ReviseAlgo Logo

Debugging & Profiling

timeit Module

Accurately measuring small code snippets — timeit for micro-benchmarks, comparing approaches, and understanding Python performance characteristics.

Interview: Performance analysis — knowing how to benchmark code accurately is essential for optimization.

Last Updated: June 12, 2026 • 7 min read

The timeit module provides accurate timing for small code snippets by running them many times and disabling garbage collection. It's designed for micro-benchmarks — comparing different ways to do the same thing. For larger-scale profiling, use cProfile instead.

Usage Modes

  • timeit.timeit(stmt, setup, number) — returns total time for N executions
  • timeit.repeat(stmt, setup, number, repeat) — runs multiple times, returns list of times
  • timeit.Timer(stmt, setup) — Timer object for more control
  • Command line: python -m timeit "expression"
  • Use min() of repeat results — best time is most representative

Best Practices

  • Use repeat and take the minimum — eliminates system noise
  • Put setup code in the setup parameter — don't include it in timing
  • Use number large enough — aim for at least 0.1 seconds total
  • Compare relative performance — absolute numbers vary by system
  • Be aware of caching effects — first run may be slower (JIT, file caches)

timeit vs Other Tools

  • timeit: Micro-benchmarks for small expressions/functions
  • cProfile: Function-level profiling of entire programs
  • time.perf_counter(): Manual timing with highest-resolution clock
  • line_profiler: Line-by-line timing within functions

Interview Insight

timeit disables garbage collection for accurate micro-benchmarks. Use repeat() and take the minimum to eliminate system noise. It's perfect for comparing approaches (list comprehension vs map vs loop) but not for profiling entire applications — use cProfile for that.

Use Cases

Comparing algorithms — which approach is faster for a specific task

Optimization verification — confirming that changes actually improve performance

Python idiom comparison — quantifying why certain patterns are preferred

Library comparison — measuring different libraries for the same operation

Quick performance checks — fast micro-benchmarks during development

Common Mistakes

Using time.time() instead of timeit — time.time() includes GC pauses and system noise

Not using repeat — single runs are affected by system activity

Including setup in timing — put one-time setup in the setup parameter

Too few iterations — ensure total time is at least 0.1s for meaningful results

Drawing conclusions from micro-benchmarks — they don't always reflect real-world performance