njtynhj

I want to define a function that has a specialized implementation for some types and a general implementation for all other types, with more magic than {-# SPECIALIZE #-} or rewrite rules can get me. (Roughly, I want to switch from IORef to Ptr when possible.) Closed type families look like they might provide a solution: ...but it's unclear how, from here, to build a function special :: Specialized a , or something more or less equivalent. Closed type families seem to work on the type level, but I can't figure out how to do the appropriate thing on the data level. If I were willing to explicitly list all types I didn't want to specialize for, I could write a type class and make an instance for all the other types, but I want to write something that specializes a finite list of types and then does something else for all other types. How can I accomplish something like this? You can accomplish it with overlapping classes. Here's an example for writing nub which cho...

New to Haskell, but I've run into something that's confusing to me, and I can't explain it or find any documentation. If I have the following code: Now, this code works perfectly well The issue comes if I attempt to move a point and assign it to itself = doesn't perform assignment; it equates two things. You are creating a recursive definition by saying that otherpoint and move otherpoint 5 5 are the same thing and are interchangeable. That means when you try to evaluate the call to move , it tries to evaluate otherpoint , which leads to the next call to move , etc. You can't simply "rebind" a variable in Haskell. Instead, use a different name. Let’s look at an example of a self-referential definition that does work: As chepner mentioned, an equals sign in Haskell is is not an assignment, but a definition. It has very different semantics than an equals sign in, for example, C or Java. In general, each variable in Haskell can only be given one definit...