Pymortem

Pymortem is a post-mortem debugging tool that lets you inspect and manipulate execution contexts after exceptions occur. Unlike traditional debuggers that require a separate interactive shell, pymortem gives you direct access to all variables and frames in the exception stack, making it valuable in Jupyter notebooks and interactive environments.

This package evolved from an educational blog post on post-mortem debugging techniques. What began as educational code examples has been refined into a practical debugging library.

1 Installation

pip install pymortem

2 Features

• Enhanced Tracebacks: Rich, visual traceback output showing code context around errors with line numbers and error indicators
• Frame Inspection: Directly examine variables at any level in the call stack without navigating through a separate command interface
• Code Execution in Context: Run arbitrary code in the context of any stack frame without restarting your program
• Chained Exception Support: Clear visualization of exception chains, showing both “raised from” and “during handling” relationships
• No Special Setup: Works with standard Python without requiring breakpoints or special execution modes

3 Usage

Examining an Exception after it Occurs

# In one cell where an error happens:
def foo():
    x = 10
    output = x / 0
    return output

foo()

# In the next cell, examine the exception:
import pymortem

# Get enhanced traceback and frame information
traceback_msg, frames = pymortem.extract_from_exception()

# Display the improved traceback
print(traceback_msg)

Inspecting Variables in the Error Context

# After running the above cells
# Let's examine the local variables in different frames

# The frame where the error occurred
print("Locals in error frame:", frames[-1]["locals"])

# Check global variables too
print("Some globals:", {k: v for k, v in list(frames[-1]["globals"].items())[:5]})

Executing Code in a Frame’s Context

import pymortem
import sys

# Get the most recent exception
exception = pymortem.retrieve_the_last_exception() # Store the exception
_, frames = pymortem.extract_from_exception(exception)

# Choose a frame to work with (e.g., frames[1] for a specific frame)
work_frame = frames[-1]

# Execute code in that frame's context
pymortem.execute(
    """
    # You can access all variables that existed when the error occurred
    print("Available variables:", list(locals().keys()))

    # Test potential fixes without rerunning the entire notebook
    try:
        # Try a fix for a ZeroDivisionError
        denominator = 2  # Was 0 before
        fixed_result = x / denominator
        print(f"Fix worked! Result = {fixed_result}")
    except Exception as e:
        print(f"Fix didn't work: {e}")
    """,
    work_frame
)

Handling Chained Exceptions

# Create a chained exception scenario
try:
    try:
        x = {"key": "value"}
        result = x["missing_key"]  # Will raise KeyError
    except KeyError:
        result = 10 + "0"  # Will raise ValueError
except Exception as e:
    chain_exception = e

# Examine the exception chain
traceback_msg, all_frames = pymortem.extract_from_exception(chain_exception)
print(traceback_msg)
print("")

# Frames are arranged in chronological order, with the first exception first
original_error_frame = all_frames[0]  # Frame from the KeyError
raised_from_frame = all_frames[-1]    # Frame from the ValueError

print(f"First exception type: {type(chain_exception.__cause__)}")
print(f"Second exception type: {type(chain_exception)}")

4 Why Use Pymortem?

Post-mortem debugging in Python traditionally requires using tools like pdb.pm() or %debug in IPython, which launch a separate command interface with its own syntax and navigation model. Pymortem takes a different approach:

1. Direct Context Access: Instead of a separate debugging shell, access frame data directly in your current Python environment
2. Better Visualization: See more context around exceptions with cleaner, more informative tracebacks
3. Natural Code Execution: Run diagnostic code directly in frame contexts using standard Python syntax
4. Stays in Flow: Particularly valuable in notebooks where switching to a separate debugging interface breaks your workflow
5. Handles Complexity: Elegantly deals with nested and chained exceptions that can be confusing in traditional debuggers

5 License

MIT