> >> variables. In other words, if we specify the constraint x < y it > >> looks like either x1 < x2 or x2 < x1 is sufficient to satisfy arc > >> consistency. In other words, i think we should always use the > >> symmetric closure of the constraint. > > > >> Does this sound valid to you? > > > > Unfortunately not. There are no values that can satisfy x1<x2 and x2<x1. > > If this was the case all csps would have no solutions > > I meant x1<x2 OR x2<x1 > > So symmetric closure was the wrong term for me to use. > > But I think we want to be using intersection of the relationship with > its inverse. > > Does that sound right to you? >
Sorry I missed the or. Unfortunately not, I mean you can have a constraint like that if you want but you don't have to have to in general. I think we're dwelling on an implementation detail. They (wikipedia) must just have the < constraint propagating both ways. My brain is fried but I did hack together an ugly search procedure. You can try it out on a sudoku puzzle if you want. For some reason I couldn't generate it using the code you gave me. http://pastebin.com/2zPB4DBA Maybe we can update the printf docs to say: load 'format/printf' http://www.jsoftware.com/help/jforc/input_and_output.htm Cheers, Mike ---------------------------------------------------------------------- For information about J forums see http://www.jsoftware.com/forums.htm
