Demangled names can exceed YAML's 1024 character key limit about stoat

[fundamental]

Oh goodness boost, that's a long type signature.

I'd side with just truncating the signatures as the YAML files do make it easier to look at the output manually if need be. To be 100% correct, yeah this would require either a change of the format s.t. the names aren't used directly and they're stored in another table, but I'm hesitant to go for that. I'm pretty sure one of the iterations used that approach (perhaps sfpv as I built tables to llvm::clang::someOtherNamespace::Function pointers rather than using the stringified names in parts of the code) and debugging issues with that was a giant pain.

[dsacre]

In boost's defense, map<string, vector<string>> alone is around 250 characters when fully spelled out. The function actually takes two of those, and each argument appears (only) twice in the type signature... :)

Does this look ok to you?
dsacre/stoat@de21782

I take it the shortened names only affect what would appear in the output/graph, and stoat uses mangled names for everything else?

[fundamental]

That might be sufficient in practice as llvm does replace lengthy templated type names with shorter versions, but it might make sense to just limit some of the other printfs to this limit (when they're getting used as keys).

Shortened names should only show up in the graphviz renderings, though they're also used in pattern matching for the white/black lists. They can't be used at the YAML file level as collisions happen too easily.

[dsacre]

Here's a more complete patch. Is this what you had in mind?
dsacre/stoat@e9b527a

[fundamental]

Yeah that looks right. I'll try to merge that in this afternoon.

[fundamental]

Merged. I've also added you to the contributor list on github, so you should have commit access as well. thanks 👍