fillna after resample doesn't work on version 0.20.1

[jmilhomens]

Code Sample, a copy-pastable example if possible

import pandas as pd
df = pd.DataFrame({ 'values': range(5)}, index = pd.date_range(start = '2017-01-01 00:00:00', periods = 5, freq='S'))
resampled_df = df.resample('500ms').mean()
resampled_df.fillna(method='ffill')

Problem description

The fillna method is not working after a resample. The result of the above code is:

	values
2017-01-01 00:00:00.000	0.0
2017-01-01 00:00:00.500	NaN
2017-01-01 00:00:01.000	1.0
2017-01-01 00:00:01.500	NaN
2017-01-01 00:00:02.000	2.0
2017-01-01 00:00:02.500	NaN
2017-01-01 00:00:03.000	3.0
2017-01-01 00:00:03.500	NaN
2017-01-01 00:00:04.000	4.0

This code worked with version 0.19.2

Expected Output

The expected output in this case can be also obtained with:

df.resample('500ms').ffill()

The result is:

	values
2017-01-01 00:00:00.000	0
2017-01-01 00:00:00.500	0
2017-01-01 00:00:01.000	1
2017-01-01 00:00:01.500	1
2017-01-01 00:00:02.000	2
2017-01-01 00:00:02.500	2
2017-01-01 00:00:03.000	3
2017-01-01 00:00:03.500	3
2017-01-01 00:00:04.000	4

However, in some cases one may need to do the re-sample and fill missing values in separate steps.

Output of pd.show_versions()

INSTALLED VERSIONS

commit: None
python: 3.6.1.final.0
python-bits: 64
OS: Darwin
OS-release: 16.5.0
machine: x86_64
processor: i386
byteorder: little
LC_ALL: en_US.UTF-8
LANG: en_US.UTF-8
LOCALE: en_US.UTF-8

pandas: 0.20.1
pytest: None
pip: 9.0.1
setuptools: 27.2.0
Cython: 0.25.2
numpy: 1.12.1
scipy: 0.19.0
xarray: None
IPython: 6.0.0
sphinx: None
patsy: None
dateutil: 2.6.0
pytz: 2017.2
blosc: None
bottleneck: None
tables: None
numexpr: None
feather: None
matplotlib: 2.0.2
openpyxl: None
xlrd: None
xlwt: None
xlsxwriter: None
lxml: None
bs4: None
html5lib: 0.999
sqlalchemy: None
pymysql: None
psycopg2: None
jinja2: 2.9.6
s3fs: None
pandas_gbq: None
pandas_datareader: None

[TomAugspurger]

Seems at least partially related to

pandas/pandas/core/internals.py

Line 915 in b0038ac

def check_int_bool(self, inplace):

returning True, when it should be False (we have ints + nans, so it should be float). I guess the root cause is resampled_df having an IntBlock

In [6]: resampled_df._data
Out[6]:
BlockManager
Items: Index(['values'], dtype='object')
Axis 1: DatetimeIndex([       '2017-01-01 00:00:00', '2017-01-01 00:00:00.500000',
                      '2017-01-01 00:00:01', '2017-01-01 00:00:01.500000',
                      '2017-01-01 00:00:02', '2017-01-01 00:00:02.500000',
                      '2017-01-01 00:00:03', '2017-01-01 00:00:03.500000',
                      '2017-01-01 00:00:04'],
              dtype='datetime64[ns]', freq='500L')
IntBlock: slice(0, 1, 1), 1 x 9, dtype: float64

potentially related to #16361

[jreback]

this is a dupe of #16361