Recipe disambiguation of recipes from other locations

phenaproxima → closed merge request !133

Nice find. Definitely should happen. But when there are recipes with the same name but different paths that's quite interesting...

Why is that interesting, though?

The human-readable name is just that -- a human-readable name. In combination with the type key, there's nothing obviously special about it. It can be as descriptive -- or, more likely, undescriptive -- as the recipe author wants. There are no guidelines, and certainly no rules, about uniqueness (if there, we'd need to explicitly validate the name's uniqueness, at the very least).

Recipes are really only uniquely identified by their path. Since they are not extensions, it would be legal for them to have the same directory name (machine name). Since the code in question needs to identify recipes unambiguously, the path seems like the cleanest way to do that. (We could also hash the contents of recipe.yml but that might be overkill.)

It's interesting because when we list recipes in the project browser what's going to happen?

This looks like what we agreed on.

The only thing I think is commit-blocking here is a strongly explained reason why discovery is not centralized, and why core is special-cased. Without that, someone (maybe one of us) is liable to come back to this code in a couple of years and try to add dynamic path resolution or other stuff which would break the intention here.

We're intentionally limiting it as a reasoned design decision, so we need to be absolutely 1000% crystal-clear that the decision is explained.

I've resolved all the MR threads, apart from two - one of which I don't think we should do - the other is actually unrelated and should get it's own issue.

Alright, I think this solves the problem reasonably well. Let's ship it.

Committed and pushed 6895717bf08 to 11.x and 8af521051eb to 10.3.x. Thanks!

Automatically closed - issue fixed for 2 weeks with no activity.