llogiq
llogiq
I don't think that will be a problem for our schema – Generics aren't likely to change much (until we get HKTs that is – but then I'll gladly extend...
Yes. Plain `Call`s are always statically dispatched, whereas `MethodCall`s may be dynamically dispatched (if `self` is a trait object).
Yes. Luckily, a) dynamic dispatch isn't all that common in Rust and b) for most intents & purposes it's sufficient to look up all possible impls and see if their...
@cramertj Just a heads-up – I renamed the project to metacollect due to some people feeling uneasy about the old name. No need to stoke emotions.
Sorry, not yet. I'll need to catch some sleep first – maybe tomorrow.
This has languished for some time. Sorry for the delays. There have been a few developments in the meantime: - We don't need to track type contents, because rustc already...
With "type contents" I mean the composition of structs and enums. I'm wary of monomorphization, because a) there are existential types and trying to monomorphize can give us an exponential...
With into() and detef() calls, we record the exact receiving types, and thus can pinpoint the trait impls. As long as all dependencies (and yes, that probably includes `std`) are...
OK, we don't analyze FFI code. We only analyze Rust code for the crate and all dependencies. Since we expect metacollect to run on the whole dependency tree, this should...
That's why we _collect_ (as in "meta_collect_") the generics and substitutions. Say, metacollect records that `Person::new(_)` is generic over a type with a trait bound of `Into`. Later it records...