ObjWatch

[ English | 中文 ]

Overview

ObjWatch is a robust Python library designed to streamline the debugging and monitoring of complex projects. By offering real-time tracing of object attributes and method calls, ObjWatch empowers developers to gain deeper insights into their codebases, facilitating issue identification, performance optimization, and overall code quality enhancement.

⚠️ Performance Warning

ObjWatch may impact your application's performance. It is recommended to use it solely in debugging environments.

Features

Nested Structure Tracing: Visualize and monitor nested function calls and object interactions with clear, hierarchical logging.
Enhanced Logging Support: Leverage Python's built-in logging module for structured, customizable log outputs, including support for simple and detailed formats. Additionally, to ensure logs are captured even if the logger is disabled or removed by external libraries, you can set level="force". When level is set to "force", ObjWatch bypasses the standard logging handlers and uses print() to output log messages directly to the console, ensuring that critical debugging information is not lost.
Logging Message Types: ObjWatch categorizes log messages into various types to provide detailed insights into code execution. The primary types include:
- run: Indicates the start of a function or class method execution.
- end: Signifies the end of a function or class method execution.
- upd: Represents the creation of a new variable.
- apd: Denotes the addition of elements to data structures like lists, sets, or dictionaries.
- pop: Marks the removal of elements from data structures like lists, sets, or dictionaries.
These classifications help developers efficiently trace and debug their code by understanding the flow and state changes within their applications.
Multi-GPU Support: Seamlessly trace distributed PyTorch applications running across multiple GPUs, ensuring comprehensive monitoring in high-performance environments.
Custom Wrapper Extensions: Extend ObjWatch's functionality with custom wrappers, allowing tailored tracing and logging to fit specific project needs.
Context Manager & API Integration: Integrate ObjWatch effortlessly into your projects using context managers or API functions without relying on command-line interfaces.

Installation

ObjWatch is available on PyPI. Install it using pip:

pip install objwatch

Getting Started

Basic Usage

ObjWatch can be utilized as a context manager or through its API within your Python scripts.

Using as a Context Manager

import objwatch

def main():
    # Your application code
    pass

if __name__ == '__main__':
    with objwatch.ObjWatch(['your_module.py']):
        main()

Using the API

import objwatch

def main():
    # Your application code
    pass

if __name__ == '__main__':
    obj_watch = objwatch.watch(['your_module.py'])
    main()
    obj_watch.stop()

Example Usage

Below is a comprehensive example demonstrating how to integrate ObjWatch into a Python script:

import objwatch
import time

class SampleClass:
    def __init__(self, value):
        self.value = value

    def increment(self):
        self.value += 1
        time.sleep(0.1)

    def decrement(self):
        self.value -= 1
        time.sleep(0.1)

def main():
    obj = SampleClass(10)
    for _ in range(5):
        obj.increment()
    for _ in range(3):
        obj.decrement()

if __name__ == '__main__':
    # Using as a Context Manager with Detailed Logging
    with objwatch.ObjWatch(['examples/example_usage.py']):
        main()

    # Using the API with Simple Logging
    obj_watch = objwatch.watch(['examples/example_usage.py'])
    main()
    obj_watch.stop()

When running the above script, ObjWatch will generate logs similar to the following:

