2

I'm going to translate into UML a C++ project I have been doing for a few days and which I estimate convenient to handle as a model for a way better management. The problem is the following: I tried to define a singleton to store precalculated values, it didn't work and the best solution ended to be declaring a static global scope variable with two functions, one to add data and other to search such data. How can I model functions and variables in a namespace?

I thought of representing the global variable as an object and the functions as interfaces. However I really have no idea of what to do.

Note: I have tried many ways to include those three elements inside a class or a struct but it has been imposible to make them work as intended.

  • Regarding representing global variables and free functions using the class notation: in UML, you can represent static members (as opposed to instance members) by underlining member names (see here), so you can model globals like that. Another thing you can do is to use a custom stereotype (the << ... >> thing) on the class itself, e.g. << static >> or << globals >> to modify the meaning of the element (just explain it in some accompanying text). Stereotypes are there to allow you to extend UML. – Filip Milovanović Aug 12 at 16:50
  • If you need to indicate that some other code depends on that pseudo class, then you can use a generalized dependency (a dashed arrow) - - - - ->, as that doesn't imply that the dependent end holds a reference, just that the names of the global functions and/or the global variable are mentioned somehow. – Filip Milovanović Aug 12 at 16:51
  • @Miguel Somehow related: stackoverflow.com/questions/13663462/… – πάντα ῥεῖ Aug 12 at 18:05
4

UML is not a graphical programming language. Transforming C++ in UML will do no magic in this regard.

Recommendation

  • The first thing that you need to do, is to reconsider your problem and find out what is worth to show visually and what you expect from the visualization.
  • At the same time, if you have a lot of static global variables, your source code encapsulated some internals at the level of the compilation unit. So even if you have no objects and classes, you have somehow a component and could see your source file as component.
  • Now, if you manage to do this, one step further could be to consider the component as an instance of a class. This could allow you to start introducing some OO design into a bunch of free standing jungle of functions.
  • Alternatively, you could consider every freestanding function as a stateless class. But this might lead to an overkill and bring us back to the question if visual programming is really a good idea for your problem.

How UML can help?

If you want to show the flow of control, that starting with some objects gets some behaviors and generates a flow of objects, you may consider using an activity diagram. Functions would be actions:

  • but you might just end up with a kind of more elaborate flowchart diagram : you might express with more work things that could be easier expressed in code;
  • on the other side, viewing a very complex flow may allow to better understand how the parts relate together, and even identify parallelisation opportunities.

If you want to show the structure of your module, and the exposed API, you may consider:

  • a component diagram could help you to abstract the interface and purpose of your different modules or submodules (e.g. compilation units);
  • a composite structure may focus on the inner structure of your modules: the parts may represent named objects (variables). But nothing is foreseen for functions, so you’d need to handle them as a parts as well. I’m not sure that it’s the best thing in your case
  • a class diagram could be used to show how the static globals relate to a group of related functions, especially if these are embedded in their own namespace.

Personally, from what I understand of the partial information on your case, I’d opt for the last approach: even if you would have no class in your implementation, at a conceptual level they seem to be there. And this could then help to develop a more object oriented design, breaking down larger modules in smaller "classes", focusing on their public interface, and making them interchangeable.

| improve this answer | |
  • 1
    TBF people certainly have tried to use UML as a graphical programming lanuage - its is generaly considered to have been a bad idea though – jk. Aug 13 at 7:31

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.