Create an indexed tabix file from a Pandas dataframe via "pure" Python

pandas_df_to_tabix_via_libraries.py

#!/usr/bin/env python

'''
Create an indexed tabix file from a Pandas dataframe 
via "pure" Python (i.e., no subprocess)
''' 

import os
import io
import pandas as pd
import pysam
import bgzip

ds = io.StringIO('''chr1	842320	842327
chr1	842328	842330
chr1	842328	842330
chr1	855426	855427
chr1	855739	855740''')

df = pd.read_csv(ds, delimiter='\t', header=None)
df.columns = ['chrom', 'start', 'stop']

out_bgz_fn = "test_pd.bed.gz"
with open(out_bgz_fn, "wb") as out_bgz:
    with bgzip.BGZipWriter(out_bgz) as out_bgz_fh:
        for index, row in df.iterrows():
            out_line = '{}\t{}\t{}\n'.format(row['chrom'], row['start'], row['stop'])
            out_bgz_fh.write(out_line.encode())

if not os.path.exists(out_bgz_fn):
    raise Exception("Error: Could not create bgzip archive")

out_index_fn = "{}.tbi".format(out_bgz_fn)
if not os.path.exists(out_index_fn):
    pysam.tabix_index(out_bgz_fn, preset="bed")
    if not os.path.exists(out_index_fn):
        raise Exception("Error: Could not create index of bgzip archive")

How does this compare (speed wise) to df.to_csv(..., compression="gzip")?

Obviously I know that a gzip'ed file can't be tabix indexed...

wow... interaction of bgzip and pandas is weird...I find that just importing bgzip means to_csv automatically writes f"{filename}.gz in bgzip format... like it's hooking into it somehow...