Add RawType property to FunctionParameter type

[suaviloquence]

I think in order to write correct witness functions in c-s-c we need to be able query the function signature more granularly than just the signature property. Adding the ability to query the type of each parameter, in addition to the already-existing GenericParameter property, should be enough to do this (but I'm not 100% sure). I can implement this, just opening an issue to make sure this is the right direction to go in.

[obi1kenobi]

You seem to have a vision in mind here: could you share some more of it? I haven't thought about the problem space nearly as much as you, and I'd love to hear more of your thinking on it. The idea seems reasonable in general, though I'm not super thrilled with the current `RawType` vertex type design — I hacked it in as a workaround originally and it hasn't seen any more love since then.

…

On Mon, Dec 23, 2024, 3:10 PM Max Carr ***@***.***> wrote: I think in order to write correct witness functions in c-s-c we need to be able query the function signature more granularly than just the signature property. Adding the ability to query the type of each parameter, in addition to the already-existing GenericParameter property, should be enough to do this (but I'm not 100% sure). I can implement this, just opening an issue to make sure this is the right direction to go in. — Reply to this email directly, view it on GitHub <#687>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAR5MSUY5KABXCXOXOYW7WT2HBU4NAVCNFSM6AAAAABUDTG4ZOVHI2DSMVQWIX3LMV43ASLTON2WKOZSG42TMNRVHAYTSMI> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[suaviloquence]

Current problem: we need to construct a witness function header (e.g., for function_missing, it needs to take parameters that have the same arguments/generics/generic bounds as the removed function). We can use signature, but it's kind of hacky, as the name of the removed function is hardcoded in the function signature, so we would have to print out a function like:

fn removed_function<'a, T>(x: &'a t, y: Cow<'static, str>) -> T::Item
where T: 'a + Iterator
{
    path::to::removed_function(x, y)
} 

which works, but may be confusing (the name of the witness function is the same as the removed function, and the return value is not technically necessary for a minimized withess?) -- another idea is to allow specifying a function name with signature, maybe?

Future problems: writing witnesses for functions that change arity/parameter types?

I was also under the assumption that RawType would eventually become (or be superseded by) a Type class that has the stringified name but also supports other queries (equality, satisfies trait bound, etc. - but this is for the future)

[obi1kenobi]

All sounds good to me, go for it! Also, if you have opinions or ideas on how to design a better Type vertex type, I'd love to chat about that sometime :)

I'd be very curious to see what the witness query and template would look like for function arity change / parameter type change lints! I wonder if we might run into more issues there.