cyclomatic_complexity
What it does
Measures the cyclomatic complexity of each function and reports the ones above max-complexity (20 by default).
When a function is above the maximum, the message also lists the expressions in its body that add more than 20% of the score, as a starting point to split it.
Why is this bad?
The score counts the paths through the flow graph of the function. A function with a high score can be hard to read, to test exhaustively, and to change without breaking one of the paths nobody had in mind.
A high score might be due to a large number of if conditions, for loops, or early return() calls, among other things. It generally indicates that the function should be refactored, for instance by creating smaller intermediate functions.
Details
While the way the score is computed tries to match the output of the cyclocomp package (which is also the output of lintr::cyclocomp_linter()), some differences exist:
- a chain of
&&or||adds one path per operator.cyclocompscoresif (a || b || c || d) stop()as 11, this rule as 5 (stop()can be reached via paths, and we add one path if the condition is never true); switch()adds one path per arm, whilecyclocompignores it;- a function defined inside another one is scored on its own and counts as a single node in the function that contains it, while
cyclocompadds all of its paths to the ones of the outer function.
Configuration options
The operation of the cyclomatic_complexity rule can be customised in the configuration file.
Use max-complexity to set the highest score a function is allowed to reach before it is reported. It must be at least 1.
Default values
max-complexity = 20TOML settings
[lint]
...
[lint.cyclomatic_complexity]
max-complexity = 25