Symbols API: different constructor overloads considered equal

[auduchinok]

[dsyme]

@auduchinok Yes, EffectivelySameAs considers these to both be a use of the type name String. I think it's really confusing if a search for uses of the type name doesn't return these instances?

It's possible a search for uses of a particular constructor should be more refined. That would mean EffectivelySameAs is no longer an equality relation though

I don't I think we should track this as an issue since it's by design, and we have a separate discussion going on somewhere about equality for symbols?

[dsyme]

@auduchinok Do you remember which issue we were discussing the symbol equality logic in? thanks

[dsyme]

OK, it is #4658. Let's discuss there

[auduchinok]

@dsyme You were faster than me. :)
This particular case was about constructor overload uses, not resolve to type entities.

[dsyme]

@auduchinok Yes, I see you have a fix for #4658 so I'll reopen and we can use this one

[dsyme]

@auduchinok we would accept a PR to make

1. EffectivelySameAs(ctor-symbol) only return other uses of that exact consrtructor
2. EffectivelySameAs(type-name) return uses of both the type name and any constructor

Though I suppose that logic could be implemented from the outside.

Do you think there is a use case for EffectivelySameAs(type-name) returning onlythe uses of the type name?

[auduchinok]

Do you think there is a use case for EffectivelySameAs(type-name) returning onlythe uses of the type name?

Yes. I think some features may rely on fact that two symbols are really the same (which is not true for type names and constructors).
For example, this behaviour is one of the reasons why we use ReSharper search in Find Usages instead of FindUsesOfSymbolInFile.