Expected Log Output

[2024-12-14 20:40:56] [DEBUG] objwatch: Processed targets: {'examples/example_usage.py'}
[2024-12-14 20:40:56] [INFO] objwatch: Starting ObjWatch tracing.
[2024-12-14 20:40:56] [INFO] objwatch: Starting tracing.
[2024-12-14 20:40:56] [DEBUG] objwatch: run main
[2024-12-14 20:40:56] [DEBUG] objwatch: | run SampleClass.__init__
[2024-12-14 20:40:56] [DEBUG] objwatch: | end SampleClass.__init__
[2024-12-14 20:40:56] [DEBUG] objwatch: | run SampleClass.increment
[2024-12-14 20:40:56] [DEBUG] objwatch: | | upd SampleClass.value
[2024-12-14 20:40:56] [DEBUG] objwatch: | | upd SampleClass.value
[2024-12-14 20:40:56] [DEBUG] objwatch: | end SampleClass.increment
[2024-12-14 20:40:56] [DEBUG] objwatch: | run SampleClass.increment
[2024-12-14 20:40:56] [DEBUG] objwatch: | | upd SampleClass.value
[2024-12-14 20:40:57] [DEBUG] objwatch: | end SampleClass.increment
[2024-12-14 20:40:57] [DEBUG] objwatch: | run SampleClass.increment
[2024-12-14 20:40:57] [DEBUG] objwatch: | | upd SampleClass.value
[2024-12-14 20:40:57] [DEBUG] objwatch: | end SampleClass.increment
[2024-12-14 20:40:57] [DEBUG] objwatch: | run SampleClass.increment
[2024-12-14 20:40:57] [DEBUG] objwatch: | | upd SampleClass.value
[2024-12-14 20:40:57] [DEBUG] objwatch: | end SampleClass.increment
[2024-12-14 20:40:57] [DEBUG] objwatch: | run SampleClass.increment
[2024-12-14 20:40:57] [DEBUG] objwatch: | | upd SampleClass.value
[2024-12-14 20:40:57] [DEBUG] objwatch: | end SampleClass.increment
[2024-12-14 20:40:57] [DEBUG] objwatch: | run SampleClass.decrement
[2024-12-14 20:40:57] [DEBUG] objwatch: | | upd SampleClass.value
[2024-12-14 20:40:57] [DEBUG] objwatch: | end SampleClass.decrement
[2024-12-14 20:40:57] [DEBUG] objwatch: | run SampleClass.decrement
[2024-12-14 20:40:57] [DEBUG] objwatch: | | upd SampleClass.value
[2024-12-14 20:40:57] [DEBUG] objwatch: | end SampleClass.decrement
[2024-12-14 20:40:57] [DEBUG] objwatch: | run SampleClass.decrement
[2024-12-14 20:40:57] [DEBUG] objwatch: | | upd SampleClass.value
[2024-12-14 20:40:57] [DEBUG] objwatch: | end SampleClass.decrement
[2024-12-14 20:40:57] [DEBUG] objwatch: end main
[2024-12-14 20:40:57] [INFO] objwatch: Stopping ObjWatch tracing.
[2024-12-14 20:40:57] [INFO] objwatch: Stopping tracing.
[2024-12-14 20:40:57] [DEBUG] objwatch: Processed targets: {'examples/example_usage.py'}
[2024-12-14 20:40:57] [INFO] objwatch: Starting ObjWatch tracing.
[2024-12-14 20:40:57] [INFO] objwatch: Starting tracing.
[2024-12-14 20:40:57] [DEBUG] objwatch: run main
[2024-12-14 20:40:57] [DEBUG] objwatch: | run SampleClass.__init__
[2024-12-14 20:40:57] [DEBUG] objwatch: | end SampleClass.__init__
[2024-12-14 20:40:57] [DEBUG] objwatch: | run SampleClass.increment
[2024-12-14 20:40:57] [DEBUG] objwatch: | | upd SampleClass.value
[2024-12-14 20:40:57] [DEBUG] objwatch: | | upd SampleClass.value
[2024-12-14 20:40:57] [DEBUG] objwatch: | end SampleClass.increment
[2024-12-14 20:40:57] [DEBUG] objwatch: | run SampleClass.increment
[2024-12-14 20:40:57] [DEBUG] objwatch: | | upd SampleClass.value
[2024-12-14 20:40:57] [DEBUG] objwatch: | end SampleClass.increment
[2024-12-14 20:40:57] [DEBUG] objwatch: | run SampleClass.increment
[2024-12-14 20:40:57] [DEBUG] objwatch: | | upd SampleClass.value
[2024-12-14 20:40:57] [DEBUG] objwatch: | end SampleClass.increment
[2024-12-14 20:40:57] [DEBUG] objwatch: | run SampleClass.increment
[2024-12-14 20:40:57] [DEBUG] objwatch: | | upd SampleClass.value
[2024-12-14 20:40:58] [DEBUG] objwatch: | end SampleClass.increment
[2024-12-14 20:40:58] [DEBUG] objwatch: | run SampleClass.increment
[2024-12-14 20:40:58] [DEBUG] objwatch: | | upd SampleClass.value
[2024-12-14 20:40:58] [DEBUG] objwatch: | end SampleClass.increment
[2024-12-14 20:40:58] [DEBUG] objwatch: | run SampleClass.decrement
[2024-12-14 20:40:58] [DEBUG] objwatch: | | upd SampleClass.value
[2024-12-14 20:40:58] [DEBUG] objwatch: | end SampleClass.decrement
[2024-12-14 20:40:58] [DEBUG] objwatch: | run SampleClass.decrement
[2024-12-14 20:40:58] [DEBUG] objwatch: | | upd SampleClass.value
[2024-12-14 20:40:58] [DEBUG] objwatch: | end SampleClass.decrement
[2024-12-14 20:40:58] [DEBUG] objwatch: | run SampleClass.decrement
[2024-12-14 20:40:58] [DEBUG] objwatch: | | upd SampleClass.value
[2024-12-14 20:40:58] [DEBUG] objwatch: | end SampleClass.decrement
[2024-12-14 20:40:58] [DEBUG] objwatch: end main
[2024-12-14 20:40:58] [INFO] objwatch: Stopping ObjWatch tracing.
[2024-12-14 20:40:58] [INFO] objwatch: Stopping tracing.

