How to cProfile

How to cProfile Python scripts written by others (where you don't know which file to edit or where to even start looking).

Problem

Assume we want to profile a pip-installed Python application which provides a script with an entry_point. This is our example script with example arguments:

$ myscript --somearg --anotheroption X

After browsing the cProfile documentation it was not clear to me where to insert the cProfile.run() for the pip installed script.

After some trial and error, the following worked:

How to profile the script

$ python -m cProfile -o restats $(which myscript) --somearg --anotheroption X

Producing statistics from the restats file

Run this script to extract statistics:

import pstats

p = pstats.Stats('restats')
p.sort_stats('time').print_stats(10)

Example output:

Tue Jan 21 21:08:28 2020    restats

         8168533 function calls (7730858 primitive calls) in 13.080 seconds

   Ordered by: internal time
   List reduced from 5735 to 10 due to restriction <10>

   ncalls  tottime  percall  cumtime  percall filename:lineno(function)
      174    2.932    0.017    2.932    0.017 {method 'executemany' of 'sqlite3.Cursor' objects}
    81975    0.743    0.000    1.389    0.000 {method 'randint' of 'numpy.random.mtrand.RandomState' objects}
395367/19848    0.582    0.000    1.498    0.000 /home/user/.local/share/virtualenvs/somecode-0C9px_8o/lib/python3.8/copy.py:128(deepcopy)
    35788    0.318    0.000    0.439    0.000 /home/user/.local/share/virtualenvs/somecode-0C9px_8o/lib/python3.8/site-packages/scipy/special/_basic.py:1938(comb)
    16562    0.222    0.000    0.227    0.000 /home/user/somecode/somecode/cluster_data.py:139(_add_excitor)
   146757    0.216    0.000    0.216    0.000 {built-in method numpy.array}
    81984    0.210    0.000    0.646    0.000 /home/user/.local/share/virtualenvs/somecode-0C9px_8o/lib/python3.8/site-packages/numpy/core/_dtype.py:333(_name_get)
     9925    0.208    0.000    2.795    0.000 /home/user/somecode/somecode/cc_rep_w_sampling.py:85(_get_cluster_in_combo)
    49161    0.176    0.000    0.321    0.000 /home/user/somecode/somecode/hvertices.py:332(_fully_specified)
    69860    0.166    0.000    0.166    0.000 {built-in method numpy.arange}