Sarah Diot-Girard - Debugging as an experimental science | PyData Amsterdam 2024

Treat debugging as an experimental science with systematic testing of hypotheses and validation through experiments

Keep a debugging log to track assumptions, experiments tried, and results - similar to how scientists maintain research logs

Build and maintain a “mental map” of the codebase to help isolate issues and understand interactions between components

Focus on making debugging experiments reproducible and systematic rather than random trial and error

Break down complex problems into smaller, testable pieces using divide and conquer approach

Be explicit about assumptions and validate them systematically - don’t trust anything without verification

Document bugs, solutions and learnings to help prevent recurrence and assist team knowledge sharing

Consider temporal aspects - check when things last worked correctly and what changed since then

Pay special attention to side effects, mutable state, and interactions between components as common bug sources

Use appropriate tools (debuggers, logs, monitoring) strategically but focus on systematic problem-solving approach

Consider debugging as a learning opportunity to better understand the codebase rather than just fixing issues

Write tests to prevent bug recurrence and improve long-term maintainability, focusing on the specific issue encountered