Merging on columns results in loss of categorical dtypes (cast to "object")

[laufere]

Code Sample, a copy-pastable example if possible

import pandas as pd
import numpy as np
a = pd.DataFrame({'numbers': np.arange(10), 'categories': list('abcabcabcd')})
a['categories'] = a['categories'].astype('category')
b = pd.DataFrame({'numbers': np.arange(10)})

print a.dtypes
print a.categories.cat.categories
print

merged = pd.merge(a, b, left_index=True, right_index=True)

print merged.dtypes
print merged['categories'].cat.categories
print 'Merge ok!'
print

merged = pd.merge(a, b, on=['numbers'], how='left')
print merged.dtypes
try:
    print merged['categories'].cat.categories #crashes
except:
    print 'Merge not ok!'
print

Expected Output

The try block should print "categories" 's categories the same way as above, with:
Index([u'a', u'b', u'c', u'd'], dtype='object')

However, the data type is replaced to object/string.

This is not fixed by the v0.18.2 release, which fixes some of the merge issues where int's would get casted to floats when merging.

output of pd.show_versions()

INSTALLED VERSIONS

commit: None
python: 2.7.11.final.0
python-bits: 64
OS: Linux
OS-release: 2.6.32-431.23.3.el6.x86_64
machine: x86_64
processor: x86_64
byteorder: little
LC_ALL: None
LANG: en_US.UTF-8

pandas: 0.17.1
nose: 1.3.7
pip: 7.1.2
setuptools: 20.3.1
Cython: 0.23.4
numpy: 1.10.4
scipy: 0.16.0
statsmodels: 0.6.1
IPython: 4.0.1
sphinx: 1.3.1
patsy: 0.4.0
dateutil: 2.4.2
pytz: 2015.7
blosc: None
bottleneck: 1.0.0
tables: 3.2.2
numexpr: 2.4.4
matplotlib: 1.5.0
openpyxl: 2.2.6
xlrd: 0.9.4
xlwt: 1.0.0
xlsxwriter: 0.7.7
lxml: 3.4.4
bs4: 4.4.1
html5lib: None
httplib2: None
apiclient: None
sqlalchemy: 1.0.9
pymysql: None
psycopg2: None
Jinja2: None

[laufere]

@chrish42

[jreback]

duplicate of #10409

which is still open (and 0.18.2 is not released as of yet)