I am building a simple python extension to process numpy.ndarray objects using rust-numpy crate. I want to save numpy.ndarray objects in static or thread local variables for later process:

use std::cell::RefCell;
use numpy::PyReadwriteArray1;
use pyo3::{Bound, pymodule, PyResult, types::PyModule};

thread_local! {
    static ARRAYS: RefCell<Vec::<PyReadwriteArray1<i32>>> = RefCell::new(Vec::<PyReadwriteArray1<i32>>::new());
}

#[pymodule]
fn rust_ext<'py>(m: &Bound<'py, PyModule>) -> PyResult<()> {
    #[pyfn(m)]
    fn register_array(mut x: PyReadwriteArray1<i32>) {
        x.as_array_mut()[0] = 100;
        ARRAYS.with_borrow_mut(|v| v.push(x));
    }

    Ok(())
}

But I got lifetime compile error:

static ARRAYS: RefCell<Vec::<PyReadwriteArray1<i32>>> = RefCell::new(Vec::<PyReadwriteArray1<i32>>::new());
  |                                                   ^ expected named lifetime parameter

So is it possible to save python objects in extension written in Rust? If so, how to fix ‘py lifetime issue?