statprof - statistical profiling for Python

This package provides a simple statistical profiler for Python.

Python's default profiler has been lsprof for several years. This is an instrumenting profiler, which means that it saves data on every action of interest. In the case of lsprof, it runs at function entry and exit. This has problems: it can be expensive due to frequent sampling, and it is blind to hot spots within a function.

In contrast, statprof samples the call stack periodically (by default, 1000 times per second), and it correctly tracks line numbers inside a function. This means that if you have a 50-line function that contains two hot loops, statprof is likely to report them both accurately.

Note

This package does not yet work on Windows! See the implementation and portability notes below for details.

Usage

It's easy to get started with statprof:

import statprof

statprof.start()
try:
    my_questionable_function()
finally:
    statprof.stop()
    statprof.display()

Or with a contextmanager :

import statprof

with statprof.profile():
    my_questionable_function()

The profiler can be invoked at more than one place inside your code and will report its findings for all of them at once at the end:

import statprof

statprof.start()
try:
    my_questionable_function()
finally:
    statprof.stop()

uninteresting_code()

statprof.start()
try:
    my_other_questionable_function()
finally:
    statprof.stop()

statprof.display()

However, when you are profiling your code by repeatedly executing it in IPython, each run will add new samples to the previously collected ones, and display results aggregated over all runs. Since you will likely just want to see the results for the last run, remember to reset the profiler first:

import statprof

statprof.reset()
with statprof.profile():
    my_questionable_function()

For more comprehensive help, run pydoc statprof.

Portability

Because statprof uses the Unix itimer signal facility, it does not currently work on Windows. (Patches to improve portability would be most welcome.)

Implementation notes

The statprof profiler works by setting the Unix profiling signal ITIMER_PROF to go off after the interval you define in the call to reset(). When the signal fires, a sampling routine is run which looks at the current procedure that's executing, and then crawls up the stack, and for each frame encountered, increments that frame's code object's sample count. Note that if a procedure is encountered multiple times on a given stack, it is only counted once. After the sampling is complete, the profiler resets profiling timer to fire again after the appropriate interval.

Meanwhile, the profiler keeps track, via os.times(), how much CPU time (system and user -- which is also what ITIMER_PROF tracks), has elapsed while code has been executing within a start()/stop() block.

The profiler also tries (as much as possible) to avoid counting or timing its own code.

History

This package was originally written and released by Andy Wingo. It was ported to modern Python by Alex Frazer, and posted to github by Jeff Muizelaar. The current maintainer is Bryan O'Sullivan.

Reporting bugs, contributing patches

The current maintainer of this package is Bryan O'Sullivan.

Please report bugs using the github issue tracker.

If you'd like to contribute patches, please do - the source is on github, so please just issue a pull request.

$ git clone git://github.com/bos/statprof.py