> I think I'd be more interested in Crystal if it was easier to integrate into existing ruby code bases via something like the FFI.
That would be very nice, I agree, though I've always found the FFI interface an under-documented pain so I'd probably end up just using Crystal or using some other way to interoperate, like pipes or HTTP.
On Sorbet, I just hate the look of it. I know it's subjective (possibly that's Ruby's own fault for being so delightful to look at, I'm spoiled!) but I can't abide the syntax. Crystal's is much nicer, and has some extra sugar sprinkled on top for the moments you're most likely to be using it (like external names in method signatures[1] or providing init args) that win out for me.
Sorry about the formatting, I can't work out HN's markdown :/
# Ruby
# from the Sorbet intro
extend T::Sig
sig {params(name: String).returns(Integer)}
def main(name)
puts "Hello, #{name}!"
name.length
end
# Crystal
def main(name : String) : Int
puts "Hello, #{name}!"
name.size
end
puts main "you"
That's so much nicer to my eyes, and you don't need most of it as it will be inferred.
Initialisation is nicer, too, (unless Ruby has been updated in this regard too? As I wrote, I use it much less now)
# Crystal
class A
getter name : String
def initialize(*, @name); end
end
a = A.new name: "Hacker News"
puts a.name
or without the getter and keyword arg:
# Crystal
class A
getter name
def initialize(@name : String); end
end
a = A.new "Hacker News"
puts a.name
Whatever suits.
# Crystal
# Example of external names
def increment(value, by)
# OK, but reads odd
value + by
end
def increment(value : Int, by amount : Int) : Int
# Better
value + amount
end
puts increment 1, 2
but those types are easily inferred so it's actually:
# Crystal
def increment(value, by amount)
# Even better
value + amount
end
I could go on because now I'm comfortable with it, it looks much better. I feel the way moving from Perl and C# to Ruby felt.
[1] https://crystal-lang.org/reference/1.3/syntax_and_semantics/...
Edit: formatting, of course