Module sparse_io_vector 

Modules

scales

Read-time column scales, and the depth factors they come from.

Structs

BackendLocation

Where a single global cell’s nonzeros come from in one of the underlying SparseIo backends. Under ColumnAlignment::Disjoint each global column has exactly one BackendLocation. Under ColumnAlignment::Union a global column can carry one entry per backend that observed the cell.

SparseIoVec

TripletsMatched

Enums

ColumnAlignment

Strategy for aligning column (cell) names across multiple pushed backends.

RowAlignment

Strategy for aligning row names across multiple pushed backends.

Functions

data_name_of

The data_name SparseIoVec::push takes for a backend read from file, one of n_files stacked side by side (ColumnAlignment::Disjoint): the file’s stem (a/pbmc.zarr.zip → pbmc) when there are several, so cells sharing a barcode across files stay apart, and none for a single file. read_data_on_shared_rows and the CLI’s readers name cells by this rule.

stacked_column_name

The name a cell name of file goes by once n_files files are stacked side by side: name@{stem} for several files (data_name_of), name for one.

Type Aliases

ColumnNameCanonicalizer

Optional canonicalizer applied to every backend column (barcode) name during push under ColumnAlignment::Union. Mirror of RowNameCanonicalizer. Two raw barcodes whose canonical forms match are treated as the same biological cell — backends share a single global column id, and read_columns_* merges their nonzeros into one output column. Default = None preserves exact-match behavior.

RowNameCanonicalizer

Optional canonicalizer applied to every backend row name during push. Two raw names that produce the same canonical form are treated as the same row. The displayed row_names_by_global stores the canonical form, so downstream readers see a single consistent label for each row across all backends. Default = None preserves exact-match behavior.