FazBrowse GitHub Viewer | Trending |
URL:
| Home
Tools: [Download Repo ZIP]   [Original HTTPS Page]

Use `str.translate` to improve performance of `normalize` by hugovk · Pull Request #529 · python/importlib_metadata · GitHub

/ importlib_metadata Public

Use str.translate to improve performance of normalize - #529

Closed
hugovk wants to merge 2 commits into
python:mainfrom
hugovk:speedup-normalize
ConversationCommits (2)ChecksFiles changed
Closed

Use str.translate to improve performance of normalize#529
hugovk wants to merge 2 commits into
python:mainfrom
hugovk:speedup-normalize

Conversation

hugovk commented Jan 13, 2026

Copy link
Copy Markdown
Member

This is the same as python/cpython#143660.

We can apply @henryiii's improvement to packaging in pypa/packaging#1030 (see also https://iscinumpy.dev/post/packaging-faster/) to improve the performance of normalize and make it ~3.7 times faster.

Benchmark

Run Prepared.normalize(n) on every name in PyPI:

# benchmark_names.py
import sqlite3
import timeit
from importlib_metadata import Prepared

# Get data with:
# curl -L https://github.com/pypi-data/pypi-json-data/releases/download/latest/pypi-data.sqlite.gz | gzip -d > pypi-data.sqlite
# Or ues pre-cached files from:
# https://gist.github.com/hugovk/efdbee0620cc64df7b405b52cf0b6e42

CACHE_FILE = "/tmp/bench/names.txt"
DB_FILE = "/tmp/bench/pypi-data.sqlite"

try:
    with open(CACHE_FILE) as f:
        TEST_ALL_NAMES = [line.rstrip("\n") for line in f]
except FileNotFoundError:
    TEST_ALL_NAMES = []
    with sqlite3.connect(DB_FILE) as conn:
        with open(CACHE_FILE, "w") as cache:
            for (name,) in conn.execute("SELECT name FROM projects"):
                if name:
                    TEST_ALL_NAMES.append(name)
                    cache.write(name + "\n")


def bench():
    for n in TEST_ALL_NAMES:
        Prepared.normalize(n)


if __name__ == "__main__":
    print(f"Loaded {len(TEST_ALL_NAMES):,} names")
    t = timeit.timeit("bench()", globals=globals(), number=1)
    print(f"Time: {t:.4f} seconds")

Benchmark data can be found at https://gist.github.com/hugovk/efdbee0620cc64df7b405b52cf0b6e42

Before

❯ python3.14 --version
Python 3.14.2

❯ python3.14 benchmark_names.py
Loaded 8,344,947 names
Time: 4.8933 seconds

After

❯ python3.14 benchmark_names.py
Loaded 8,344,947 names
Time: 1.3266 seconds

3.7 times faster.

hugovk commented Jan 19, 2026 •
edited
Loading

Copy link
Copy Markdown
Member Author

Following on from pypa/packaging#1064, Python 3.12 and 3.13 are slower, even though 3.10, 3.11 and 3.14 are much faster.

Measured with hyperfine on python.org versions on macOS:

Python main (s) PR (s) Result
3.10 6.632 ± 0.515 3.456 ± 0.122 PR 1.92x faster
3.11 6.331 ± 0.541 3.302 ± 0.031 PR 1.92x faster
3.12 5.690 ± 0.183 6.640 ± 0.069 main 1.17x faster
3.13 5.784 ± 0.109 5.957 ± 0.074 main 1.03x faster
3.14 5.749 ± 0.060 2.163 ± 0.125 PR 2.66x faster

Marking as draft for now.

Details 
❯ hyperfine --warmup 1 -r 3 \
    -n main --prepare 'git checkout main' 'python3.10 benchmark_names.py' \
    -n PR --prepare 'git checkout speedup-normalize -q' 'python3.10 benchmark_names.py'
Benchmark 1: main
  Time (mean ± σ):      6.632 s ±  0.515 s    [User: 6.138 s, System: 0.359 s]
  Range (min … max):    6.237 s …  7.214 s    3 runs

Benchmark 2: PR
  Time (mean ± σ):      3.456 s ±  0.122 s    [User: 3.257 s, System: 0.169 s]
  Range (min … max):    3.373 s …  3.596 s    3 runs

Summary
  PR ran
    1.92 ± 0.16 times faster than main

importlib_metadata on  speedup-normalize [?] via 🐍 v3.14.2 via 💎 v3.1.3 took 42s
❯ hyperfine --warmup 1 -r 3 \
    -n main --prepare 'git checkout main' 'python3.11 benchmark_names.py' \
    -n PR --prepare 'git checkout speedup-normalize -q' 'python3.11 benchmark_names.py'
Benchmark 1: main
  Time (mean ± σ):      6.331 s ±  0.541 s    [User: 5.415 s, System: 0.420 s]
  Range (min … max):    5.859 s …  6.921 s    3 runs

