function node_rep = node_merge ( dim_num, node_num, node_xy, tolerance )
%*****************************************************************************80
%
%% node_merge() detects nodes that should be merged.
%
% Discussion:
%
% Two nodes "should" be merged if they are within TOLERANCE distance
% of each other.
%
% With a tolerance of 0, only exactly equal nodes are counted.
%
% With a positive tolerance, a pair of nodes inside a circle of
% radius TOLERANCE result in a count of 1 duplicate.
%
% However, what do we do if nodes A, B and C are arranged in a line,%
% with A and B just within TOLERANCE of each other, and B and C just
% within tolerance of each other? What we do here is make a choice
% that can be defended consistently. A and B define an equivalence
% class because they are closer than TOLERANCE. C is then added to
% this equivalence class, because it is within TOLERANCE of at least
% on thing in that equivalence class.
%
% Thus, if 100 nodes are separated pairwise by slightly less
% than TOLERANCE, a total of 99 duplicates will be counted.
%
% The program starts out by giving each node its own label.
% If it finds that two nodes should be merged, then the index of
% one node is used as the label for both. This process continues
% until all nodes have been considered. The number of unique nodes
% is the number of unique values in the output quantity NODE_REP.
%
% Licensing:
%
% This code is distributed under the GNU LGPL license.
%
% Modified:
%
% 24 August 2006
%
% Author:
%
% John Burkardt
%
% Input:
%
% integer DIM_NUM, the spatial dimension.
%
% integer NODE_NUM, the number of nodes.
%
% real NODE_XY(DIM_NUM,NODE_NUM), the nodes.
%
% real TOLERANCE, the maximum distance between
% two nodes regarded as duplicate.
%
% Output:
%
% integer NODE_REP(NODE_NUM), the "representative" of each node.
% NODE_REP(NODE) is the index of a node which is within TOLERANCE of node
% NODE, or for which a chain of nodes can be found, all having the
% same representative, and all of which are pairwise closer than TOLERANCE.
%
for node1 = 1 : node_num
node_rep(node1) = node1;
end
for node1 = 1 : node_num
rep_dist(1:node_num) = inf;
for node2 = 1 : node_num
dist = sqrt ( sum ( ...
( node_xy(1:dim_num,node1) - node_xy(1:dim_num,node2) ).^2 ) );
rep = node_rep(node2);
if ( dist < rep_dist(rep) )
rep_dist(rep) = dist;
end
end
for node2 = 1 : node_num
rep = node_rep(node2);
if ( rep_dist(rep) <= tolerance )
node_rep(node2) = node1;
end
end
end
return
end