parse_gff3.py

import attr
import typing
from collections import defaultdict

@attr.s(auto_attribs=True)
class GFF3Record:
    seq_id : str = ""
    source : str = ""
    type_ : str = ""
    start : int = 0
    end : int = 0
    score : str = "."
    strand : str = "."
    attributes : typing.Dict[str, str] = attr.Factory(dict)
    id_ : str = ""
    parent : "GFF3Record" = None
    childs : typing.List["GFF3Record"] = attr.Factory(list)

def parse_attributes(attrs_str):
    attrs = dict()
    for it in attrs_str.split(";"):
        k, v = it.split('=')    
        attrs[k] = v
    return attrs

def parse_gff3(path):
    records = dict()
    collection_by_type = defaultdict(list)
    
    with open(path) as f:
        for line in f:
            if line.startswith('#'):
                continue
            itms = line.strip().split("\t")
            attrs = parse_attributes(itms[-1])
            rec = GFF3Record(*itms[:3], start=int(itms[3]), end=int(itms[4]),
                             strand=itms[6], attributes=attrs)
            if 'ID' not in attrs:
                rec.id_ = rec.type_ + ':' + attrs['Parent'].split(':')[1]
            else:
                rec.id_ = attrs['ID']

            if 'Parent' in attrs:
                rec.parent = records[attrs['Parent']]
                rec.parent.childs.append(rec)
            records[rec.id_] = rec
            collection_by_type[rec.type_].append(rec)

    return records, collection_by_type