Template Haskell

[tfausak]

The README says: https://github.com/ocharles/weeder/blob/66fbba0523bf93abf30323f51670366d577e4f98/README.md#template-haskell

Weeder is currently unable to parse the result of a Template Haskell splice. If some Template Haskell code refers to other source code, this dependency won't be tracked by Weeder, and thus Weeder might end up with false positives.

I couldn't find an open issue for this, so I figured I'd make one.

Obviously this is a known limitation. I'm curious if it's essentially impossible for Weeder to support, or has it just not been implemented yet? I suspect it's more the former, but I can't find any discussion about it.

[ocharles]

It's mostly down to what goes in to the HIE AST. I think sometimes for splices it does actually splice in the entire tree, which would mean weeder would be able to look through TH. I think we need some examples of where weeder currently fails, and then we can try and see if it's weeder or GHC that needs to change

[jkeuhlen]

After testing with GHC 9.10.1, it seems like things work much better. I don't have any definitive tests for full coverage, but I think this might "just work" with the recent changes to HIE files.

[jkeuhlen]

To get more specific for tracking purposes, one false positive class that remains in GHC 9.10.1 is QuasiQuoter usage.