SAM: PNEXT 0 with RNEXT =

[jkbonfield]

The SAM spec has the following two statements.

If RNEXT is *, no assumptions can be made on PNEXT and bit 0x20.

If PNEXT is 0, no assumptions can be made on RNEXT and bit 0x20.

Firstly it should probably say "FLAG bit 0x20".

Also if RNEXT is = and PNEXT is 0 then logically we're saying we can't make assumptions on RNEXT and due to PNEXT and therefore we can't complain that PNEXT is invalid because of RNEXT not being *. That seems a little circular, but perhaps it's consistent.

I raise it though because this specific case causes htslib to turn PNEXT to * as it says position 0 must mean unmapped. Htsjdk also warns about this case, but at least doesn't modify the data.

Personally I think it's an htslib bug as it has no right to change the data, even if it wishes to warn about it, but thought I'd run it past the other maintainers first.

The only spec nuance here really is is it justified to want to say the mate maps somewhere else on this chromosome, but we don't have the exact coordinate (maybe it was just a crude kmer hit) so we wish to list it as unknown? The spec appears to forbid this, but it's a bit tenuous.