Read an AnnData lazily

read_lazy() reads an h5ad file or a zarr store (v3 or v2) into an AnnData() whose X, layers, obsm, varm, obsp, varp, obs, var and raw are lazy (see read_elem_lazy()) and whose uns is read into memory. It matches Python's read_lazy().

Usage

read_lazy(store, load_annotation_index = TRUE)

Arguments

store

An h5ad file or a zarr directory, given as a path or as an open store handle.

load_annotation_index

Whether the obs and var indexes are read into memory. If FALSE the names are Python's dummy range ("0", "1", ...).

Value

An AnnData().

Details

Like Python it only reads the current encodings, so files from old anndata releases (such as an unencoded X, a dataframe version 0.1.0 or raw.X) raise an anndata_error, and uns$neighbors matrices are not moved to obsp. Use read_h5ad() or read_zarr() for those files, which raise anndata_old_format_warning (Python's OldFormatWarning).

Deviations from Python

Arrays inside uns are read into memory (Python keeps them as dask arrays), t() of a lazy AnnData stays lazy (it fails in Python) and lazy frames are indexed by row name where Python's Dataset2D has no loc. See vignette("deviations") for the full list.

Examples

adata <- AnnData(matrix(as.numeric(1:6), nrow = 3))
path <- tempfile(fileext = ".zarr")
write_zarr(adata, path)
read_lazy(path)
#> AnnData object with n_obs × n_vars = 3 × 2
#>     layers: None (.X)