Serializing float columns with `format="json"` turns inf/nan values into null

[stinodego]

Checks

• [X] I have checked that this issue has not already been reported.
• [X] I have confirmed this bug exists on the latest version of Polars.

Reproducible example

import io

import polars as pl

df = pl.DataFrame({"a": [1.0, float("inf"), float("-inf"), float("nan")]})
ser = df.serialize()
print(ser)  # {"columns":[{"name":"a","datatype":"Float64","bit_settings":"","values":[1.0,null,null,null]}]}
result = pl.DataFrame.deserialize(io.StringIO(ser))
print(result)

Log output

shape: (4, 1)
┌──────┐
│ a    │
│ ---  │
│ f64  │
╞══════╡
│ 1.0  │
│ null │
│ null │
│ null │
└──────┘

Issue description

Roundtripping float columns doesn't work for nan/inf values.

Expected behavior

original values are preserved

Installed versions

main

[ritchie46]

Ah yes. That's because JSON doesn't support those values.