Allow elements to be passed in to enable selective parsing

[leviathan747]

Selective parsing increases performance significantly when using a small bit of an external project. It also enables pulling in a dependency project for parsing without pulling in the whole dependency tree if the part you need does not rely on all the other dependencies.

[coveralls]

Coverage decreased (-0.8%) to 79.217% when pulling 67fd993 on leviathan747:selective-parsing into 2f740fe on xtuml:master.

[leviathan747]

TODO:

• support package following package references
• handle case when parent name is a prefix of a different parent element

[john-tornblom]

Python has a somewhat unexpected behavior with mutable default values:

def append(number, number_list=[]):
    number_list.append(number)
    print(number_list)
    return number_list

append(5) # expecting: [5], actual: [5]
append(7) # expecting: [7], actual: [5, 7]
append(2) # expecting: [2], actual: [5, 7, 2]

Perhaps it does not matter in this case, but I usually do it like this instead:

def prebuild_model(metamodel, parent_paths=None):
    parent_paths = parent_paths or list()
    ...

Or you could use a default tuple() value instead, which is immutable.

[leviathan747]

I think I will change it to have no default since the option parser will handle passing a default value if the user does not provide one.

[leviathan747]

Actually, I will go with your suggestion since maybe that is more expressive when reading the code and does not assume the utility is run from the command line...

[leviathan747]

Package references are supported. Now the model_paths function returns a list of all possible paths to an element (including through package references).

I also changed the implementation so that paths are represented as lists of individual segments instead of strings. This solves the weakness with using startswith that I mentioned before.

[john-tornblom]

I am still not entirely sure what scenarios this PR cover, and I believe further documentation (on scenarios in /doc) is needed. However, before we commit time on review to identify missing corner cases, is this feature still relevant and desirable?

[leviathan747]

I'm going to close this one for now and postpone work on it. It will be tracked further here: https://support.onefact.net/issues/12251