niceSOFT/python3-charset-normalizer: Truly universal encoding detector in pure Python.

bin/performance.py
156 строк6 KB
Ahmed TAHRI
:wrench: silent performance.py exit non-zero
05 мар 2026, 16:32
05 мар 2026, 16:320010f0c
from __future__ import annotations

import argparse
from glob import glob
from math import ceil
from os.path import isdir
from statistics import mean, stdev
from sys import argv
from time import perf_counter_ns

from chardet import detect as chardet_detect

from charset_normalizer import detect


def calc_percentile(data, percentile):
    n = len(data)
    p = n * percentile / 100
    sorted_data = sorted(data)

    return sorted_data[int(p)] if p.is_integer() else sorted_data[int(ceil(p)) - 1]


def performance_compare(arguments):
    parser = argparse.ArgumentParser(
        description="Performance CI/CD check for Charset-Normalizer"
    )

    parser.add_argument(
        "-s",
        "--size-increase",
        action="store",
        default=1,
        type=int,
        dest="size_coeff",
        help="Apply artificial size increase to challenge the detection mechanism further",
    )

    args = parser.parse_args(arguments)

    if not isdir("./char-dataset"):
        print(
            "This script require https://github.com/Ousret/char-dataset to be cloned on package root directory"
        )
        exit(1)

    chardet_results = []
    charset_normalizer_results = []

    file_list = sorted(glob("./char-dataset/**/*.*"))
    total_files = len(file_list)

    for idx, tbt_path in enumerate(file_list):
        with open(tbt_path, "rb") as fp:
            content = fp.read() * args.size_coeff

        before = perf_counter_ns()
        chardet_detect(content)
        chardet_time = round((perf_counter_ns() - before) / 1000000000, 5)
        chardet_results.append(chardet_time)

        before = perf_counter_ns()
        detect(content)
        charset_normalizer_time = round((perf_counter_ns() - before) / 1000000000, 5)
        charset_normalizer_results.append(charset_normalizer_time)

        charset_normalizer_time = charset_normalizer_time or 0.000005
        cn_faster = (chardet_time / charset_normalizer_time) * 100 - 100
        print(
            f"{idx + 1:>3}/{total_files} {tbt_path:<82} C:{chardet_time:.5f}  "
            f"CN:{charset_normalizer_time:.5f}  {cn_faster:.1f} %"
        )

    # Print the top 10 rows with the slowest execution time
    print(
        f"\n{'-' * 102}\nTop 10 rows with the slowest execution time of charset_normalizer:\n"
    )
    sorted_results = sorted(
        enumerate(charset_normalizer_results), key=lambda x: x[1], reverse=True
    )
    for idx, time in sorted_results[:10]:
        tbt_path = file_list[idx]
        print(f"{idx + 1:>3}/{total_files} {tbt_path:<82}  CN:{time:.5f}")

    # Print charset normalizer statistics
    min_time = min(charset_normalizer_results)
    max_time = max(charset_normalizer_results)
    stdev_time = stdev(charset_normalizer_results)
    mean_time = mean(charset_normalizer_results)
    cv = (stdev_time / mean_time) * 100  # Coefficient of variation
    print(f"\n{'-' * 102}\nCharset Normalizer statistics:\n")
    print(f"Minimum Execution Time: {min_time:.5f} seconds")
    print(f"Maximum Execution Time: {max_time:.5f} seconds")
    print(f"Mean Execution Time: {mean_time:.5f} seconds")
    print(f"Standard Deviation: {stdev_time:.5f} seconds")
    print(f"Coefficient of Variation (CV): {cv:.1f} %")

    # Print comparison statistics for chardet and charset normalizer
    chardet_avg_delay = round(mean(chardet_results) * 1000)
    chardet_99p = round(calc_percentile(chardet_results, 99) * 1000)
    chardet_95p = round(calc_percentile(chardet_results, 95) * 1000)
    chardet_50p = round(calc_percentile(chardet_results, 50) * 1000)

    charset_normalizer_avg_delay = round(mean(charset_normalizer_results) * 1000)
    charset_normalizer_99p = round(
        calc_percentile(charset_normalizer_results, 99) * 1000
    )
    charset_normalizer_95p = round(
        calc_percentile(charset_normalizer_results, 95) * 1000
    )
    charset_normalizer_50p = round(
        calc_percentile(charset_normalizer_results, 50) * 1000
    )

    # mypyc can offer performance ~1ms in the 50p. When eq to 0 assume 1 due to imprecise nature of this test.
    if charset_normalizer_50p == 0:
        charset_normalizer_50p = 1

    print(f"\n{'-' * 102}\nCharset Normalizer vs Chardet statistics:\n")

    print("------------------------------")
    print("--> Chardet Conclusions")
    print("   --> Avg: " + str(chardet_avg_delay) + "ms")
    print("   --> 99th: " + str(chardet_99p) + "ms")
    print("   --> 95th: " + str(chardet_95p) + "ms")
    print("   --> 50th: " + str(chardet_50p) + "ms")

    print("------------------------------")
    print("--> Charset-Normalizer Conclusions")
    print("   --> Avg: " + str(charset_normalizer_avg_delay) + "ms")
    print("   --> 99th: " + str(charset_normalizer_99p) + "ms")
    print("   --> 95th: " + str(charset_normalizer_95p) + "ms")
    print("   --> 50th: " + str(charset_normalizer_50p) + "ms")

    # print("------------------------------")
    # print("--> Charset-Normalizer / Chardet: Performance Сomparison")
    # print(
    #     "   --> Avg: x"
    #     + str(round(chardet_avg_delay / charset_normalizer_avg_delay, 2))
    # )
    # print("   --> 99th: x" + str(round(chardet_99p / charset_normalizer_99p, 2)))
    # print("   --> 95th: x" + str(round(chardet_95p / charset_normalizer_95p, 2)))
    # print("   --> 50th: x" + str(round(chardet_50p / charset_normalizer_50p, 2)))

    # return (
    #     0
    #     if chardet_avg_delay > charset_normalizer_avg_delay
    #     and chardet_99p > charset_normalizer_99p
    #     else 1
    # )

    return 0


if __name__ == "__main__":
    exit(performance_compare(argv[1:]))