Radon

Radon computes code metrics for Python: cyclomatic complexity per function with a letter grade, raw line counts, Halstead metrics and the Maintainability Index per file [1].

The quality question it answers: where is the code hardest to understand and change? It ranks; it does not decide. A grade C function is a candidate to look at, not a defect.

Run for this card with radon 6.0.1, October 2026.

Install and quick start 

poetry add --group dev radon
poetry run radon cc -s -a src/       # complexity per function, with scores and the average
poetry run radon mi -s src/          # Maintainability Index per file
poetry run radon raw -s src/         # LOC, SLOC, comments, blank lines
poetry run radon cc -s -n C src/     # only functions graded C or worse

Reading the output

On the parking-fee function from the pytest card and the receipt function from the flake8 card (constructed examples):

$ radon cc -s -a src/
src/parking/fee.py
    F 5:0 fee - A (4)
src/parking/receipt.py
    F 5:0 receipt - C (11)

2 blocks (classes, functions, methods) analyzed.
Average complexity: B (7.5)

Read a line as: F (function; M method, C class), line 5, column 0, name, grade, score. The complexity grades are A 1–5, B 6–10, C 11–20, D 21–30, E 31–40, F 41 and above.

$ radon mi -s src/
src/parking/__init__.py - A (100.00)
src/parking/fee.py - A (63.80)
src/parking/receipt.py - A (51.49)

The Maintainability Index is per file, on a 0–100 scale; A is above 19, so almost every real file is an A. Compare files with each other, and watch the trend over time, rather than reading the grade. The 63.80 is for fee.py as printed on the pytest card: after the mutmut card adds one comment, # pragma: no mutate, the same file scores 80.91.

Common mistakes

  • Expecting flake8 and Radon to agree. On the same function, Radon scored 11 and flake8’s C901 scored 10. Radon adds one for each boolean operator (minutes > m and member); mccabe, which flake8 uses, does not. A limit of “10” therefore means different things in the two tools.
  • Averaging away the problem. Average complexity: B (7.5) hides the one C function. Gate on the worst function (-n C), not the average.
  • Treating MI as a measurement of maintainability. It is a regression formula from the 1990s, sensitive to comment lines; the Maintainability Index page shows how much one comment moves it.

In CI

Radon itself always exits 0. To fail a build on complexity, use flake8’s --max-complexity, or the companion tool xenon (poetry add --group dev xenon), which wraps Radon with thresholds. On the example, xenon --max-absolute B src/ (fail if any single block is graded worse than B; xenon 0.9.3) printed ERROR:xenon:block "src/parking/receipt.py:5 receipt" has a rank of C and exited 1.

Go deeper

References

  1. M. Lacchia,
    “Radon Documentation: Introduction to Code Metrics and Command-Line Usage.” 2023. [Online]. Available: https://radon.readthedocs.io/

Disclaimer: AI is used for text summarization, polishing and explaining. Authors have verified all facts and claims. In case of an error, feel free to file an issue.

This site uses Just the Docs, a documentation theme for Jekyll.