7.9 Distribution

Given a set M, let {\it lowerBound}({\tt M}) and {\it
upperBound}({\tt M}) denote the greatest lower bound and the least upper bound currently known for M. Also define {\it unknown}({\tt M}) =
{\it upperBound}({\tt M}) \backslash {\it lowerBound}({\tt M}).

distribute

{FS.distribute +Dist *Ms}

The vector Ms is distributed according to the specification Dist. The following values for Dist are supported:

  • naive is equivalent to generic, i. e. the default settings apply.

  • generic(order:    +Order   <= order
            filter:   +Filter  <= true 
            select:   +Select  <= id
            element:  +Element <= element
            rrobin:   +RRobin  <= false 
            weights:  +Weights <= {FS.makeWeights nil}
            procedure:+Proc    <= proc {$skip end)

    • Order

      • naive selects the left-most variable.

      • order(sel:  +Sel  <= min
              cost: +Cost <= card  
              comp: +Comp <= unknown)

        • Sel = min selects the left-most variable S from Ss with the minimal cost according to Cost.

        • Sel = max selects the left-most variable S from Ss with the maximal cost according to Cost.

        • Cost = card: The cost is the cardinality of the set determined by Comp.

        • Cost = weightSum: The cost is the sum of the weights associated with the elements of the set determined by Comp.

        • Cost = weightMin: The cost is the minimal weight determined by Comp.

        • Cost = weightMax: The cost is the maximal weight associated with an element of the set determined by Comp.

        • Comp = unknown selects {\it unknown}({\tt S}).

        • Comp = lowerBound selects {\it lowerBound}({\tt S}).

        • Comp = upperBound selects {\it upperBound}({\tt S}).

      • fun {Order +Ss} ... end

    • Filter determines if an element S of Ss is choosen for distribution. That is the case if {IsDet S} and the filter yields true.

      • true skips values in Ss.

      • fun {Filter +E} ... end

    • Select is used to access the actual finite set variable. Self-defined functions resp. procedures have to apply an appropriate selection function by themselves.

      • id is the identity function.

      • fun {Select +E} ... end

    • Element

      • element(sel: +Sel <= min
                wrt: +Wrt <= unknown) 

        • Sel = min selects the minimal element with respect to Wrt.

        • Sel = max selects the maximal element with respect to Wrt.

        • Wrt = unknown chooses an element from {\it unknown}(S). and interprets it as an integer.

        • Wrt = weight chooses an element from {\it unknown}(S) and takes its weight as selection criterion.

      • fun {Element +E} ... end

    • RRobin

      • true causes the distribution to step through the variable list in a round-robin fashion.

      • false causes the distribution to completely enumerate the head of the variable list and then proceeds with the head of the tail of the variable list.

    • Weights must be a list of the form [E#...]. The variable E denotes an element and W the element's weight. An list element of the form default#W assigns to all not explicitely mentioned elements the weight W. If there is no element default#W then default#0 is implicitely added.

    • Proc is applied when stability is reached. Since this application may cause instability, distribution is continued when stability is reached again.


Denys Duchier, Leif Kornstaedt, Martin Homik, Tobias Müller, Christian Schulte and Peter Van Roy
Version 1.4.0 (20080702)