Make rasa data validate check for duplicated intents, forms, responses and slots when using domains split between multiple files

[alwx]

Proposed changes:

• fix rasa data validate should fail if conflicting slots are created in different domain files #9555

Status (please check what you already did):

• added some tests for the functionality
• updated the documentation
• updated the changelog (please check changelog for instructions)
• reformat files using black (please check Readme for instructions)

[ancalita]

Looks great 👍🏼
I left mainly some documentation related suggestions.
I was also wondering if we want to raise InvalidDomain when duplicates are not None in from_dict? What happens now if you train with a Domain with conflicting slots for example?
This could be purely out of scope for the issue which refers only to rasa data validate, but it would be just an addition of a few lines of code 🤔

[ancalita]

Looks good 🎉
I'd still explore with Ty and the rest of the squad looking at the other Domain.merge related issues whether a holistic approach is required for this entire umbrella + if to raise InvalidDomain whenever duplicates are found?

[joejuzl]

Did we consider just raising an error/warning here so that we don't need to store the duplicates on the class?

[alwx]

The idea was to show this information only when running rasa data validate, as it's written in the description of this issue. We can update it to show the warning every time the domain is getting merged but in this case I don't see the point in doing any changes to rasa data validate and even link this issue to the validator which means the whole task will be different.

[alwx]

Maybe @TyDunn can take a look at it and say which approach we prefer.

[TyDunn]

Does the approach that @joejuzl describe happen when you run rasa data validate too?

[alwx]

@TyDunn If we do it the way @joejuzl proposes then the warnings/exceptions will be raised every time the domain is getting merged (e.g. when just running rasa train or literally any other command that uses domain), not only when doing rasa data validate

[joejuzl]

For context we do already have quite a lot of domain validation that happens just from loading it e.g. _check_domain_sanity in the __init__ looks for duplicates etc.

[TyDunn]

@alwx I believe Joe is out for the rest of the year, and I am not going to be able to gain enough context atm. Can you make a judgement call here yourself?

[alwx]

@TyDunn I think it makes sense to keep this one as it as and maybe create a new issue because what Joe was talking about makes the scope much bigger

[alwx]

@TyDunn please approve to merge this one.