BUG: Pandas closes user-provided file handles that it doesn't own.

[tpllaha]

• I have checked that this issue has not already been reported.

• I have confirmed this bug exists on the latest version of pandas.

• (optional) I have confirmed this bug exists on the master branch of pandas.

---

read_csv closes user-provided file handles in specific cases. In particular, if we pass a BytesIO or a file opened in binary mode, and pass an encoding kwarg. If the encoding kwarg is not passed, pandas does not close the file handle (as expected).
Much earlier (in 2016) a related issue has been reported and fixed. The test cases that were mentioned there still pass today, but only in case the file handle is opened in binary mode & the encoding kwarg is passed. (version 1.1.2 & python 3.8.5)

Code Sample, a copy-pastable example

import sys
import pandas
from io import BytesIO, StringIO

print(f'Python version: {sys.version}')
print(f'pandas version: {pandas.__version__}')

string_io = StringIO('a,b\n1,2')
bytes_io_1 = BytesIO(b'a,b\n1,2')
bytes_io_2 = BytesIO(b'a,b\n1,2')

pandas.read_csv(string_io)
print(f'Was StringIO closed? - {string_io.closed}')

pandas.read_csv(bytes_io_1)
print(f'Was BytesIO closed when encoding is NOT passed? - {bytes_io_1.closed}')

pandas.read_csv(bytes_io_2, encoding='utf-8')
print(f'Was BytesIO closed when encoding is passed? - {bytes_io_2.closed}')

prints:

Python version: 3.8.5 (v3.8.5:580fbb018f, Jul 20 2020, 12:11:27) 
[Clang 6.0 (clang-600.0.57)]
pandas version: 1.1.2
Was StringIO closed? - False
Was BytesIO closed when encoding is NOT passed? - False
Was BytesIO closed when encoding is passed? - True
                                              ^

Problem description

Expectation is that pandas should never close user-provided handles. If pandas didn't open the file, then it doesn't own it & therefore shouldn't close it.

Expected Output

I would expect the script above to output:

Python version: 3.8.5 (v3.8.5:580fbb018f, Jul 20 2020, 12:11:27) 
[Clang 6.0 (clang-600.0.57)]
pandas version: 1.1.2
Was StringIO closed? - False
Was BytesIO closed when encoding is NOT passed? - False
Was BytesIO closed when encoding is passed? - False

Output of pd.show_versions()

INSTALLED VERSIONS

commit : d9fff27
python : 3.8.5.final.0
python-bits : 64
OS : Darwin
OS-release : 19.6.0
Version : Darwin Kernel Version 19.6.0: Thu Jun 18 20:49:00 PDT 2020; root:xnu-6153.141.1~1/RELEASE_X86_64
machine : x86_64
processor : i386
byteorder : little
LC_ALL : en_US.UTF-8
LANG : en_US.UTF-8
LOCALE : en_US.UTF-8

pandas : 1.1.0
numpy : 1.17.2
pytz : 2020.1
dateutil : 2.7.3
pip : 20.2.3
setuptools : 40.7.1
Cython : 0.29.7
pytest : 4.6.5
hypothesis : 4.55.2
sphinx : None
blosc : None
feather : None
xlsxwriter : None
lxml.etree : 4.5.2
html5lib : None
pymysql : None
psycopg2 : None
jinja2 : 2.11.2
IPython : 7.18.1
pandas_datareader: None
bs4 : None
bottleneck : None
fsspec : 0.8.2
fastparquet : None
gcsfs : None
matplotlib : None
numexpr : None
odfpy : None
openpyxl : None
pandas_gbq : None
pyarrow : None
pytables : None
pyxlsb : None
s3fs : 0.3.5
scipy : 1.5.2
sqlalchemy : 1.3.10
tables : None
tabulate : None
xarray : None
xlrd : None
xlwt : None
numba : None

[twoertwein]

@tpllaha thank you for the report! I thought master wouldn't have that problem anymore (I was thinking about to_csv), but pandas.read_csv(bytes_io_2, encoding='utf-8') still seems to close that file handle

[twoertwein]

@tpllaha if you need a quick workaround, only the c engine is affected. Until it is fixed, you can use engine='python'

[tpllaha]

@twoertwein thanks. For now, I've worked around it by simply not passing the encoding parameter. It looks like the default encoding is utf-8 and my files are utf-8 encoded, but the bug only surfaces if the encoding is passed explicitly.