tools/sort.py

import json, sys, glob, lzma

input_files = sys.argv[1]
output_file = sys.argv[2]

print(">>> Inputs", input_files)
if "_" in input_files:
	input_files = glob.glob(input_files.replace("_", "*"))
else:
	input_files = [ input_files ]
print(">>> Inputs", input_files)


def process(input_file):
	print(">>> read file", input_file)
	if "jsonl.xz" in input_file: fin = lzma.open(input_file, "rt")
	else:                        fin = open(input_file, "rt")

	line_txtlen_pairs = []
	for line in fin:
		line_txtlen_pairs.append((line, len(line)))
	fin.close()
	return line_txtlen_pairs


line_txtlen_pairs = []
from multiprocessing.pool import Pool
with Pool() as pool:
    for pairs in pool.imap_unordered(process, input_files):
    	line_txtlen_pairs += pairs

print("=> total docs:", len(line_txtlen_pairs))
print("sorting docs by len ...")

total_txtlen = 0
line_txtlen_pairs.sort(key=lambda x: x[1])
with open(output_file, "wt") as fout:
	for text, n in line_txtlen_pairs:
		if n < 1500: continue
		fout.write(text)
		total_txtlen += n
		print(n, end=" ")

print(f"\nmin len {line_txtlen_pairs[0][1]}, max len {line_txtlen_pairs[-1][1]}")
print(f"avg len {total_txtlen / len(line_txtlen_pairs)}")