read_csv ignores dtype for bool columns with missing values

[finnhacks42]

Code Sample, a copy-pastable example if possible

import pandas as pd
from pandas.compat import StringIO
data = "false,1\n,1\ntrue,"
df = pd.read_csv(StringIO(data), header=None, names=['a','b'], dtype={'a': 'bool'}, engine='c')
print(df['a'].dtype)
print(df)

outputs:

object
a b
0 False 1.0
1 NaN 1.0
2 True NaN

import pandas as pd
from pandas.compat import StringIO
data = "false,1\n,1\ntrue,"
df = pd.read_csv(StringIO(data), header=None, names=['a','b'], dtype={'a': 'bool'},engine='python')
print (df['a'].dtype)
print(df)

outputs:

bool
a b
0 False 1.0
1 True 1.0
2 True NaN

Problem description

Missing values cannot be coerced into a column of dtype bool. The expected behaviour, to be consistent with the analogous case for integers, is to throw a ValueError. The user is then aware of the issue and can specify a converter or parse the boolean as a float (subject to #16698).

Instead the c engine silently ignores the requested dtype and returns a column of type object. This behaviour is especially confusing on large datasets with low_memory=True as you get the warning: Column 1 has mixed types. Specify dtype option on import or set low_memory=False.

The python engine silently converts all missing values to True and returns a column of dtype bool.

Expected Output

ValueError('Boolean column has NA values in column 1')

Output of pd.show_versions()

INSTALLED VERSIONS

commit: None
python: 3.6.5.final.0
python-bits: 64
OS: Darwin
OS-release: 14.5.0
machine: x86_64
processor: i386
byteorder: little
LC_ALL: None
LANG: None
LOCALE: en_US.UTF-8

pandas: 0.23.0.dev0+725.gf67c6fa80
pytest: 3.5.0
pip: 9.0.3
setuptools: 39.0.1
Cython: 0.28.1
numpy: 1.14.2
scipy: 1.0.1
pyarrow: 0.9.0
xarray: 0.10.2
IPython: 6.3.0
sphinx: 1.7.2
patsy: 0.5.0
dateutil: 2.7.2
pytz: 2018.3
blosc: None
bottleneck: 1.2.1
tables: 3.4.2
numexpr: 2.6.4
feather: 0.4.0
matplotlib: 2.2.2
openpyxl: 2.5.1
xlrd: 1.1.0
xlwt: 1.2.0
xlsxwriter: 1.0.2
lxml: 4.2.1
bs4: 4.6.0
html5lib: 1.0.1
sqlalchemy: 1.2.6
pymysql: 0.8.0
psycopg2: None
jinja2: 2.10
s3fs: 0.1.4
fastparquet: 0.1.5
pandas_gbq: None
pandas_datareader: None

[chris-b1]

Yes, I agree this should match the integer behavior, thanks for the report, PR welcome!

df = pd.read_csv(StringIO(data), header=None, names=['a','b'], dtype={'a': 'bool', 'b': 'int64'}, engine='c')

ValueError                                Traceback (most recent call last)
<snip>
pandas/_libs/parsers.pyx in pandas._libs.parsers.TextReader.read()
pandas/_libs/parsers.pyx in pandas._libs.parsers.TextReader._read_low_memory()
pandas/_libs/parsers.pyx in pandas._libs.parsers.TextReader._read_rows()
pandas/_libs/parsers.pyx in pandas._libs.parsers.TextReader._convert_column_data()
pandas/_libs/parsers.pyx in pandas._libs.parsers.TextReader._convert_tokens()
pandas/_libs/parsers.pyx in pandas._libs.parsers.TextReader._convert_with_dtype()
ValueError: Integer column has NA values in column 1

[atulagrwl]

Here is the current behavior

Bool Value	Engine = C	Engine = Python
NA	NaN	True
0	False	False
1	True	True
-1	error	True
100	error	True
abc	error	True

What is expected behavior for above cases? @chris-b1