anndata.abc.CSRDataset

class anndata.abc.CSRDataset

Bases: ABC

On disk CSR sparse matrix.

Analogous to h5py.Dataset or zarr.Array, but for sparse matrices.

Attributes

backend

Which file type is used on-disk.

dtype

The numpy.dtype of the data attribute of the sparse matrix.

format: ClassVar[Literal['csr', 'csc']] = 'csr'

The format of the sparse matrix.

group

The group underlying the backed matrix.

shape

Shape of the matrix.

Methods

abstractmethod __getitem__(index)

Load a slice or an element from the sparse dataset into memory.

Parameters:

index anndata.typing.Index

Index to load.

Return type:

float | csr_matrix | csc_matrix | csr_array | csc_array

Returns:

The desired data read off disk.

abstractmethod append(sparse_matrix)

Append an in-memory or on-disk sparse matrix to the current object’s store.

Parameters:

sparse_matrix csr_matrix | csc_matrix | csr_array | csc_array | CSRDataset | CSCDataset

The matrix to append.

Raises:

• NotImplementedError – If the matrix to append is not one of csr_array, csc_array, csr_matrix, or csc_matrix.

• ValueError – If both the on-disk and to-append matrices are not of the same format i.e., csr or csc.

• OverflowError – If the underlying data store has a 32 bit indptr, and the new matrix is too large to fit in it i.e., would cause a 64 bit indptr to be written.

• AssertionError – If the on-disk data does not have csc or csr format.

Return type:

None

abstractmethod to_memory()

Load the sparse dataset into memory.

Return type:

csr_matrix | csc_matrix | csr_array | csc_array

Returns:

The in-memory representation of the sparse dataset.