Odd behaviour of resample+mean+interpolate on int64 series

[myyc]

this issue is present on the latest stable release (as well as latest master at the time of this writing). for frames with only int64 values, the following has strange behaviour

df = {"a": [1,3,1,4]}
df = pd.DataFrame(df, index=pd.date_range("2017-01-01", "2017-01-04"))

# these two are not the same
df.resample("H").mean()["a"].interpolate("cubic")  # bad
df.resample("H")["a"].mean().interpolate("cubic")  # good

# this works
df.astype("float64").resample("H").mean()["a"].interpolate("cubic")

my workaround is more than enough for me but i figured i'd report it anyway...

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_IE.UTF-8 LANG: en_IE.UTF-8 LOCALE: en_IE.UTF-8

pandas: 0.21.0.dev+31.g0ea0f25bf
pytest: 3.0.7
pip: 9.0.1
setuptools: 35.0.2
Cython: 0.25.2
numpy: 1.12.1
scipy: 0.19.0
xarray: None
IPython: 6.0.0
sphinx: None
patsy: 0.4.1
dateutil: 2.6.0
pytz: 2017.2
blosc: None
bottleneck: None
tables: None
numexpr: None
feather: None
matplotlib: 2.0.0
openpyxl: None
xlrd: None
xlwt: None
xlsxwriter: None
lxml: None
bs4: None
html5lib: 0.999999999
sqlalchemy: None
pymysql: None
psycopg2: None
jinja2: 2.9.6
s3fs: None
pandas_gbq: None
pandas_datareader: None

[chris-b1]

Thanks for the report! It looks like the in the first case the internal data structures are getting into an invalid state.

In [106]: s1 = df.resample("H").mean()["a"]

In [107]: s1._data.blocks[0]
Out[107]: IntBlock: 73 dtype: float64

In [108]: s2 = df.resample("H")["a"].mean()

In [109]: s2._data.blocks[0]
Out[109]: FloatBlock: 73 dtype: float64

[TomAugspurger]

@jreback I think this blocker for 0.20.2? I can take a look tomorrow if you don't have time.

[jreback]

hmm did this work on 0.19.2?

[TomAugspurger]

I think so

In [6]: pd.__version__
Out[6]: '0.19.2'

In [7]: df = {"a": [1,3,1,4]}
   ...: df = pd.DataFrame(df, index=pd.date_range("2017-01-01", "2017-01-04"))
   ...:

In [8]: df.resample("H").mean()["a"]._data.blocks[0]
Out[8]: FloatBlock: 73 dtype: float64