Support for Keyword Arguments in from_csv Function with use_pyarrow=True for Enhanced CSV Parsing Flexibility

[Max0u]

Description

Summary
The current implementation of from_csv function does not allow kwargs. This feature request proposes enabling the use of kwargs in the context of use_pyarrow=True to enhance functionality and user flexibility, particularly for handling non-standard CSV files with custom delimiters or escape characters.

Motivation
In practice, CSV files do not always conform to a strict standard; many third-party CSV files use non-standard delimiters or escape characters. While using the polar library, I encountered a limitation where I could not adjust the CSV delimiter or escape character in the from_csv function when using PyArrow for parsing. This restriction forced me to use pandas as reader and transfer to polar which cause memory bump.

I would gladly make the PR if this is accepted

[alexander-beedie]

I think if you need to use pyarrow parsing with all of its possible kwargs, you will be better off calling it directly; we want to eventually drop this pass-through completely. (Being able to lean on pyarrow was meant as a temporary convenience for the early days when our CSV parser was significantly less developed than it is now).

But! Could you detail more specifically which features the Polars CSV reader wasn't able to handle in your file? If it's something we can address natively then you wouldn't need "use_pyarrow" at all 👍