Perfwatch

Perfwatch is a python package that allows you to monitor the performance of your code. It is designed to be used in a Jupyter notebook, but can also be used in a Python script.

Table of Contents

• Installation
• Usage
• Development
• Usage
  • Available Profiler Types
  • Basic Usage
  • Customizing Profiling
  • Logging
• License

Installation

To install perfwatch, run the following command:

pip install perfwatch

Development

To install perfwatch for development, clone the repository and run the following command:

poetry install

To setup pre-commit hooks, run the following command:

poetry run pre-commit install

To run the tests, run the following command:

poetry run pytest

Usage

Available Profiler Types

The profiler supports the following profiler types:

• cpu: Profiles CPU usage using the cProfile module.
• memory: Profiles memory usage using the memory_profiler module.
• thread: Profiles thread creation and usage.
• io: Profiles I/O operations.
• network: Profiles network traffic using the NetworkProfiler class.
• gpu: Profiles GPU usage using the GPUProfiler class.
• cache: Profiles cache performance (not implemented).
• exception: Profiles exception handling (not implemented).
• system: Profiles system performance (not implemented).
• distributed: Profiles distributed system performance (not implemented).
• line: Profiles line-by-line execution using the line_profiler module.
• time: Profiles execution time.

Basic Usage

from perfwatch import watch

@watch(["line", "cpu", "time"])
def test():
    for _ in range(1000000):
        pass

if __name__ == "__main__":
    test()

Customizing Profiling

You can customize the profiling behavior by passing additional keyword arguments to the watch decorator. For example:

@watch("network", packet_src="localhost")
def my_function(x, y):
    # function implementation
    pass

Logging

The Profiler Service uses a logger to output profiling results. You can specify a log file path using the log_file_path keyword argument:

@watch("cpu", log_file_path="profiling.log")
def my_function(x, y):
    # function implementation
    pass

License

This project is licensed under the MIT License - see the LICENSE file for details.