Misc. software package metrics for python:
- (Ca) Afferent coupling -> number of modules/packages that depend on a module/package
- (Ce) Efferent coupling -> number of modules/packages that are dependencies of a module/package
- Instability -> Ce/(Ce+Ca)
- Abstractness -> ratio of abstract classes in a module/package
- Distance to main sequence -> distance of module/package to the line A + I = 1
It also generates a PlantUML diagram of the analyzed project! You can easily visualize this diagram e.g. on https://www.planttext.com/
Example usage:
python3 -m coupling_metrics --project_root tests/example_project --package_name my_module
This will produce a report with metrics like:
------------------
Distance to main sequence:
------------------
my_module.lib3 -> 1.0
my_module.lib4 -> 0.6666666666666667
my_module.lib2 -> 0.5
my_module.lib1 -> 0.0
-> average: 0.54
And a PlantUML diagram like:
Note that in order to implement these metrics, which are mostly related to
object-oriented programming, I have assumed that every python module file
is a class by itself.
Abstractness is calculated as abstract/(concrete + abstract) where
abstract is the amount of classes that have at least one explicit
@abstractmethod decorator, and concrete is the amount of concrete
methods defined in the module, or classes that do not have abstract methods.
So in order for one module to count as 100% abstract, it does not have to
declare extra methods outside the scope of the abstract classes defined in it.
Some features that are still not implemented:
- Allow analyzing references to packages outside of the considered namespace, with configurable levels for each such external package.
- Draw a diagram of the position of each analyzed module in the main sequence line.
See this nice blog post for context. This tool will print warnings whenever the instability of a dependency is bigger than that of its dependant.