Extremelly slow checks

[pycqa]

Extremelly slow checks

[JensRantil]

Oh -- I am running Python 2.7.1.

[JensRantil]

This can be reproduced from command line. Execution took 17 seconds with:

$ ../bin/pep257 rewind/*.py rewind/test/*.py

[JensRantil]

Oh, and one last thing, if you would like me to publish my current pep257 branch for you to reproduce this, I can do that.

[JensRantil]

Just looking at the code it seems like parse_methods(...) might be called many times with the same arguments in parse_contexts(...) (and takes account for 17 seconds of execution). Refactoring so that each source is parsed only once and having the results cached, one might decrease the time spent in that function a lot. This could possibly be done by implementing something similar to this or this.

[JensRantil]

This is what the profiling result looks like after the changes that pull request #12 introduces:

 ncalls  tottime  percall  cumtime  percall filename:lineno(function)
      1    0.002    0.002    6.025    6.025 pep257.py:67(<module>)
      1    0.003    0.003    5.996    5.996 pep257.py:366(main)
  190/6    0.001    0.000    5.991    0.999 pep257.py:130(<lambda>)
    118    0.008    0.000    5.991    0.051 pep257.py:323(check_source)
     96    0.000    0.000    4.988    0.052 pep257.py:242(parse_contexts)
    533    0.180    0.000    4.269    0.008 pep257.py:188(parse_top_level)
   1264    1.169    0.001    3.480    0.003 StringIO.py:168(readlines)
   1040    0.007    0.000    2.894    0.003 pep257.py:137(abs_pos)
     65    0.000    0.000    2.495    0.038 pep257.py:208(parse_classes)
1152457    1.685    0.000    2.222    0.000 StringIO.py:139(readline)
 331865    1.060    0.000    2.033    0.000 tokenize.py:264(generate_tokens)
 314711    0.080    0.000    2.022    0.000 {next}
     78    0.000    0.000    1.774    0.023 pep257.py:204(parse_functions)
     78    0.000    0.000    0.944    0.012 pep257.py:118(cached_func)
    135    0.003    0.000    0.944    0.007 pep257.py:226(parse_methods)
 309500    0.771    0.000    0.771    0.000 {built-in method match}
    112    0.002    0.000    0.707    0.006 pep257.py:278(__init__)
    224    0.060    0.000    0.703    0.003 pep257.py:144(rel_pos)
1152693    0.362    0.000    0.362    0.000 {method 'find' of 'str' objects}
   2108    0.018    0.000    0.200    0.000 pep257.py:172(parse_docstring)

[JensRantil]

Every file checked generates two calls to parse_functions(...), parse_classes(...) and parse_methods(...) through parse_contexts(...). By caching those results whenever possible a one would be able to gain 2.495/2+1.774/2+ 0.944/2=2.59 seconds more in the example above.

[JensRantil]

Every file checked generates two calls to parse_functions(...), parse_classes(...) and parse_methods(...) through parse_contexts(...). By caching those results whenever possible a one would be able to gain 2.495/2+1.774/2+ 0.944/2=2.59 seconds more in the example above.

This fix was added in pull request #12.

[keleshev]

#12 is merged now. pep257 is probably still 10 times slower than it could be, but good enough for now.

If you come up with more speed improvements—you are welcome!

[JensRantil]

#12 is merged now. pep257 is probably still 10 times slower than it could be, but good enough for now.

I agree, but works for now!

If you come up with more speed improvements—you are welcome!

Sure! But for now it doesn't hinder me in running my tests.