BUG: regression in read_csv reading from file stream with encoding

[coroa]

Pandas version checks

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

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

• I have confirmed this bug exists on the main branch of pandas.

Reproducible Example

import pandas as pd
import tempfile

with tempfile.NamedTemporaryFile("wt", encoding="utf-8", delete=False) as f:
    f.write("### DATA\nkey,value\ntables,rectangular\n")

with open(f.name, 'r', encoding="utf-8") as fin:
    fin.readline()
    pd.read_csv(fin)

Issue Description

Fails in TextReader.__cinit__ with EmptyDataError in pandas 1.5.0, while it completed correctly in 1.4.4. Using the system default encoding of "UTF-8" (uppercase) completes correctly as well (as does omitting the encoding argument in open).

File ~/.local/conda/envs/datatoolbox_test/lib/python3.8/site-packages/pandas/io/parsers/c_parser_wrapper.py:92, in CParserWrapper.__init__(self, src, **kwds)
     89     kwds.pop(key, None)
     91 kwds["dtype"] = ensure_dtype_objs(kwds.get("dtype", None))
---> 92 self._reader = parsers.TextReader(src, **kwds)
     94 self.unnamed_cols = self._reader.unnamed_cols
     96 # error: Cannot determine type of 'names'

File ~/.local/conda/envs/datatoolbox_test/lib/python3.8/site-packages/pandas/_libs/parsers.pyx:554, in pandas._libs.parsers.TextReader.__cinit__()

EmptyDataError: No columns to parse from file

Expected Behavior

No raise. pd.read_csv reads in table.

Installed Versions

INSTALLED VERSIONS

commit : 87cfe4e
python : 3.8.13.final.0
python-bits : 64
OS : Darwin
OS-release : 21.6.0
Version : Darwin Kernel Version 21.6.0: Sat Jun 18 17:07:25 PDT 2022; root:xnu-8020.140.41~1/RELEASE_X86_64
machine : x86_64
processor : i386
byteorder : little
LC_ALL : None
LANG : en_GB.UTF-8
LOCALE : en_GB.UTF-8

pandas : 1.5.0
numpy : 1.23.3
pytz : 2022.2.1
dateutil : 2.8.2
setuptools : 65.3.0
pip : 22.2.2
Cython : None
pytest : 7.1.3
hypothesis : None
sphinx : None
blosc : None
feather : None
xlsxwriter : 3.0.3
lxml.etree : None
html5lib : None
pymysql : None
psycopg2 : None
jinja2 : 3.1.2
IPython : 8.5.0
pandas_datareader: None
bs4 : None
bottleneck : None
brotli : None
fastparquet : None
fsspec : None
gcsfs : None
matplotlib : 3.6.0
numba : None
numexpr : None
odfpy : None
openpyxl : 3.0.10
pandas_gbq : None
pyarrow : None
pyreadstat : None
pyxlsb : None
s3fs : None
scipy : 1.9.1
snappy : None
sqlalchemy : None
tables : None
tabulate : 0.8.10
xarray : 2022.6.0
xlrd : 2.0.1
xlwt : 1.3.0
zstandard : None
tzdata : None

[twoertwein]

I can confirm that it works on 1.4.4 (engine="c" and "python") and fails on 1.5.0 for engine="c" but works with engine="python".

I must admit I slightly prefer the failure with the c-engine in 1.5.0. After .readline() the position of the file handle reached the end and should therefore correctly be empty. If you add fin.seek(0) before the read_csv call, it works.
Forget my comment - you intentionally read only one line - not the entire file.

[phofl]

Can confirm

commit 9292530c9eadd70fbff68a3e71c2fc2bea10b3d9
Author: Torsten Wörtwein <twoertwein@users.noreply.github.com>
Date:   Mon May 30 16:45:01 2022 -0400

    CLN: Simplify mmap code (#47175)

cc @twoertwein #47175

[twoertwein]

I assume that this shortcut (to make read_csv+mmap faster) is causing an issue:

pandas/pandas/io/parsers/c_parser_wrapper.py

Line 79 in 5550fd9

src = src.buffer # type: ignore[assignment]

In this example

(Pdb) src.tell()
9
(Pdb) src.buffer.tell()
38

TextIOWrapper seems to have read-ahead (since its internal buffer is already at the end).

The following seems to fix this issue src.seek(src.tell()) but I don't know whether that is a good solution.