Configuration

ObjWatch offers customizable logging formats and tracing options to suit various project requirements. Utilize the simple parameter to toggle between detailed and simplified logging outputs.

Parameters

targets (list): Files or modules to monitor.
ranks (list, optional): GPU ranks to track when using torch.distributed.
output (str, optional): Path to a file for writing logs.
output_xml (str, optional): Path to the XML file for writing structured logs. If specified, tracing information will be saved in a nested XML format for easy browsing and analysis.
level (str, optional): Logging level (e.g., DEBUG, INFO).
simple (bool, optional): Enable simple logging mode with the format "DEBUG: {msg}".
wrapper (FunctionWrapper, optional): Custom wrapper to extend tracing and logging functionality.
with_locals (bool, optional): Enable tracing and logging of local variables within functions during their execution.
with_module_path (bool, optional): Control whether to prepend the module path to function names in logs.

Advanced Usage

Multi-GPU Support

ObjWatch seamlessly integrates with distributed PyTorch applications, allowing you to monitor and trace operations across multiple GPUs. Specify the ranks you wish to track using the ranks parameter.

import objwatch

def main():
    # Your distributed application code
    pass

if __name__ == '__main__':
    obj_watch = objwatch.watch(['distributed_module.py'], ranks=[0, 1, 2, 3], output='./dist.log, simple=False)
    main()
    obj_watch.stop()

Custom Wrapper Extensions

ObjWatch provides the FunctionWrapper abstract base class, enabling users to create custom wrappers that extend and customize the library's tracing and logging capabilities. By subclassing FunctionWrapper, developers can implement tailored behaviors that execute during function calls and returns, offering deeper insights and specialized monitoring suited to their project's specific needs.

FunctionWrapper Class

The FunctionWrapper class defines two essential methods that must be implemented:

wrap_call(self, func_name: str, frame: FrameType) -> str:

This method is invoked at the beginning of a function call. It receives the function name and the current frame object, which contains the execution context, including local variables and the call stack. Implement this method to extract, log, or modify information before the function executes.
wrap_return(self, func_name: str, result: Any) -> str:

This method is called upon a function's return. It receives the function name and the result returned by the function. Use this method to log, analyze, or alter information after the function has completed execution.

