On Tue, 2009-12-29 at 14:47 +0000, pbrowne wrote:
Hi, I am studying the underlying semantics behind Haskell and to what degree those semantics are actually implemented. I need to clarify what a *type synonym* actual means in relation to Haskell's logic (or formal semantics). I used the following type synonym:
type Name = String getName(n) = n
Don't use parentheses. You can think it looks like function call oin other languages but it is not. (n) is a "one-element tuple" i.e. n so if you have function with 2 arguments f (n, k) will not work (it will apply one argument which is a tuple). usually it is written as: getName n = n or getName = id
I checked the types with two tests: -- test 1 :t "ww" "ww" :: [Char]
-- test 2 :t getName("ww") getName("ww") :: Name
Obviously I get two different types. In the case of the function Haskells type system seems to pick up enough information to determine that “ww” is a Name. But I am not sure what is happening with the literal "ww" in the first test.
Pat
ghci> :t getName "ww" getName "ww" :: [Char] ghci> :t getName "ww" :: [Char] getName "ww" :: [Char] :: [Char] ghci> :t (getName "ww" :: [Char]) :: String (getName "ww" :: [Char]) :: String :: String ghci> :t ((getName "ww" :: [Char]) :: String) :: Name ((getName "ww" :: [Char]) :: String) :: Name :: Name it works only because [Char], String and Name are synonimes (:: works differently then () cast in C/C++ where (int)((double)((int)0)) is legal). As far as compiler/interpreter is concerned it is indifferent on the exact naming. However it tried to guess what name You would like to have. For example: type FullPath = String type RelativePath = String expand :: RelativePath -> FullPath expand = ... it is much more clear what you get (FullPath) - not just the raw type. Of course it will not stop you when you state expand . expand as it is rather decorative thing. If you want to create new type you need to write: newtype Name = Name String Regards