When comparing Ceylon vs Idris, the Slant community recommends Ceylon for most people. In the question“What are the best (productivity-enhancing, well-designed, and concise, rather than just popular or time-tested) programming languages?” Ceylon is ranked 25th while Idris is ranked 28th. The most important reason people chose Ceylon is:
The compiler prevents you from using a potentially null variable, unless you check it is not null. Ie. it forces you to check a potentially null value before using it. The type system is strict, but flexible, allowing union and intersection of types, covariant and contravariant types, reified types, etc. Type inference and union types allows a dynamic programming style, close of JS spirit.
Ranked in these QuestionsQuestion Ranking
Pro Strong static typing, null safe and flexible, almost dynamic type system
The compiler prevents you from using a potentially null variable, unless you check it is not null. Ie. it forces you to check a potentially null value before using it.
The type system is strict, but flexible, allowing union and intersection of types, covariant and contravariant types, reified types, etc.
Type inference and union types allows a dynamic programming style, close of JS spirit.
It brings type safety to JS, allowing to define interfaces to existing JS APIs, yet using the
dynamic keyword for flexible calls in the JS ecosystem.
Pro Excellent IDE support
Ceylon has reified generics, so it doesn't loose the type of collections at runtime. This makes autocompletion, debugging, etc. first-class. The Eclipse plugin makes it a full-fledged Ceylon IDE, and an IntelliJ IDEA plugin is in the works.
Pro Great tutorial
Gavin King, main author of the language, has a great, clear technical writing style, making understandable difficult concepts like variance or sound type system.
Pro Try it out in the browser
It has a Web IDE: http://try.ceylon-lang.org/ with impressive demos: http://try.ceylon-lang.org/?gist=bd41b47f325b6d32514a so you can try it without installing anything, and see the JS generation / interop in action.
Pro Excellent documentation
The language specification is very complete and up to date; also, the language module is very well documented.
Pro Easy to learn even if you don't have prior programming experience
Ceylon is indeed fairly easy and readable. Of course those ones who know OOP and a bit of functional programming concepts will feel almost at home right from the start.
Pro Generate HTML
HTML generation is supported right in the SDK.
Pro Same code in backend and frontend
Pro Full dependent types
Idris not only has support for type classes, but is a fully dependently typed language, giving you the full power to statically verify your code.
Pro Domain driven design and type driven development
Because of full dependent types in Idris, the programmer can focus more on modelling the domain with types and waste less time fixing common bugs that the type checker will catch. Dependent types help apply type driven development and a lot of code auto generation, making the compiler and type checker an ally in developing working software instead of just getting in the way.
Con Lack of physical or electronic books
We should hope Red Hat or anyone interested would take the time and write one. That would strengthen the maturity of the language, but Ceylon is rapidly developing which can make the author's efforts futile because his or hers work will become obsolete soon.
The second hindrance is, of course, popularity of the language which can't give much to the pockets of the author (however, Dart's unpopularity at start didn't prevent it to have a lot of printed material, but that's Google's child, we know).
Con Currently has large runtime
Ceylon 1.2 needs a language runtime of 1.55 MiB, and the Collection library adds another 370 KiB. That's a lot for the Web...
Now, this has to be put in perspective: if you use Ceylon to make a web application, these files will be loaded once, then cached by the browser (that's not casual browsing).
Moreover, most servers compress such resource, and the numbers become respectively 234 KiB and 54 KiB, which is more reasonable...
Con Weaker type inference
As type inference is undecidable for dependently-typed languages, Idris cannot offer the full type inference that Haskell supports, and so more type annotations will be needed.
Con Different semantics from Haskell
Idris, while similar to Haskell, has strict semantics, which may cause some confusion if your backend is done in Haskell. If using Idris, it would make sense to do the backend in Idris as well, if not for the fact that Idris currently has fewer libraries available for web development than Haskell.