pytest-line-profiler

line-by-line profiling for code executed by pytest, using line-profiler.

Why?

Line profiler is a wonderful tool to easily identify bottlenecks inside specific functions of your code, and quantify the improvements after a refactor.

Using it is straightforward but required to instrument the functions you want to profile with a "virtual" @profile decorator and then execute "a trigger script" (code that calls the decorated functions somehow) via kernprof.py which works as a python wrapper that understands the decorator, register the functions to be profiled, and print the stats when the script finishes.

Altought it does its job, is a bit invasive: you need to have an special "instrumented" version of your code, and execute it in a way that potentially clashes with the way you do normally (for instance, through a shortcut command from your editor, a test runner, another script, etc.)

Moreover, frequently in real case scenarios, "a trigger script" isn't just a simple function call. You need to prepare input data, connect to external resources, etc. And that's exactly what a test can do, right?

Installation

You can install "pytest-line-profiler" via pip from PyPI.

pip install git+https://github.com/ArchangelImaging/dep-pytest-line-profiler@main

Alternatively, you can add the package to your poetry project's pyproject.toml file as follows

[tool.poetry.dependencies]
pytest-line-profiler = {git = "https://github.com/ArchangelImaging/dep-pytest-line-profiler", rev = "main"}

Usage

Mark your test passing the functions you wants to profile as positional arguments, like @pytest.mark.line_profile.with_args(function1, function2, [...])

If your test exercises any of those functions, you'll get the profile result as a report.

For example:

import pytest

def f(i):
    return i * 10

def g(n=10):
    return sum(f(i) for i in range(10))


@pytest.mark.line_profile.with_args(f, g)
def test_as_mark():
    assert g() == 450

After that test is executed, you'll get the stats from the line profiler instance.

============ Line Profile result for tests/test_line_profiler.py::test_as_mark ============
Timer unit: 1e-06 s

Total time: 4e-06 s
File: /home/tin/lab/pytest-line-profiler/tests/test_line_profiler.py
Function: f at line 4

Line #      Hits         Time  Per Hit   % Time  Line Contents
==============================================================
     4                                           def f(i):
     5        10          4.0      0.4    100.0      return i * 10

Total time: 3e-05 s
File: /home/tin/lab/pytest-line-profiler/tests/test_line_profiler.py
Function: g at line 7

Line #      Hits         Time  Per Hit   % Time  Line Contents
==============================================================
     7                                           def g(n=10):
     8         1         30.0     30.0    100.0      return sum(f(i) for i in range(10))

Alternatively, you can run any test passing the function/s to profile from the command line

$ pytest --line-profile path.to.function_to_be_profiled [...] 

The line-profiler library offer flexibility to change the output time units, to change the default output time unit you can set output_unit in the function decorator. For example:

@pytest.mark.line_profile.with_args(f, g, output_unit=1e-3)
def test_as_mark():
    assert g() == 450

Archangel Edit -> change output units

I have added the option to specify the output units in the decorator, e.g.

import pytest

def f(i):
    return i * 10

def g(n=10):
    return sum(f(i) for i in range(10))


@pytest.mark.line_profile.with_args(f, g, output_unit=1e-3)
def test_as_mark():
    assert g() == 450

this to make the output more legible

Contributing

Contributions are very welcome. Tests can be run with [pytest][https://github.com/pytest-dev/pytest], please ensure the coverage at least stays the same before you submit a pull request.

License

Distributed under the terms of the MIT license, "pytest-line-profiler" is free and open source software

Issues

If you encounter any problems, please file an issue along with a detailed description.