On Sat, 2008-11-22 at 11:33 +0000, Thomas Schilling wrote:
Be careful, though. This only works if there's a single constructor for your exception type. If there are multiple, you should write it like this:
thing_to_try `catch` \(e :: MyErrorType) -> case e of MyError1 _ -> ..; MyError2 _ -> ...
If you write `catch` (MyError1 ...) and a MyError2 is thrown, you will get a pattern match error exception.
Since I am trying to replicate the behavior of errorCalls, I have only the single constructor ErrorCall to worry about. I don't understand though, how this works: I have a predicate errorCalls e@(ErrorCall _) = Just e In the following transcript it seems to work correctly even if something beside ErrorCall is thrown. Passing NonTermination to errorCalls get a type error as expected. Why does it work inside tryJust? *Main> tryJust errorCalls $ print $ [] !! 23 tryJust errorCalls $ print $ [] !! 23^JLeft Prelude.(!!): index too large *Main> tryJust errorCalls $ print $ throw NonTermination tryJust errorCalls $ print $ throw NonTermination^J*** Exception: <<loop>> *Main> errorCalls (ErrorCall "What?") errorCalls (ErrorCall "What?")^JJust What? *Main> errorCalls NonTermination errorCalls NonTermination^J <interactive>:1:11: Couldn't match expected type `ErrorCall' against inferred type `NonTermination' In the first argument of `errorCalls', namely `NonTermination' In the expression: errorCalls NonTermination In the definition of `it': it = errorCalls NonTermination