Reputation: 852
I have 2 lists - nodes and links... Now what I would want is the most efficient way to add all the directly/indirectly linked elements into different groups.... For eg, 0 is connected to 1 which is connected to 2 so nodes 0,1,2 become group 1.... node 3 is connected to 4 so it becomes group 2 and so on.... Thanks in advance for your help :) This is part of a d3 implementation am doing..
PS: These lists will easily be in tens of thousands of nodes and links.
"nodes":[
{
"id":0,
"x":1509.9862,
"y":-609.1013
},
{
"id":1,
"x":1645.9578,
"y":-85.06705
},
{
"id":2,
"x":1948.1533,
"y":-469.3646
},
{
"id":3,
"x":348.1533,
"y":-669.3646
},
{
"id":4,
"x":1448.1533,
"y":-1469.3646
}
...
]
"links":[
{
"from":0,
"to":1
},
{
"from":1,
"to":2
},
{
"from":3,
"to":4
}
...
]
Upvotes: 1
Views: 484
Reputation: 1977
This is a classic UnionFind problem. The idea is to see each node as a set that has a pointer point to its father. Nodes with the same father are in the same group. So for your problem, we can create n sets at the beginning. And then iterate through the link to group everyone connected by the same link. The complexity is O(n), where n is the number of nodes.
nodes = [{
"id": 0,
"x": 1509.9862,
"y": -609.1013
},
{
"id": 1,
"x": 1645.9578,
"y": -85.06705
},
{
"id": 2,
"x": 1948.1533,
"y": -469.3646
},
{
"id": 3,
"x": 348.1533,
"y": -669.3646
},
{
"id": 4,
"x": 1448.1533,
"y": -1469.3646
}
];
links = [{
"from": 0,
"to": 1
},
{
"from": 1,
"to": 2
},
{
"from": 3,
"to": 4
}
];
// union-find is a data structure that can union two sets and check
// whether two element in the same set.
var father = {};
function group(nodes, links) {
// create n set with each set has the node as its only element
nodes.forEach(function(node, i) {
father[node.id] = node.id;
});
// union each set that has a link between them
links.forEach(function(link, i) {
union(link.from, link.to);
});
// for each unioned set, group nodes together
var id = 1;
var groupIdCnt = {};
var groupIds = {};
nodes.forEach(function(node, i) {
var f = find(node.id);
if (typeof groupIds[f] === 'undefined') {
groupIds[f] = id;
groupIdCnt[id] = 1;
id++;
} else {
groupIdCnt[groupIds[f]]++;
}
});
var groups = {};
nodes.forEach(function(node, i) {
var f = find(node.id);
if (groupIdCnt[groupIds[f]] === 1) {
node['group'] = 0;
} else {
node['group'] = groupIds[f];
}
if (typeof groups[node['group']] === 'undefined') {
groups[node['group']] = [];
}
groups[node['group']].push(node);
});
return Object.values(groups);
}
// find father of each set
function find(node) {
// if it is the root, return
if (father[node] === node) {
return node;
}
// if not, find the father and point to it
father[node] = find(father[node]);
return father[node];
}
// update the father of set which includes node1 to the father of set which includes node 2
function union(node1, node2) {
father[find(node1)] = find(node2);
}
// O(n), since we visit each node once
var groups = group(nodes, links);
console.log(nodes);
console.log(groups);
Upvotes: 3
Reputation: 12458
In the solution below I am creating groups of link
s that are, well, linked to each other. I do so by looping through all of the from
/to
combinations, and finding out if either has already been added to any of the accumulating groups of link
s. If they have, then I just add (or re-add) both the from
and to
value to that group. If neither the from
nor to
value has yet been grouped, then I make a new group and add both the from
and to
values to it. Note that these "groups" are actually Javascript sets, a new ES6/ES2015 data type that makes it much easier to deal with "groups" of elements for which no duplicates are needed and/or allowed.
Once the groups/sets of links are established, I then simply add an attribute to each node
that indicates which group of link
s it is a part of.
Note that, for the sake of this demo code, I've simplified/de-cluttered some node
values. I've also added some extra links, just to demonstrate some further cases that need handling.
const groupNodes = (nodes, links) => {
const groups = links.reduce((grps, {from, to}) => {
if (!grps.some(grp => {
if (grp.has(from) || grp.has(to)) return grp.add(from).add(to);
})) grps.push(new Set([from, to]));
return grps;
}, []);
nodes.forEach(node => {
groups.forEach((grp, i) => { if (grp.has(node.id)) node.group = i; });
});
return nodes;
};
const nodes = [
{
"id":0,
"x":0,
"y":0
},
{
"id":1,
"x":11,
"y":-11
},
{
"id":2,
"x":22,
"y":-22
},
{
"id":3,
"x":33,
"y":-33
},
{
"id":4,
"x":44,
"y":-44
},
{
"id":5,
"x":55,
"y":-55
},
{
"id":6,
"x":66,
"y":-66
}
];
const links = [
{
"from": 0,
"to" : 1
},
{
"from": 1,
"to" : 2
},
{
"from": 2,
"to" : 0
},
{
"from": 3,
"to" : 4
},
{
"from": 4,
"to" : 5
},
{
"from": 6,
"to" : 0
}
];
console.log(JSON.stringify(groupNodes(nodes, links)));
Upvotes: 1
Reputation: 22247
This code spits out an object whose keys are the node id and whose values are a group id, not necessarily sequential.
var obj = {
"links":[
{
"from":0,
"to":1
},
{
"from":1,
"to":2
},
{
"from":5,
"to":4
},
{
"from":3,
"to":4
}
]
};
var groups = {};
var nextGrp = 1;
for (var i=0, l; l = obj.links[i]; i++) {
if (groups[l.from]) {
if (groups[l.to]) {
if (groups[l.to] != groups[l.from]) {
// the two items span two different groups which must now be joined into 1
for (var j in groups) {
if (groups[j] == groups[l.to]) {
groups[j] = groups[l.from];
}
}
}
} else {
groups[l.to] = groups[l.from];
}
} else if (groups[l.to]) {
groups[l.from] = groups[l.to];
} else {
groups[l.from] = nextGrp;
groups[l.to] = nextGrp;
nextGrp++;
}
}
console.log(groups);
Upvotes: 1