`fast_array_utils`¶

Fast array utilities with minimal dependencies.

fast_array_utils.conv: This submodule is always available and contains conversion utilities.
fast_array_utils.stats: This submodule requires numba to be installed and contains statistics utilities.
fast_array_utils.typing and fast_array_utils.types: These submodules contain types for annotations and checks, respectively. Stubs for these types are available even if the respective packages are not installed.

See also the test utilities in testing.fast_array_utils.

usage¶

fast-array-utils supports the following array types:

numpy.ndarray
scipy.sparse.cs{rc}_{array,matrix}
cupy.ndarray and cupyx.scipy.sparse.cs{rc}_matrix
dask.array.Array
h5py.Dataset and zarr.Array
anndata.abc.CS{CR}Dataset (only supported by .conv.to_dense at the moment)

Use fast_array_utils.conv.to_dense to densify arrays and optionally move them to CPU memory:

from fast_array_utils.conv import to_dense

numpy_arr = to_dense(sparse_arr_or_mat)
numpy_arr = to_dense(dask_or_cuda_arr, to_cpu_memory=True)
dense_dask_arr = to_dense(dask_arr)
dense_cupy_arr = to_dense(sparse_cupy_mat)

Use fast_array_utils.conv.* to calculate statistics across one or both axes of a 2D array. All of them support an axis and dtype parameter:

from fast_array_utils import stats

all_equal = stats.is_constant(arr_2d)
col_sums = stats.sum(arr_2d, axis=0)
mean = stats.mean(arr_2d)
row_means, row_vars = stats.mean_var(arr_2d, axis=1)

installation¶

To use fast_array_utils.stats or fast_array_utils.conv:

(uv) pip install 'fast-array-utils[accel]'

To use testing.fast_array_utils:

(uv) pip install 'fast-array-utils[testing]'

fast_array_utils¶

usage¶

installation¶

`fast_array_utils`¶