For more details on frame objects, refer to the official Python documentation.

TensorShapeLogger

As an example of a custom wrapper, ObjWatch includes the TensorShapeLogger class within the objwatch.wrappers module. This wrapper automatically logs the shapes of tensors involved in function calls, which is particularly beneficial in machine learning and deep learning workflows where tensor dimensions are critical for model performance and debugging.

Creating and Integrating Custom Wrappers

To create a custom wrapper:

Subclass FunctionWrapper: Define a new class that inherits from FunctionWrapper and implement the wrap_call and wrap_return methods to define your custom behavior.
Initialize ObjWatch with the Custom Wrapper: When initializing ObjWatch, pass your custom wrapper via the wrapper parameter. This integrates your custom tracing logic into the ObjWatch tracing process.

By leveraging custom wrappers, you can enhance ObjWatch to capture additional context, perform specialized logging, or integrate with other monitoring tools, thereby providing a more comprehensive and tailored tracing solution for your Python projects.

Example Use Cases

For example, the TensorShapeLogger can be integrated as follows:

from objwatch.wrappers import TensorShapeLogger

# Initialize ObjWatch with the custom TensorShapeLogger
obj_watch = objwatch.ObjWatch(['your_module.py'], simple=False, wrapper=TensorShapeLogger))
with obj_watch:
    main()

Example of Using a Custom Wrapper

Extend ObjWatch's functionality by creating custom wrappers. This allows you to tailor the tracing and logging mechanisms to fit specific needs within your projects.

from objwatch.wrappers import FunctionWrapper

class CustomWrapper(FunctionWrapper):
    def wrap_call(self, func_name, frame):
        return f" - Called {func_name} with args: {frame.f_locals}"

    def wrap_return(self, func_name, result):
        return f" - {func_name} returned {result}"

# Integrate the custom wrapper
obj_watch = objwatch.watch(['your_module.py'], simple=False, wrapper=CustomWrapper)
main()
obj_watch.stop()

Support

If you encounter any issues or have questions, feel free to open an issue on the ObjWatch GitHub repository or reach out via email at [email protected].

Acknowledgements

Inspired by the need for better debugging and understanding tools in large Python projects.
Powered by Python's robust tracing and logging capabilities.

Name		Name	Last commit message	Last commit date
Latest commit History 40 Commits
.github		.github
examples		examples
objwatch		objwatch
requirements		requirements
tests		tests
.gitignore		.gitignore
CODE_OF_CONDUCT.md		CODE_OF_CONDUCT.md
CONTRIBUTING.md		CONTRIBUTING.md
LICENSE		LICENSE
MANIFEST.in		MANIFEST.in
README.md		README.md
README_zh.md		README_zh.md
SECURITY.md		SECURITY.md
setup.py		setup.py
version.txt		version.txt

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Repository files navigation

ObjWatch

Overview

Features

Installation

Getting Started

Basic Usage

Using as a Context Manager

Using the API

Example Usage

Configuration

Parameters

Advanced Usage

Multi-GPU Support

Custom Wrapper Extensions

FunctionWrapper Class

TensorShapeLogger

Creating and Integrating Custom Wrappers

Example Use Cases

Example of Using a Custom Wrapper

Support

Acknowledgements

About

Releases 7

Languages

License

aeeeeeep/objwatch

Folders and files

Latest commit

History

Repository files navigation

ObjWatch

Overview

Features

Installation

Getting Started

Basic Usage

Using as a Context Manager

Using the API

Example Usage

Configuration

Parameters

Advanced Usage

Multi-GPU Support

Custom Wrapper Extensions

FunctionWrapper Class

TensorShapeLogger

Creating and Integrating Custom Wrappers

Example Use Cases

Example of Using a Custom Wrapper

Support

Acknowledgements

About

Topics

Resources

License

Code of conduct

Security policy

Stars

Watchers

Forks

Releases 7

Languages