benchpy

benchpy is a Python utility to benchmark code.

Installation

pip install benchpy

Usage

Benchmarks are grouped and named. The default group is None and the default name is the name of the function being benchmarked.

from benchpy import benchmarked

@benchmarked(group='passing', name='')
def function():
  pass

You may also benchmark snippet of code using a context manager, in this case, you must name your benchmark.

from benchpy import benchmarked

with benchmarked(group='passing', name='passing in a context'):
    pass

Then you can use the statistics function to get an overview of the results of a given function.

from benchpy import benchmarked

@benchmarked()
def foobar():
    pass

print(benchmarked.results('foobar'))
print(benchmarked.statistics('foobar'))

Examples covering basic usage are providen in the examples folder.

Results from results is stored as a numpy array:

[[0.15900000000000003, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1.0],
  [0.15900000000000003, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1.0],
  [0.15900000000000003, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1.0]]

You can then apply any data treatment numpy provides, which is practical for plotting your results.

The statistics function will compute simple statistics about your result set:

# [user time, system time, ...]
max: [0.15900000000000003, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1.0],
sum: [0.15900000000000003, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1.0],
avg: [0.15900000000000003, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1.0],
min: [0.15900000000000003, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1.0],
med: [0.15900000000000003, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1.0]

The fields for each calculation will match those of resource.getrusage: getrusage documentation.

benchmarked also allow you to define what kind of rusage data you want, defaulted to RUSAGE_SELF.

@benchmarked(rusage=RUSAGE_BOTH)
def foo():
    pass

The resource module is not available in Windows, so only the second column will contain relevant information.