.loc and .iloc returns different dtypes on empty dataframe

[artemyk]

.iloc returns object dtype for a non-object empty series:

In [1]: import pandas as pd
In [2]: df = pd.DataFrame({'a':[1,2,3],'b':['b','b2','b3']})

In [3]: df.ix[[],:].loc[:,'a']
Out[3]: Series([], name: a, dtype: int64)

In [4]: df.ix[[],:].iloc[:,0]
Out[4]: Series([], name: a, dtype: object)

(came across this as I think it is related to the issue reported http://stackoverflow.com/questions/29749356/python-pandas-export-structure-only-no-rows-of-a-dataframe-to-sql/ , i.e. in sql code which is trying to infer series datatypes)

 >>> pd.show_versions()

INSTALLED VERSIONS
------------------
commit: None
python: 2.7.9.final.0
python-bits: 64
OS: Darwin
OS-release: 14.1.0
machine: x86_64
processor: i386
byteorder: little
LC_ALL: None
LANG: en_US.UTF-8

pandas: 0.15.2
nose: 1.3.6
Cython: 0.21.1
numpy: 1.9.2
scipy: 0.14.0
statsmodels: 0.6.1
IPython: 3.1.0
sphinx: 1.2.3
patsy: 0.2.1
dateutil: 2.4.2
pytz: 2015.2
bottleneck: None
tables: None
numexpr: None
matplotlib: 1.4.3
openpyxl: None
xlrd: None
xlwt: None
xlsxwriter: None
lxml: 3.4.2
bs4: 4.3.2
html5lib: None
httplib2: 0.9
apiclient: None
rpy2: None
sqlalchemy: 0.9.8
pymysql: 0.6.2.None
psycopg2: 2.5.4 (dt dec pq3 ext)

[sinhrks]

Thanks for the report. Though it looks like a bug, is there any reason once you convert it to empty DataFrame? All the followings work properly.

import pandas as pd
df = pd.DataFrame({'a':[1,2,3],'b':['b','b2','b3']})
df.loc[[], 'a']
# Series([], Name: a, dtype: int64)
df.ix[[], 'a']
# Series([], Name: a, dtype: int64)
df.iloc[[], 0]
# Series([], Name: a, dtype: int64)
df.ix[[], 0]
# Series([], Name: a, dtype: int64)

[artemyk]

@sinhrks It's a bit involved, but basically there as a function (to write to a table to a sql database) that takes in a dataframe and infers types from the columns. Sometimes people would like to pass in an empty dataframe in order to create an empty table.