BUG: read_csv with pyarrow engine cannot handle single-line CSV files

[jorisvandenbossche]

As long as the file contains a newline, it works fine:

>>> import io
... import pandas as pd
... pd.read_csv(
...     io.StringIO("1,2,3\n"),
...     names=["col1", "col2", "col3"],
...     engine="pyarrow",
... )
   col1  col2  col3
0     1     2     3

But reading an actual one-line file raises inside pyarrow:

>>> pd.read_csv(
...     io.StringIO("1,2,3"),
...     names=["col1", "col2", "col3"],
...     engine="pyarrow",
... )
---------------------------------------------------------------------------
ArrowInvalid     
...
ParserError: CSV parse error: Empty CSV file or block: cannot infer number of columns

While the default c or python engine handle this fine.
And if the header is in the file, it also works fine.

Tested with current latest versions on Ubuntu:

>>> pd.__version__
'3.0.0.dev0+2236.g3c4586fde9'
>>> pa.__version__
'21.0.0'

[khemkaran10]

@jorisvandenbossche I think the issue is that we are not setting column_names here, adding this arg should fix the issue. if it seems reasonable, I am happy to open a PR:

pandas/pandas/io/parsers/arrow_parser_wrapper.py

Lines 142 to 148 in bbcf86f

	self.read_options = {
	"autogenerate_column_names": self.header is None,
	"skip_rows": self.header
	if self.header is not None
	else self.kwds["skiprows"],
	"encoding": self.encoding,
	}

    self.read_options = {
            "column_names": self.kwds["names"],
            "autogenerate_column_names": self.header is None and self.kwds["names"] is None,
            "skip_rows": self.header
            if self.header is not None
            else self.kwds["skiprows"],
            "encoding": self.encoding,
        }

[KevsterAmp]

@khemkaran10 Seems like that fixes the issue. I suggest you open a PR to have the github CI tests verify this change

[rahulratho15]

take