Leo White

> We would end up with two distinct (non-linear) extensions of the same environment in the same function, which sounds confusing. Rather than confusing I would describe this as just...

I don't think those features would really make the code easier to read. They make the binding structure of the code more complicated, and so it requires more thinking to...

Fair enough, but then I think the language extension you are looking for is just structural record types that support punning. There is no need for some kind of `open`...

This bug has just popped up for us again. I think that we're going to go ahead and merge #8900 in our internal compiler to avoid this issue. I think...

> when the program stays almost all the time within at least one level of effect handlers, there is essentially no performance gain as we run the CPS versions of...

I agree with Hugo that the key to performance with this approach is to identify call sites that cannot involve effects and call the direct version there. Using either a...

I like this general idea, but I'm not too sure about using the type of the argument to distinguish it. How about using the label of the argument instead? So...

> Would it be possible to use the ppx mechanism to this purpose? I'm thinking of something like: > > let f [%auto.location loc] = ... You could possibly use...

> I agree, this needs support at the type level. I think Gabriel was suggesting to use extension node syntax, rather than suggesting an actual ppx AST transformer. > Another...

> The fact that it is a type makes the helper easy to write: reexporting happens naturally by unification. > If we were to replace this with extension or attributes,...