How to cast string column in Polars to numeric coercing errors

Question

I'm working with Polars (version 0.19.0) in Python and facing an issue with converting a string column to numeric while coercing errors. My goal is to convert strings to integers, where non-numeric strings are replaced with nulls.

Here's an example of my DataFrame:

import polars as pl

df = pl.DataFrame({
    "str_column": ["3", "4", "unavailable", "random_String", "17"]
})

I want to convert the str_column to integers, expecting the following output: | str_column | |------------| | 3 | | 4 | | null | | null | | 17 |

I tried using to_integer(strict=False) like this:

df.with_columns(pl.col("str_column").str.parse_int(strict=False))

However, this converts everything to null instead of the expected output.

How can I achieve the desired conversion in Polars (preferably using native polars operations)?

Answer 1

like this

In [4]: import polars as pl
   ...:
   ...: df = pl.DataFrame({
   ...:     "str_column": ["3", "4", "unavailable", "random_String", "17"]
   ...: })

In [5]: df.with_columns(parsed=pl.col('str_column').cast(pl.Int64, strict=False))
Out[5]:
shape: (5, 2)
┌───────────────┬────────┐
│ str_column    ┆ parsed │
│ ---           ┆ ---    │
│ str           ┆ i64    │
╞═══════════════╪════════╡
│ 3             ┆ 3      │
│ 4             ┆ 4      │
│ unavailable   ┆ null   │
│ random_String ┆ null   │
│ 17            ┆ 17     │
└───────────────┴────────┘