What
Break out logic from _get_grouped_contribution so that contributions and refactors are easier. Measure is likely one of cyclomatic complexity, so consider adding the C901 lint as well for a simpler DoD.
https://github.com/lvgig/model_interpreter/blob/0ecb43ce1e40ac7160527a56953419dec40c1a4b/model_interpreter/interpreter.py#L217
Why
Came up in context of review #2; it is very hard to reason about what is happening here.
Generic dispatch is a great OOP means of breaking out logical complexity from if-elfif-else chains and decrease the amount of cognitive bandwidth required to process code.
How
What are the types of model covered by the function? For each, create an implementation of 'getting grouped contributions' and then use a generic dispatched to choose which is the correct method. Then you only need to figure out how to tell the 'chooser' (aka the generic) how to choose the right kind - I'd base this off the model type passed in or potentially the model attributes from sklearn etc.
Links:
What
Break out logic from
_get_grouped_contributionso that contributions and refactors are easier. Measure is likely one of cyclomatic complexity, so consider adding the C901 lint as well for a simpler DoD.https://github.com/lvgig/model_interpreter/blob/0ecb43ce1e40ac7160527a56953419dec40c1a4b/model_interpreter/interpreter.py#L217
Why
Came up in context of review #2; it is very hard to reason about what is happening here.
Generic dispatch is a great OOP means of breaking out logical complexity from if-elfif-else chains and decrease the amount of cognitive bandwidth required to process code.
How
What are the types of model covered by the function? For each, create an implementation of 'getting grouped contributions' and then use a generic dispatched to choose which is the correct method. Then you only need to figure out how to tell the 'chooser' (aka the generic) how to choose the right kind - I'd base this off the model type passed in or potentially the model attributes from sklearn etc.
Links: