Fully Qualified Type names for the Source Generator (to avoid namespace clashes)

[thomhurst]

I've got a solution where I'm getting compilation errors because of namespace naming conflicts. Fully qualifying the types in generated code (so they begin with global::) should fix this.

[romfir]

shouldn't we also update similar usage in line 80?

[thomhurst]

Nope namespaces don't support global:: - Only types do

[romfir]

right, sorry I thought it can also be used for namespaces

[romfir]

@thomhurst Please run OneOf.SourceGenerator.AnalyzerTests and fix them (I think they should now fail) and add a test for your specific case (for the error that you encountered), could you also add "global::" prefix to AttributeNamespace and remove "global::" from filtering inside GetSemanticTargetForGeneration?
@mcintyre321 could we possibly run these tests in CI?

[thomhurst]

@thomhurst Please run OneOf.SourceGenerator.AnalyzerTests and fix them (I think they should now fail) and add a test for your specific case (for the error that you encountered), could you also add "global::" prefix to AttributeNamespace and remove "global::" from filtering inside GetSemanticTargetForGeneration? @mcintyre321 could we possibly run these tests in CI?

The comparison inside GetSemanticTargetForGeneration is just with strings, so I have to prepend global:: there. But we're comparing it anyway against a fully qualified type, so we know we're getting the right one.

Also all tests are still passing.

I can look at adding a test though.

[thomhurst]

@romfir They have also run on CI here:
https://ci.appveyor.com/project/mcintyre321/oneof/builds/47938445

[romfir]

Interesting I thought that this test would fails because of the missing namespace in implicit operators

[romfir]

GeneratesPublicClass fails on my machine with (4,11): error CS7000: Unexpected use of an aliased name

[thomhurst]

Interesting I thought that this test would fails because of the missing namespace in implicit operators

Looks like the assertions aren't doing full equivalency. It seems to be doing things like checking no diagnostics, or that certain elements match?

[thomhurst]

GeneratesPublicClass fails on my machine with (4,11): error CS7000: Unexpected use of an aliased name

You haven't changed the namespace to fully qualified have you?

[romfir]

Interesting I thought that this test would fails because of the missing namespace in implicit operators

Looks like the assertions aren't doing full equivalency. It seems to be doing things like checking no diagnostics, or that certain elements match?

They are using SyntaxTree.IsEquivalentTo, perhabs adding "global::" is a trivia difference mentioned in the method description 😄 also the comments seems to be ignored

You haven't changed the namespace to fully qualified have you?

I did 🥲

Could you move this new test to AnalyzerTests? All OneOf.SourceGenerator.Tests should be moved there but I did not have time to do it, perhaps the name is a bit misleading but its a better way to test source generators.

[thomhurst]

@romfir Have added a test to the analyzer project GeneratesClassWithConflictingNamespaces

[thomhurst]

@mcintyre321 Any change of getting this in please? I'm blocked from using the source generator at the moment because of this

[romfir]

LGTM!