)

Problem description

the above generates

              a
datetime       
2017-04-03  1.0
2017-04-04  NaN
2017-04-05  5.0

when use datetimeIndex, if a row has NaN value it is not fed to input of apply when creating output for that row. To illustrate this, I use:

def foo(x):
    print(x)
    return np.nanmin(x)
print(df.rolling('2d', 1).apply(foo))

and got printed value:

[ 1.]
[ nan   5.]

the second row is skipped.
This issue doesn't not exist when integer index is used:

df = pd.DataFrame({'a': [1, np.nan, 5]})
def foo(x):
    print(x)
    return np.nanmin(x)
print(df.rolling(2, 1).apply(foo))

print the following:

[ 1.]
[  1.  nan]
[ nan   5.]

and yields correct results

Expected Output

              a
datetime       
2017-04-03  1.0
2017-04-04  1.0
2017-04-05  5.0

Output of pd.show_versions()

INSTALLED VERSIONS

---

commit: None
python: 3.5.2.final.0
python-bits: 64
OS: Darwin
OS-release: 15.6.0
machine: x86_64
processor: i386
byteorder: little
LC_ALL: None
LANG: None
LOCALE: en_US.UTF-8

pandas: 0.20.3
pytest: 2.8.5
pip: 8.1.2
setuptools: 25.1.6
Cython: 0.23.4
numpy: 1.13.1
scipy: 0.17.0
xarray: None
IPython: 5.1.0
sphinx: 1.4.1
patsy: 0.4.0
dateutil: 2.6.1
pytz: 2017.2
blosc: None
bottleneck: 1.0.0
tables: 3.2.2
numexpr: 2.4.6
feather: None
matplotlib: 1.5.1
openpyxl: 2.3.2
xlrd: 0.9.4
xlwt: 1.0.0
xlsxwriter: 0.8.4
lxml: 3.5.0
bs4: 4.4.1
html5lib: None
sqlalchemy: 1.0.11
pymysql: None
psycopg2: None
jinja2: 2.8
s3fs: None
pandas_gbq: None
pandas_datareader: None

[gfyoung]

@staftermath : That does look a little strange. Investigation and PR to patch are welcome!

[jreback]

duplicate of this: #15305

fixes are welcome.