pairtools parsing engine abstraction

[agalitsyna]

Premise:

• pysam is fast but introduces multiple logistic complications. We may want to replace it with manual bam parser (Anton's style). pysam:

  • is complicated to link in isolated environments
  • leads to pairtools compilation errors without installing cython before pairtools
  • is not supported on osx
  • is not supported for python 3.11
  • New alignment formats emerge that can be parsed by pairtools with minor modifications, e.g. paf (pafpy)

  Proposal:

Introduce an abstraction for parsing engine that can input data from APIs from different tools

Possible technical solutions:

1. Make a universal wrapper class for "alignment" and create an io library for parsing it from data with different engines

2. ?

Complications:

1. Each new engine introduces a new dependency that we want to make optional

[nvictus]

What if the API were based on dataframes, instead of abstract alignment records?