dijkstra.pl : Dijkstra's single-source shortest path algorithm Single-source shortest path algorithm based on [Dijkstra 1959]. Given a weighted directed graph and a source node, shortest distances to all other nodes are computed. Nodes are numbered from 1, weights can be any numeric type.
How to use: The graph is represented using edge/3 constraints: edge(A,B,W): there is an edge between node A and node B, with weight W The following operation is supported: dijkstra(S): compute all distances from source node S The resulting output is represented as: distance(B,D): the distance from the source node to node B is D Internal operations are: scan(N,L): scan the neighbours of node N, which is at distance L from the source node, then get the next non-scanned node with the smallest label (distance) relabel(N,L): if N is already scanned, do nothing; otherwise set the label of N to L
See also: Uses Fibonnacci heaps (see fib_heap.pl). Reference paper: Dijkstra's Algorithm with Fibonacci Heaps: An Executable Description in CHR. Jon Sneyers, Tom Schrijvers and Bart Demoen. WLP 2006. See also shortestpaths.pl for an all-pairs shortest path algorithm.
Program: Change the code, then submit! /* dijkstra.pl: Dijkstra's single-source shortest path algorithm (C) Jon.Sneyers at cs.kuleuven.be, 2006 This program is distributed under the terms of the GNU General Public License: http://www.gnu.org/licenses/gpl.html %% DESCRIPTION Single-source shortest path algorithm based on [Dijkstra 1959].# Given a weighted directed graph and a source node, shortest distances to all other nodes are computed. Nodes are numbered from 1, weights can be any numeric type. %% HOW TO USE The graph is represented using edge/3 constraints: # *edge(A,B,W)*: there is an edge between node A and node B, with weight W The following operation is supported: # *dijkstra(S)*: compute all distances from source node S The resulting output is represented as: # *distance(B,D)*: the distance from the source node to node B is D Internal operations are: # *scan(N,L)*: scan the neighbours of node N, which is at distance L from the source node, then get the next non-scanned node with the smallest label (distance) # *relabel(N,L)*: if N is already scanned, do nothing; otherwise set the label of N to L %% SEE ALSO Uses Fibonnacci heaps (see fib_heap.pl).# Reference paper: Dijkstra's Algorithm with Fibonacci Heaps: An Executable Description in CHR. Jon Sneyers, Tom Schrijvers and Bart Demoen. WLP 2006.# See also shortestpaths.pl for an all-pairs shortest path algorithm. %% SAMPLE QUERIES Q: edge(1,2,3),edge(2,4,8),edge(1,3,5),edge(3,4,2),edge(2,3,1), dijkstra(1). A: edge(1,3,5), edge(1,2,3), edge(2,3,1), edge(2,4,8), edge(3,4,2), distance(1,0), distance(2,3), distance(3,4), distance(4,6). */ :- module(dijkstra,[edge/3,dijkstra/1,distance/2]). user:library_directory('.'). :- use_module(library(chr)). :- use_module(library(fib_heap)). %% Deprecated syntax used for SICStus 3.x %handler dijkstra. %constraints % edge/3, dijkstra/1, distance/2, scan/2, relabel/2. %% Syntax for SWI / SICStus 4.x :- chr_constraint edge(?node,+int,+number), dijkstra(+int), distance(?node,+number), scan(+int,+number), relabel(+int,+number). %:- chr_type node == dense_int. % efficient: arrays :- chr_type node == int. % less efficient: hashtables start_scanning @ dijkstra(A) <=> scan(A,0). label_neighb @ scan(N,L), edge(N,N2,W) ==> L2 is L+W, relabel(N2,L2). scan_next @ scan(N,L) <=> distance(N,L), (extract_min(N2,L2) -> scan(N2,L2) ; true). scanned @ distance(N,_) \ relabel(N,_) <=> true. not_scanned @ relabel(N,L) <=> decr_or_ins(N,L).
Console: Enter query or select example from below, then submit and wait for answer! % loading graph/dijkstra.pl | ?- consult(...). yes [10.573 seconds] | ?-