Trying to get `CsrMatrix` from `AnnData`

[allan2]

Hi Kai,

I'm trying to read CSR from a .h5ad file in Rust. It works with anndata. I'm using the Git versions from this repo.

I'm doing something like this:

let file = H5::open(path)?;
let adata = AnnData::<H5>::open(file)?;
let slot = adata.get_x();

let mut inner = slot.extract().unwrap();
println!("{}", inner.dtype());
println!("{}", inner.shape());

let data = inner.data::<CsrMatrix<f32>>()?;
println!("{:?}", data);

CsrMatrix(f32)
54220 x 20845
thread 'main' panicked at 'called `Result::unwrap()` on an `Err` value: SparseFormatError { kind: InvalidStructure, error: "Column indices are not monotonically increasing (sorted) within each row." }'

InnerArrayElem requires specifying type D. Is it possible that I'm missing a step before getting the CsrMatrix?

Thanks in advance.

[kaizhang]

Currently, anndata-rs assumes the column indices of a csr matrix are stored in a monotonically increasing order for performance reasons.

The column indices of csr_matrix from scipy and anndata package "are not necessarily sorted. Likewise for CSC row indices. Use the .sorted_indices() and .sort_indices() methods when sorted indices are required (e.g., when passing data to other libraries)."