benchmarking_python_code.md

You can get detailed performance per function call with this snippet. Edit the lark main and add the setup to the beginning and the result printing to the bottom. Optimally, lark could have a command line option which activates this on its main.

import sys

import io
import pstats
import cProfile

profiler = cProfile.Profile()
profiler.enable()

# Core here

profiler.disable()
with io.StringIO() as outputstream:
    profiller_status = pstats.Stats( profiler, stream=outputstream )
    profiller_status.sort_stats( "time" )
    profiller_status.print_stats()
    sys.stderr.write( outputstream.getvalue()[:2000] + '\n' )

Running this solemnly will output:

         1 function calls in 0.000 seconds

   Ordered by: internal time

      ncalls     tottime     percall     cumtime     percall filename:lineno(function)
           1       0.000       0.000       0.000       0.000 {method 'disable' of '_lsprof.Profiler' objects}

Adding a few cool stuff inside it like:

import mmap
import json

# https://docs.python.org/3.9/library/mmap.html
# https://blog.askesis.pl/post/2019/02/mmap.html
memory = mmap.mmap( -1, 20, access=mmap.ACCESS_WRITE )
memory.write( b'{ "var": 10 }' )

end_of_the_file = memory.tell()
print( 'memory {:,}, end_of_the_file {}\n'.format( len( memory ), end_of_the_file ) )

memory.seek( 0 )
print( 'raw', memory.read( end_of_the_file ) )

memory.seek( 0 )
loaded = json.loads( memory.read( end_of_the_file ) )

print( 'json', loaded )
print( 'loaded', memory[:end_of_the_file] )

Will output:

         3,031 function calls (2,921 primitive calls) in 0.006 seconds

   Ordered by: internal time

      ncalls     tottime     percall     cumtime     percall filename:lineno(function)
          17       0.001       0.000       0.001       0.000 {built-in method nt.stat}
        15/6       0.000       0.000       0.001       0.000 F:\Python\lib\sre_parse.py:475(_parse)
           4       0.000       0.000       0.000       0.000 {built-in method marshal.loads}
           4       0.000       0.000       0.000       0.000 <frozen importlib._bootstrap_external>:914(get_data)
           4       0.000       0.000       0.001       0.000 <frozen importlib._bootstrap_external>:793(get_code)
        25/6       0.000       0.000       0.000       0.000 F:\Python\lib\sre_compile.py:71(_compile)
           4       0.000       0.000       0.000       0.000 {method 'read' of '_io.FileIO' objects}
          14       0.000       0.000       0.000       0.000 F:\Python\lib\sre_compile.py:276(_optimize_charset)
           6       0.000       0.000       0.001       0.000 <frozen importlib._bootstrap_external>:1356(find_spec)
           4       0.000       0.000       0.000       0.000 {built-in method builtins.print}
         102       0.000       0.000       0.000       0.000 F:\Python\lib\sre_parse.py:233(__next)
       31/12       0.000       0.000       0.000       0.000 F:\Python\lib\sre_parse.py:174(getwidth)
           3       0.000       0.000       0.000       0.000 {built-in method builtins.__build_class__}
           2       0.000       0.000       0.000       0.000 {built-in method _imp.create_builtin}
         6/2       0.000       0.000       0.006       0.003 <frozen importlib._bootstrap>:978(_find_and_load)
         123       0.000       0.000       0.000       0.000 F:\Python\lib\sre_parse.py:164(__getitem__)
          39       0.000       0.000       0.000       0.000 {method 'format' of 'str' objects}
           1       0.000       0.000       0.000       0.00

To visualize the data, you can adapt the script to use some tool as snakeviz: https://stackoverflow.com/questions/29630667/how-can-i-analyze-a-file-created-with-pstats-dump-statsfilename-off-line

import os
import sys

import snakeviz
from snakeviz.cli import main

...
profiler.disable()

profilepath = "path_to_data.prof"
profiler.dump_stats(profilepath)

sys.argv.append(profilepath)
sys.exit(main())

It is a webpage you can visuallly browse the cPrifle results:

evandrocoan/benchmarking_python_code.md