Spurious casting to float when assigning both integers and floats

[toobaz]

Code Sample, a copy-pastable example if possible

In [2]: df = pd.DataFrame([[1,2],[3,4]], columns=['a', 'b'])

In [3]: df.loc[2] = pd.Series({'a' : 5, 'b' : 6})

In [4]: df.dtypes  # right!
Out[4]: 
a    int64
b    int64
dtype: object

In [5]: df.loc[3] = pd.Series({'a' : 7})

In [6]: df.dtypes
Out[6]: 
a    float64
b    float64
dtype: object

Problem description

Column a should not be casted to float, since at no time it is assumed to contain null values.

Expected Output

In [6]: df.dtypes
Out[6]: 
a    int64
b    float64
dtype: object

Output of pd.show_versions()

INSTALLED VERSIONS

commit: None
python: 3.5.2.final.0
python-bits: 64
OS: Linux
OS-release: 4.7.0-1-amd64
machine: x86_64
processor:
byteorder: little
LC_ALL: None
LANG: it_IT.utf8
LOCALE: it_IT.UTF-8

pandas: 0.19.0+376.gdc40058.dirty
nose: 1.3.7
pip: 8.1.2
setuptools: 28.0.0
Cython: 0.23.4
numpy: 1.12.0
scipy: 0.18.1
statsmodels: 0.8.0.dev0+f80669e
xarray: None
IPython: 5.1.0.dev
sphinx: 1.4.8
patsy: 0.3.0-dev
dateutil: 2.5.3
pytz: 2015.7
blosc: None
bottleneck: 1.2.0.dev
tables: 3.2.2
numexpr: 2.6.0
feather: None
matplotlib: 2.0.0rc2
openpyxl: None
xlrd: 1.0.0
xlwt: 1.1.2
xlsxwriter: 0.9.3
lxml: None
bs4: 4.5.1
html5lib: 0.999
httplib2: 0.9.1
apiclient: 1.5.2
sqlalchemy: 1.0.15
pymysql: None
psycopg2: None
jinja2: 2.8
s3fs: None
pandas_datareader: 0.2.1

[toobaz]

It's actually not related to the input being a Series, and not even to missing values:

In [2]: df = pd.DataFrame([[1,2],[3,4]], columns=['a', 'b'])

In [3]: df.loc[2] = [5,6.0]

In [4]: df
Out[4]: 
     a    b
0  1.0  2.0
1  3.0  4.0
2  5.0  6.0

Retitling accordingly

[jreback]

yeah I suppose, there are lots of possibilities for input so these get put into a numpy array first, which casts them IIRC.

if you want to submit a pr that would be great (w/o breaking anything else of course)!

[toobaz]

By the way (mostly to remind myself when I will have time to look at it): df.loc[2] = [5,"6"] works fine

[jreback]

yes numpy generally will cast ints and floats to floats

but ints and string will be object (preserving the dtypes)

[antopolskiy]

Perhaps a related issue: http://stackoverflow.com/questions/43366763/loc-indexing-changes-type

[jreback]

@antopolskiy

no that is already fixed in master (0.20.0)

xref #11617

In [1]: df_test = pd.DataFrame({'ints':[1,2,3], 'floats': [4.5,5.5,6.5]})
   ...: 

In [2]: df_test['ints'].dtype
Out[2]: dtype('int64')

In [3]: df_test.loc[0,'ints']
Out[3]: 1

In [4]: df_test['ints'].dtype
Out[4]: dtype('int64')

[NicolasDumoulin]

Anyone does have a better workaround that recasting concerned columns with astype after using loc?