type system in erlang?
Thu Dec 15 23:40:39 CET 2005
> The basic problem with type systems is that there is no single one
> that will satisfy all wishes. What sort of type system would you
> want to see?
The sensible definition of the type of a value is "everything the
compiler knows about a value at the time the compilation is done". The
simple way to allow the user to communicate extra information to the
compiler is with some sort of assert system.
I would note again that there is more to be gained by doing further
optimization at run time than by squeezing every drop of optimization
during traditional "before any input" compilation. I would also like to
completely reject the idea that you shouldn't do optimizations that will
mess up benchmarks. Benchmarks should use the value created to have an
external side effect, otherwise they are not well defined.
More information about the erlang-questions