borrowed data escapes outside of function error for Vec<&str>

[paladin158]

I am testing out groupby, agg and apply custom functions feature with Polars, Polars-PyO3, and PyO3. However, the simple example as shown below won't compile (seems to be borrow lifetime related according to the error message).

So, I am curious to know how am I supposed to fix such an error to make the following code compile?

use polars::prelude::*;
use pyo3::prelude::*;
use pyo3_polars::error::PyPolarsErr;
use pyo3_polars::PyDataFrame;

fn helper(s: Series, cols: Vec<&str>) -> Result<Option<Series>, PolarsError> {
    Ok(Some(Series::new("out", cols)))
}

#[pyfunction]
fn test_groupby(
    pydf: PyDataFrame,
    cols1: Vec<&str>,
    cols2: Vec<&str>,
    group_cols: Vec<&str>,
) -> PyResult<PyDataFrame> {
    let df: DataFrame = pydf.into();
    let res = df
        .lazy()
        .groupby(group_cols)
        .agg(
            cols1
                .iter()
                .map(|&col1| {
                    let cols22 = cols2.clone();
                    col(col1).apply(move |s| helper(s, cols22.clone()), GetOutput::default())
                })
                .collect::<Vec<_>>(),
        )
        .collect()
        .map_err(PyPolarsErr::from)?;
    Ok(PyDataFrame(res))
}

#[pymodule]
fn test_module(_py: Python, m: &PyModule) -> PyResult<()> {
    m.add_function(wrap_pyfunction!(test_groupby, m)?)?;
    Ok(())
}

Compiler error message:

error[E0521]: borrowed data escapes outside of function                                                                                                                         
   --> src/lib.rs:437:21
    |
425 |     cols2: Vec<&str>,
    |     -----      - let's call the lifetime of this reference `'1`
    |     |
    |     `cols2` is a reference that is only valid in the function body
...
437 |                     col(col1).apply(move |s| helper(s, cols22.clone()), GetOutput::default())
    |                     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    |                     |
    |                     `cols2` escapes the function body here
    |                     argument requires that `'1` must outlive `'static`

As another attempt, when I add 'static lifetime annotations on the function signature, I got another compiler error.

New function signature:

#[pyfunction]
fn test_groupby(
    pydf: PyDataFrame,
    cols1: Vec<&'static str>,
    cols2: Vec<&'static str>,
    group_cols: Vec<&str>,
) -> PyResult<PyDataFrame>

Compile error message:

error: lifetime may not live long enough 259/260: test_module                                                                                                                         
   --> src/lib.rs:358:1
    |
358 | #[pyfunction]
    | ^^^^^^^^^^^^^
    | |
    | lifetime `'py` defined here
    | cast requires that `'py` must outlive `'static`
    |
    = note: this error originates in the attribute macro `pyfunction` (in Nightly builds, run with -Z macro-backtrace for more info)

[adamreichold]

Since the lazy evaluation seems to require owned data, the most straight forward solution would be to use Vec<String> instead of Vec<&str>.