teaguesterling/plinking_duck v0.7.0

on GitHub

Multi-file input (LIST(VARCHAR))

read_pfile() and read_pvar() now accept a list of prefixes/paths in addition to a single one, row-concatenating variants across files — for reading a variant-sharded fileset (many .pgen/.pvar sharing one identical .psam) as a single table.

SELECT * FROM read_pfile(['/data/chr1', '/data/chr2', '/data/chr3']);

• variant and genotype orient row-concatenate across files. region := filters across shards. Verified against whole-file reads (ground-truth equivalence) including all genotype modes and under parallelism.
• All listed files must share the same samples (positional, trust-the-caller); a cross-file sample_ct mismatch is a hard error. Row order is not guaranteed (add ORDER BY).
• Pairs with the scalarfs pathmacro: protocol (a catalog macro returns the shard prefixes).

Not yet supported (clean errors; planned)

• orient := 'sample' with multiple files (Phase 2).
• variants := [...] with a multi-file list — use region := for cross-file selection.
• read_pgen list input (Phase 3).