Read from generated csv file in memory without writing to hard disk using `scan_csv`

Question

numpy supports reading from StringIO. I can create a string representation of csv and feed it to numpy.

How to do it in pypolars?

I came to know that I can use read_csv() with lazy(), but I want to use scan_csv to get the benefits of lazy eval. How to go about doing that?

EDIT: Updated question to specify scan_csv.

Answer 1

This now works as of Polars 1.7.0

# bytes

pl.scan_csv(b"a,b,c\n1,2,3").collect()

shape: (1, 3)
┌─────┬─────┬─────┐
│ a   ┆ b   ┆ c   │
│ --- ┆ --- ┆ --- │
│ i64 ┆ i64 ┆ i64 │
╞═════╪═════╪═════╡
│ 1   ┆ 2   ┆ 3   │
└─────┴─────┴─────┘

# io.BytesIO

import io

f = io.BytesIO(b"a,b,c\n1,2,3")

pl.scan_csv(f).collect()

shape: (1, 3)
┌─────┬─────┬─────┐
│ a   ┆ b   ┆ c   │
│ --- ┆ --- ┆ --- │
│ i64 ┆ i64 ┆ i64 │
╞═════╪═════╪═════╡
│ 1   ┆ 2   ┆ 3   │
└─────┴─────┴─────┘