readpaf is a fast parser for minimap2 PAF (Pairwise mApping Format) files. It is written in pure python with no required dependencies unless a pandas DataFrame is required.

Minimal install:

pip install readpaf

With optional pandas dependency:

pip install readpaf[pandas]

https://raw.githubusercontent.com/alexomics/read-paf/main/readpaf.py

https://raw.githubusercontent.com/alexomics/read-paf/v0.0.5/readpaf.py

readpaf only has one user function, parse_paf that accepts of file-like object; this is any object in python that has a file-oriented API (sys.stdin, stdout from subprocess, io.StringIO, open files from gzip or open).

The following script demonstrates how minimap2 output can be piped into readpaf

from readpaf import parse_paf
from sys import stdin

for record in parse_paf(stdin):
    print(record.query_name, record.target_name)

readpaf can also generate a pandas DataFrame:

from readpaf import parse_paf

with open("test.paf", "r") as handle:
    df = parse_paf(handle, dataframe=True)

readpaf has a single user function

parse_paf(file_like=file_handle, fields=list, na_values=list, na_rep=numeric, dataframe=bool)

Parameters:

• file_like: A file like object, such as sys.stdin, a file handle from open or io.StringIO objects
• fields: A list of 13 field names to use for the PAF file, default:

  "query_name", "query_length", "query_start", "query_end", "strand",
  "target_name", "target_length", "target_start", "target_end",
  "residue_matches", "alignment_block_length", "mapping_quality", "tags"

  These are based on the PAF specification.
• na_values: A list of values to interpret as NaN. This is only applied to numeric fields, default: ["*"]
• na_rep: Value to use when a NaN value specified in na_values is found. This should ideally be 0 to match minimap2's output default: 0
• dataframe: bool, if True, return a pandas.DataFrame with the tags expanded into separate Series

If used as an iterator, then each object returned is a named tuple representing a single line in the PAF file. Each named tuple has field names as specified by the fields parameter. The SAM-like tags are converted into their specified types and stored in a dictionary with the tag name as the key and the value a named tuple with fields name, type, and value. When print or str are called on PAF record (named tuple) a formated PAF string is returned, which is useful for writing records to a file. The PAF record also has a method blast_identity which calculates the blast identity for that record.

If used to generate a pandas DataFrame, then each row represents a line in the PAF file and the SAM-like tags are expanded into individual series.