On 18 March 2005 20:20, Clemens Ley wrote:
I've had some trouble with the happy error token: What is it actually there for (apart form implementing the Haskell layout rule). I've not been able to find an example where it matters weather I include the error token or not. Can someone help me?
Happy's error token is a restricted version of the error token found in Yacc. It is sufficient to implement the Haskell layout rule, but not to do general error recovery. The reason I restricted it in this way is because I couldn't find a sensible semantics for Yacc's error token. Different Yacc implementations seem to do different things (I found differences between byacc and bison, for example), and using Yacc's error token to implement the Haskell layout rule was unreliable. It would be nice to have some kind of error recovery in Happy too, though. Cheers, Simon