search

cucapra / pollen

generating hardware accelerators for pangenomic graph queries
MIT License
24 stars 1 forks source link

Cigar overlaps in paths #80

Closed anshumanmohan closed 1 year ago

anshumanmohan commented 1 year ago

At some point I removed our CIGAR-string parsing just started putting *. The CIGAR string is optional, and a None is rendered as a * in GFA files, so this was relatively okay. However, I worried that there may be cases where odgi was doing something careful with CIGAR strings and we were not.

This PR brings CIGAR-string overlaps back into play. In the algorithms slow-odgi covers thus far, odgi never does anything clever with these strings. It carries them around when the path is unchanged, and drops them on the ground when the paths have changed. We now mimic this behavior in slow-odgi.

anshumanmohan commented 1 year ago

Whoops, sorry, the previous commit about GFAs instead of OGs when running the oracle of validate totally belongs in the basic-bench branch. I was overly hasty in merging that in, and then overly hasty in sticking the change in here 🙈

This is a band-aid solution; the issue is documented in #81