Pandas (v0.22) fails to roundtrip category types.

bug_categories.py

import pandas as pd

# Create a DataFrame df_orig and write to Parquet

size = 1000
df_orig = pd.DataFrame(
    data={
        'a': size*['foo', 'bar', 'baz'],
        'b': size*['foo', 'bar', 'baz'],
        'c': size*['foo', 'bar', 'baz'],
    })
print(df_orig.head(10),'\n')
df_orig.to_parquet(fname='df_orig.prq', engine='fastparquet')

# Coerce df_orig to a categorical variables: df_cat

df_cat0 = df_orig.copy()
for col in df_cat0.columns:
    df_cat0[col] = df_cat0[col].astype('category')

# Roundtripping through Parquet

df_cat0.to_parquet(fname='df_cat0.prq', engine='fastparquet')
df_cat1 = pd.read_parquet(path='df_cat0.prq')

# Examine read Parquet DataFrame

print('df_orig - original all dtype object')
df_orig.info(memory_usage='deep')
print()

print('df_cat0 - all dtype categories')
df_cat0.info(memory_usage='deep')
print()

print('df_cat1 - df_cat0 read from disk - all dtype object')
df_cat1.info(memory_usage='deep')
print()

Script output

# python3 bug_categories.py

     a    b    c
0  foo  foo  foo
1  bar  bar  bar
2  baz  baz  baz
3  foo  foo  foo
4  bar  bar  bar
5  baz  baz  baz
6  foo  foo  foo
7  bar  bar  bar
8  baz  baz  baz
9  foo  foo  foo

df_orig - original all dtype object
<class 'pandas.core.frame.DataFrame'>
RangeIndex: 3000 entries, 0 to 2999
Data columns (total 3 columns):
a    3000 non-null object
b    3000 non-null object
c    3000 non-null object
dtypes: object(3)
memory usage: 527.4 KB

df_cat0 - all dtype categories
<class 'pandas.core.frame.DataFrame'>
RangeIndex: 3000 entries, 0 to 2999
Data columns (total 3 columns):
a    3000 non-null category
b    3000 non-null category
c    3000 non-null category
dtypes: category(3)
memory usage: 9.6 KB

df_cat1 - df_cat0 read from disk - all dtype object
<class 'pandas.core.frame.DataFrame'>
RangeIndex: 3000 entries, 0 to 2999
Data columns (total 3 columns):
a    3000 non-null object
b    3000 non-null object
c    3000 non-null object
dtypes: object(3)
memory usage: 527.4 KB

Shell file sizes

-rw-r--r--   1 mookerji  staff   1.8K May  7 09:56 df_cat0.prq
-rw-r--r--   1 mookerji  staff   3.8K May  7 09:56 df_orig.prq

versions

>>> import pandas as pd
>>> pd.show_versions()

INSTALLED VERSIONS
------------------
commit: None
python: 3.6.5.final.0
python-bits: 64
OS: Darwin
OS-release: 17.5.0
machine: x86_64
processor: i386
byteorder: little
LC_ALL: None
LANG: en_US.UTF-8
LOCALE: en_US.UTF-8

pandas: 0.22.0
pytest: 3.5.0
pip: 9.0.3
setuptools: 39.0.1
Cython: None
numpy: 1.14.2
scipy: 1.1.0rc1
pyarrow: 0.7.1
xarray: None
IPython: 6.3.1
sphinx: None
patsy: None
dateutil: 2.6.1
pytz: 2018.4
blosc: None
bottleneck: None
tables: 3.4.2
numexpr: 2.6.4
feather: None
matplotlib: 2.1.2
openpyxl: None
xlrd: None
xlwt: None
xlsxwriter: None
lxml: None
bs4: None
html5lib: 1.0.1
sqlalchemy: 1.2.7
pymysql: None
psycopg2: 2.7.4 (dt dec pq3 ext lo64)
jinja2: 2.8.1
s3fs: 0.1.0
fastparquet: 0.1.5
pandas_gbq: None
pandas_datareader: None