On Monday, November 14, 2011, Luke Vilnis wrote: > Has anyone taken a look at the upcoming language > Ceylon<http://in.relation.to/Tutorials/IntroductionToCeylon>? > It feels to me like a pretty aggressively blub-y Java clone, but it has two > features that seem like they're ripped straight out of Typed Racket. >
I had seen a little bit about Ceylon, but it's good to know that other people are finding similar ideas useful. > It looks like the language includes a basically verbatim copy of TR's > dotted type parameters, which it calls "sequenced type > parameters<http://in.relation.to/Bloggers/IntroductionToCeylonPart8#H-RepresentingTheTypeOfAFunction>". > For example, the supertype of all tuples is written > Tuple<T,P...><http://in.relation.to/Bloggers/TuplesOrNot> > . > In the interests of full disclosure, they're probably copying C++ templates here, which have a similar but less expressive design to Typed Racket's variable-arity polymorphism. > It also has untagged union > types<http://in.relation.to/Bloggers/IntroductionToCeylonPart5>(written like > Int|Bool) which can be deconstructed with switch-case statements, sort of > like a watered down version of occurence typing: > > switch (foo) > case (is Bool) { > // this type checks > local isTrue = foo || true; > ... > } > case (is Int) { > // so does this! > local x = foo + 1; > ... > } > One interesting note here is that this almost certainly compiles to the usual `instanceof' and cast in the JVM bytecode, which is then analyzed to eliminate the cast by the JIT compiler. sam th -- sam th [email protected]
_________________________________________________ For list-related administrative tasks: http://lists.racket-lang.org/listinfo/users
