Column loses category when using .loc for a one row dataframe

[nyejon]

If I convert a list of columns to type 'category'

import pandas as pd 

d1 = {'one' : ['a'],
     'two' : ['a']}

d2 = {'one' : ['a', 'b'],
     'two' : ['a', 'b']}


df1 = pd.DataFrame(d1)
df2 = pd.DataFrame(d2)
df1.loc[: , 'one']= df1['one'].astype('category')
df2.loc[: , 'one'] = df2['one'].astype('category')

print('df1')
print(df1)
print(df1.dtypes)
print('df2')
print(df2)
print(df2.dtypes)

df1
  one two
0   a   a
one    category
two      object
dtype: object
df2
  one two
0   a   a
1   b   b
one    category
two      object
dtype: object

df1['one'] = df1['one'].cat.set_categories(df2['one'].cat.categories)

print('Assigning without loc')
print(df1)
print(df1.dtypes)

Assigning without loc
  one two
0   a   a
one    category
two      object
dtype: object

df1.loc[:, 'one'] = df1['one'].cat.set_categories(df2['one'].cat.categories)
print('Assigning with loc')
print(df1)
print(df1.dtypes)

Assigning with loc
  one two
0   a   a
one    object
two    object
dtype: object

df2.loc[:, 'one'] = df2['one'].cat.set_categories(df2['one'].cat.categories)
print('Assigning df2 with loc')
print(df2)
print(df2.dtypes)

Assigning df2 with loc
  one two
0   a   a
1   b   b
one    category
two      object
dtype: object

Problem description

I am trying to convert my defined categorical columns to the category type. It works when the dataframe is longer than one row, but if it is only one row it keeps the datatype as object.

With only one row I get the following column outputs:

df.dtypes

  one two
0   a   a
one    object
two    object
dtype: object

Expected Output

I would expect the column to be type category even for one row.

df.dtypes

  one two
0   a   a
one    category
two    object
dtype: object

Output of pd.show_versions()

INSTALLED VERSIONS ------------------ commit: None python: 3.5.3.final.0 python-bits: 64 OS: Windows OS-release: 7 machine: AMD64 processor: Intel64 Family 6 Model 60 Stepping 3, GenuineIntel byteorder: little LC_ALL: None LANG: None LOCALE: None.None

pandas: 0.20.1
pytest: No

ne
pip: 9.0.1
setuptools: 35.0.2
Cython: None
numpy: 1.12.1
scipy: 0.18.1
xarray: None
IPython: 5.1.0
sphinx: None
patsy: None
dateutil: 2.6.0
pytz: 2017.2
blosc: None
bottleneck: None
tables: None
numexpr: None
feather: None
matplotlib: 2.0.0
openpyxl: 2.4.1
xlrd: 1.0.0
xlwt: None
xlsxwriter: 0.9.6
lxml: None
bs4: None
html5lib: 0.9999999
sqlalchemy: 1.1.5
pymysql: None
psycopg2: None
jinja2: 2.9.4
s3fs: None
pandas_gbq: None
pandas_datareader: None

[TomAugspurger]

Simpler example:

In [78]: df = pd.DataFrame({"A": [1]})

In [79]: df.loc[:, 'B'] = pd.Categorical(['a'])

In [80]: df.dtypes
Out[80]:
A     int64
B    object
dtype: object

[jreback]

There is a coercion step where df.loc[:, column]= will try to do exactly df[column]= so I guess this should work.

[eamag]

Also have this issue

[mroeschke]

Looks to work on master. Could use a test

In [47]: In [78]: df = pd.DataFrame({"A": [1]})
    ...:
    ...: In [79]: df.loc[:, 'B'] = pd.Categorical(['a'])

In [48]: In [80]: df.dtypes
    ...:
Out[48]:
A       int64
B    category
dtype: object

In [49]: pd.__version__
Out[49]: '1.1.0.dev0+1974.g0159cba6e'