On Fri, 2009-01-02 at 10:48 +1100, S. Günther wrote: > > Whether circular or not, sharing values by using a "back pointer" is > > problematic for any update. Why not use a zipper instead? > I looked into zippers before and the problem I had was that they never > really matched the structure which I needed and which led me to think > about this whole knot tying thing again.[1] The things I read about them > always assumed either a list like (i.e. linear) or a tree like (i.e. existence > of a root) structure on the type to be plugged into the zipper. Now I know > that this is not necessary and there are references to a generic zipper > implementation but the thing is that I only found the source code and > decided to first look into the knot tying thing before opening another > can of worms with delimited continuations since I already spend too > muchtime thinking about that stuff.[2] So if anyone has pointers to a > good generic zipper explanation I would be thankful. Note that I don't > need full blown error handling. I would first like to understand the basics. > Now I think I came to the conclusion that locally updating a tied knot in a > pure way really is hard and that it's not me just not seeing some obvious > solution. So I just have to decide whether to use IORefs/Vars (clunky) > or to implement zippers for the structure I need (probably toohard for me). > Anyways thanks for all the answers from which I learned a few unexpected > things and which assured me that my instincts aren't that far off.
Read this http://strictlypositive.org/diff.pdf (again) _______________________________________________ Haskell-Cafe mailing list [email protected] http://www.haskell.org/mailman/listinfo/haskell-cafe
