Which immediately brings back the issue with the UClass approach versus type descriptors that are plain old data.
-
-
Replying to @pervognsen @TimSweeneyEpic
E.g. I shouldn't have to intern a UClass to describe an instantiated TArray with an element type to the rest of the system.
1 reply 0 retweets 0 likes -
Replying to @pervognsen @TimSweeneyEpic
Parametric templates with dictionary passing as the default and monomorphic instantiation as an optimization: easy interop.
1 reply 0 retweets 4 likes -
Replying to @pervognsen @TimSweeneyEpic
Unfortunately, that doesn't describe C++ templates, which is why I'm skeptical a good solution to your problem is possible.
1 reply 0 retweets 1 like -
Replying to @pervognsen
For each specialized template, the compiler could generate a function from type arguments running arbitrary code, right?
1 reply 0 retweets 1 like -
Replying to @TimSweeneyEpic
Hah, I suppose, although that feels like a classic C++ joke ("and then we had to do X") in the making. :)
1 reply 0 retweets 1 like -
Replying to @pervognsen @TimSweeneyEpic
"To avoid bloating the binary with the reified template resolvers, we included LLVM in every exe for JIT compilation."
1 reply 0 retweets 0 likes -
Replying to @pervognsen @TimSweeneyEpic
Incidentally, it's stuff like this that makes me hesistant about Ontic-style free-for-all mixing of terms and types.
1 reply 0 retweets 0 likes -
Replying to @pervognsen @TimSweeneyEpic
Even if you want to offer all that expressive power in extreme cases, you want well-defined, layered, tameable subsets.
1 reply 0 retweets 0 likes -
Replying to @pervognsen
Agreed. Subsets are the best one can hope for when the type system is Turing complete.
1 reply 0 retweets 0 likes
More interesting type systems require theorem provers for quantifier instantiation, so the automatable subset is small.
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.