perf: Query distro.version sources lazily

[ichard26]

Hey, I first wanted to say thanks for maintaining distro. From the pip team, we appreciate it!

---

When /etc/os-release provides sufficiently high-quality version information directly, there is no need to call lsb_release in a subprocess or query any other version.

This speeds up distro.version() massively when lsb_release is slow to execute, like on Ubuntu 26.04 LTS where it can easily take north of 15ms.

For context, pip uses distro.version() to populate its user-agent header so we'd like this function to be as fast as reasonably possible.

For benchmarking, you can use this which is equivalent to what pip does:

import distro
import time

def run():
    t0 = time.perf_counter()
    linux_distribution = distro.name(), distro.version(), distro.codename()
    t1 = time.perf_counter()

    print(f"elapsed: {(t1 - t0)*1000:.2f}ms")

run()

Under CPython 3.15's new sampling profiler on Ubuntu 26.04 LTS:

~/dev/oss/distro ❯ python -m profiling.sampling run -r 30khz test.py
elapsed: 16.21ms
Captured 1,028 samples in 0.03 seconds
Sample rate: 30,302.98 samples/sec
Error rate: 24.61
Profile Stats:
       nsamples   sample%  tottime (ms)    cumul%  cumtime (ms)  filename:lineno(function)
        429/429      55.5        14.157      55.5        14.157  subprocess.py:1376(Popen.communicate)
        157/291      20.3         5.181      37.6         9.603  <frozen importlib._bootstrap>:549(_call_with_frames_removed)
          45/45       5.8         1.485       5.8         1.485  <frozen importlib._bootstrap_external>:500(_compile_bytecode)
          24/24       3.1         0.792       3.1         0.792  subprocess.py:2080(Popen._execute_child)
            8/8       1.0         0.264       1.0         0.264  <frozen genericpath>:40(isfile)
            5/5       0.6         0.165       0.6         0.165  <frozen importlib._bootstrap_external>:213(_write_atomic)
            5/5       0.6         0.165       0.6         0.165  _parser.py:450(_uniq)
            4/4       0.5         0.132       0.5         0.132  <frozen importlib._bootstrap_external>:923(FileLoader.get_data)
            3/3       0.4         0.099       0.4         0.099  <frozen importlib._bootstrap_external>:517(_code_to_timestamp_pyc)
            3/3       0.4         0.099       0.4         0.099  <frozen importlib._bootstrap_external>:152(_path_stat)
            3/3       0.4         0.099       0.4         0.099  __init__.py:447(namedtuple)
            3/3       0.4         0.099       0.4         0.099  distro.py:1351(LinuxDistribution._distro_release_info)
            2/2       0.3         0.066       0.3         0.066  <frozen importlib._bootstrap_external>:211(_write_atomic)
            2/2       0.3         0.066       0.3         0.066  argparse.py:1020(<module>)
            2/2       0.3         0.066       0.3         0.066  __init__.py:515(namedtuple)

Legend:
  nsamples: Direct/Cumulative samples (direct executing / on call stack)
  sample%: Percentage of total samples this function was directly executing
  tottime: Estimated total time spent directly in this function
  cumul%: Percentage of total samples when this function was on the call stack
  cumtime: Estimated cumulative time (including time in called functions)
  filename:lineno(function): Function location and name

[HorlogeSkynet]

Hello @ichard26, and thanks for your patch. I agree it's a great improvement when best=False.

[nir0s]

@HorlogeSkynet, did you not merge on purpose?

[HorlogeSkynet]

@HorlogeSkynet, did you not merge on purpose?

Yes, just in case one wanted to express a different opinion on the matter. We can merge if you approve too !

[ichard26]

Thanks / merci beaucoup ! I'm looking forward to a release.