delocalizer / mumbai Goto Github PK

Currently we use pysam to read records from bams because it provides nice abstractions but it's not ideal — in particular AlignmentFile — because:

1. we have to supply the constructor with a fake index
2. there's no way to force it to not read the header, which wastes a disk seek given we already know everything we need to know about reference names etc. in order to reconstitute AlignedSegment instances

Currently we store only ref contig names but that's a poor check for identity of a reference genome; we should store and check LN as well.

pileup mode returns pileup for just one bam

looking at the code it's obvious why: