bugfix: make namespace directive not churn unique types

[mitchellwrosen]

Overview

Fixes #5489

This PR tweaks decl parsing to delay assigning a unique type guid until after the namespace pragma is applied. Previously (as seen in first transcript commit), one could accidentally be assigned a new unique type guid in the presence of a namespace directive.

Test coverage

I've added a transcript that demonstrates the fix.

[aryairani]

@mitchellwrosen

💥 annotations.refs.Test annotations for types with effects.typechecked file

[mitchellwrosen]

This change is due to the fact that our parser has two different error constructors that represent the duplicate type names. This PR performed only minimal unification (switching one over to the other), but there's a little bit more cleanup to do, namely delete the other one entirely.

[mitchellwrosen]

@aryairani Out of expedience I decided to leave the UniqueName generation alone, since I realized we could reuse the namespace field in the parser state that was already there for a different purpose, but I can make a separate ticket for refactoring it to be the gen -> (Bytes, gen) that we talked about if you'd like.

[aryairani]

@aryairani Out of expedience I decided to leave the UniqueName generation alone, since I realized we could reuse the namespace field in the parser state that was already there for a different purpose, but I can make a separate ticket for refactoring it to be the gen -> (Bytes, gen) that we talked about if you'd like.

I read over the code but I didn't get the new strategy. If you're convinced it will work, that's fine; if you want to discuss it let me know.

[mitchellwrosen]

@aryairani Yeah I'm confident it will work.

When we query the parser environment for whether a unique type already exists for a particular name, we prefix the given name with the current namespace directive, if any.

It's really that simple, I didn't think of this solution right away though. :(