[pyreverse] Filtering out specific classes from diagrams

[LouisDesdoigts]

Question

Hey there, firstly pyreverse is awesome so thanks!

I just wanted some clarification on how to customise the output UML diagram from pyreverse. Essentially I'm automatically generating a UML diagram for each class in my software to be hosted in the docs. There are a few things that I would like to remove from the diagrams.

Here is an example diagram from my software, generating by running pyreverse -c dLux.layers.optical_layers.Tilt -o html --colorized dLux

I have a few things that I would like to change about this.

1. How can I remove the 'has a' that adds the 'Array' class to the diagram. This is just not useful information for my purposes. Is there a way in general to only show parent classes and not other classes that are held by it?
2. Some classes have their attributes doubled. In this example the 'angles' attribute is displayed twice. What would be the reason for this, and how can I avoid this?
3. Every class in my software inherits from a class called Base in an external package. Is there a way to only display class inheritance inside of my software?

I've had a look through the options and had a play with different arguments but can't seem to make any headway. Any help is greatly appreciated!

Documentation for future user

Inside the main pyreverse docs page.

Additional context

No response

[DudeNr33]

Hi @LouisDesdoigts,

happy to hear you like Pyreverse!

To your questions:

1. You may try using -s0 for that.
2. This may happen if you for example declare an attribute at the class level and then assign a value to it in your __init__. Thanks to @nickdrozd this should be fixed with the next release of 2.17.x: [Backport maintenance/2.17.x] Fix duplicate fields Pyreverse bug #9011
   A minimal example for what could this could happen can be seen from the functional test file for the fix.
3. The problem is that the -c option implicitly sets the options -ASmy. You may try experimenting with setting -aX with X being the number of ancestors to show, but I am not sure how this will interact with the implicit -ASmy. Actually this also applies to the first question.

I don't know when I will have time to try it myself, so if you want to go ahead and give it a try with -s0 and -a1 (or something) I would be interested in the results!

[LouisDesdoigts]

Hey @DudeNr33,

Thanks for the quick response, those tips works perfectly!

Running: pyreverse -c dLux.layers.optical_layers.Tilt --output html -s0 -a2 --colorized dLux outputs exactly what I want!

I'm also glad to hear the parameter duplication will be fixed shortly too!

I have one other question too, using the -sX flag counts 'downwards' from the target class (apologies for my fumbling terminology), which is not a problem as I can reverse engineer the desired depth. However ideally there would be an option to count 'upwards', in the sense that you can exclude the N most ancestor classes, since this would be a static value, rather than dynamic in my use case. This would be kind of like -s -2 to index in the other direction, if that makes sense. Would something like this be possible?

Thanks again for the help, and let me know if you'd be interested in me expanding on the pyreverse section of the docs with a small example of this use case, I'd be more that happy to!

Cheers

[DudeNr33]

Can you give an example of how this 'downwards' counting should behave? I am not certain if I understand what you mean, as '-s' controls how many associations are shown, but in the text you talk about ancestors.

Expanding the docs is always welcome! I can imagine having a Example use cases section in the pyreverse page. ❤️