Benchmark 2: PR
  Time (mean ± σ):      3.302 s ±  0.031 s    [User: 3.023 s, System: 0.209 s]
  Range (min … max):    3.273 s …  3.334 s    3 runs

Summary
  PR ran
    1.92 ± 0.16 times faster than main

importlib_metadata on  speedup-normalize [?] via 🐍 v3.14.2 via 💎 v3.1.3 took 38s
❯ hyperfine --warmup 1 -r 3 \
    -n main --prepare 'git checkout main' 'python3.12 benchmark_names.py' \
    -n PR --prepare 'git checkout speedup-normalize -q' 'python3.12 benchmark_names.py'
Benchmark 1: main
  Time (mean ± σ):      5.690 s ±  0.183 s    [User: 5.281 s, System: 0.274 s]
  Range (min … max):    5.492 s …  5.852 s    3 runs

Benchmark 2: PR
  Time (mean ± σ):      6.640 s ±  0.069 s    [User: 6.169 s, System: 0.296 s]
  Range (min … max):    6.560 s …  6.688 s    3 runs

Summary
  main ran
    1.17 ± 0.04 times faster than PR

importlib_metadata on  speedup-normalize [?] via 🐍 v3.14.2 via 💎 v3.1.3 took 50s
❯ hyperfine --warmup 1 -r 3 \
    -n main --prepare 'git checkout main' 'python3.13 benchmark_names.py' \
    -n PR --prepare 'git checkout speedup-normalize -q' 'python3.13 benchmark_names.py'
Benchmark 1: main
  Time (mean ± σ):      5.784 s ±  0.109 s    [User: 5.359 s, System: 0.253 s]
  Range (min … max):    5.674 s …  5.891 s    3 runs

Benchmark 2: PR
  Time (mean ± σ):      5.957 s ±  0.074 s    [User: 5.601 s, System: 0.241 s]
  Range (min … max):    5.876 s …  6.022 s    3 runs

Summary
  main ran
    1.03 ± 0.02 times faster than PR

importlib_metadata on  speedup-normalize [?] via 🐍 v3.14.2 via 💎 v3.1.3 took 50s
❯ hyperfine --warmup 1 -r 3 \
    -n main --prepare 'git checkout main' 'python3.14 benchmark_names.py' \
    -n PR --prepare 'git checkout speedup-normalize -q' 'python3.14 benchmark_names.py'
Benchmark 1: main
  Time (mean ± σ):      5.749 s ±  0.060 s    [User: 5.442 s, System: 0.223 s]
  Range (min … max):    5.684 s …  5.802 s    3 runs

Benchmark 2: PR
  Time (mean ± σ):      2.163 s ±  0.125 s    [User: 1.922 s, System: 0.169 s]
  Range (min … max):    2.020 s …  2.257 s    3 runs

Summary
  PR ran
    2.66 ± 0.16 times faster than main

hugovk marked this pull request as draft January 19, 2026 18:50
hugovk mentioned this pull request Jan 20, 2026
Merged

hugovk commented Feb 6, 2026

Copy link
Copy Markdown
Member Author

Updated to match python/cpython#144083 and now 3.0-3.6x speedup across all Python versions:

Python main (s) PR (s) Result
3.10 6.322 ± 0.229 2.046 ± 0.055 PR 3.09x faster
3.11 5.827 ± 0.083 1.612 ± 0.012 PR 3.62x faster
3.12 5.557 ± 0.041 1.889 ± 0.510 PR 2.94x faster
3.13 5.572 ± 0.069 1.640 ± 0.015 PR 3.40x faster
3.14 5.886 ± 0.266 1.667 ± 0.008 PR 3.53x faster
Details 
❯ hyperfine --warmup 1 -r 3 \
    -n main --prepare 'git checkout main' 'python3.10 benchmark_names.py' \
    -n PR --prepare 'git checkout speedup-normalize -q' 'python3.10 benchmark_names.py'
Benchmark 1: main
  Time (mean ± σ):      6.322 s ±  0.229 s    [User: 5.905 s, System: 0.292 s]
  Range (min … max):    6.057 s …  6.460 s    3 runs

  Warning: Statistical outliers were detected. Consider re-running this benchmark on a quiet system without any interferences from other programs.

Benchmark 2: PR
  Time (mean ± σ):      2.046 s ±  0.055 s    [User: 1.892 s, System: 0.128 s]
  Range (min … max):    2.004 s …  2.108 s    3 runs

Summary
  PR ran
    3.09 ± 0.14 times faster than main

importlib_metadata on  speedup-normalize [?⇡] via 🐍 v3.14.3 via 💎 v3.1.3 took 33s
❯ hyperfine --warmup 1 -r 3 \
    -n main --prepare 'git checkout main' 'python3.11 benchmark_names.py' \
    -n PR --prepare 'git checkout speedup-normalize -q' 'python3.11 benchmark_names.py'
Benchmark 1: main
  Time (mean ± σ):      5.827 s ±  0.083 s    [User: 5.293 s, System: 0.315 s]
  Range (min … max):    5.731 s …  5.877 s    3 runs

  Warning: Statistical outliers were detected. Consider re-running this benchmark on a quiet system without any interferences from other programs.

Benchmark 2: PR
  Time (mean ± σ):      1.612 s ±  0.012 s    [User: 1.514 s, System: 0.095 s]
  Range (min … max):    1.602 s …  1.625 s    3 runs

Summary
  PR ran
    3.62 ± 0.06 times faster than main

importlib_metadata on  speedup-normalize [?⇡] via 🐍 v3.14.3 via 💎 v3.1.3 took 32s
❯ hyperfine --warmup 1 -r 3 \
    -n main --prepare 'git checkout main' 'python3.12 benchmark_names.py' \
    -n PR --prepare 'git checkout speedup-normalize -q' 'python3.12 benchmark_names.py'
Benchmark 1: main
  Time (mean ± σ):      5.557 s ±  0.041 s    [User: 5.265 s, System: 0.216 s]
  Range (min … max):    5.516 s …  5.598 s    3 runs

Benchmark 2: PR
  Time (mean ± σ):      1.889 s ±  0.510 s    [User: 1.564 s, System: 0.179 s]
  Range (min … max):    1.578 s …  2.478 s    3 runs

  Warning: Statistical outliers were detected. Consider re-running this benchmark on a quiet system without any interferences from other programs.

Summary
  PR ran
    2.94 ± 0.79 times faster than main

importlib_metadata on  speedup-normalize [?⇡] via 🐍 v3.14.3 via 💎 v3.1.3 took 32s
❯ hyperfine --warmup 1 -r 3 \
    -n main --prepare 'git checkout main' 'python3.13 benchmark_names.py' \
    -n PR --prepare 'git checkout speedup-normalize -q' 'python3.13 benchmark_names.py'
Benchmark 1: main
  Time (mean ± σ):      5.572 s ±  0.069 s    [User: 5.322 s, System: 0.195 s]
  Range (min … max):    5.492 s …  5.614 s    3 runs

  Warning: Statistical outliers were detected. Consider re-running this benchmark on a quiet system without any interferences from other programs.

Benchmark 2: PR
  Time (mean ± σ):      1.640 s ±  0.015 s    [User: 1.544 s, System: 0.092 s]
  Range (min … max):    1.626 s …  1.655 s    3 runs

Summary
  PR ran
    3.40 ± 0.05 times faster than main

importlib_metadata on  speedup-normalize [?⇡] via 🐍 v3.14.3 via 💎 v3.1.3 took 31s
❯ hyperfine --warmup 1 -r 3 \
    -n main --prepare 'git checkout main' 'python3.14 benchmark_names.py' \
    -n PR --prepare 'git checkout speedup-normalize -q' 'python3.14 benchmark_names.py'
Benchmark 1: main
  Time (mean ± σ):      5.886 s ±  0.266 s    [User: 5.361 s, System: 0.283 s]
  Range (min … max):    5.579 s …  6.050 s    3 runs

  Warning: Statistical outliers were detected. Consider re-running this benchmark on a quiet system without any interferences from other programs.

Benchmark 2: PR
  Time (mean ± σ):      1.667 s ±  0.008 s    [User: 1.573 s, System: 0.091 s]
  Range (min … max):    1.661 s …  1.676 s    3 runs

Summary
  PR ran
    3.53 ± 0.16 times faster than main

hugovk marked this pull request as ready for review February 6, 2026 18:37
This was referenced Mar 20, 2026
Merged
Merged

jaraco commented Mar 20, 2026

Copy link
Copy Markdown
Member

Superseded by #533

jaraco closed this Mar 20, 2026
hugovk deleted the speedup-normalize branch March 21, 2026 10:12
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters. Learn more about bidirectional Unicode characters
Show hidden characters
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

No reviews

Assignees

No one assigned

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

2 participants

Back | FazBrowse Home | New Git URL