pytc.df.thc.convert_x_to_rank_major¶

pytc.df.thc.convert_x_to_rank_major(src, dst, *, dataset='X', row_block=8)[source]¶

Convert one X dataset to rank-major layout, chunked over an nmo axis.

src/dst are paths or open h5py.File objects. The loop materializes one (nmo, row_block, rank) slab at a time, so peak host memory is nmo * row_block * rank * 8 bytes (~1.6 GiB at the default block). Returns the destination path or file unchanged. The destination dataset is written contiguous (no HDF5 chunking) so panel reads stay single sequential extents.