competencegraph/js/sigma/sigma.plugins.filter

sigma.plugins.filter

Plugin developed by Sébastien Heymann for Linkurious.

---

General

This plugin filters nodes and edges in a fancy manner:

• Define your own filters on nodes and edges using the nodesBy and edgesBy methods, or execute more complex filters using the neighborsOf method.
• Register multiple filters before applying them anytime at once.
• Undo any filter while preserving the execution order.
• Chain all methods for concise style.

See the following example code and unit tests for full usage.

To use, include all .js files under this folder. Then initialize it as follows:

var filter = new sigma.plugins.filter(sigInst);

Predicates

Predicates are truth tests (i.e. functions which return a boolean) on a single node or a single edge. They return true if the element should be visible. For instance:

// Only edges of size above one should be visible:
function(e) {
  return e.size > 1;
}

In this example, notice that if the size attribute is undefined, the edge will be hidden. If you still want to display edges with no size attribute defined, you have to modify the predicate a bit:

// Only edges of size above one should be visible:
function(e) {
  return e.size === undefined || e.size > 1;
}

Predicates are applied by predicate processors.

Predicate processors

Predicate processors are functions which wrap one predicate and apply it to the graph. Three predicate processors are available:

• nodesBy
• edgesBy
• neighborsOf

For each node of the graph, the nodesBy processor sets the attribute hidden to false if the predicate is true for the node. It also sets the hidden attribute of edges to true if one of the edge's extremities is hidden. For instance:

// Only connected nodes (i.e. nodes of positive degree) should be visible:
filter.nodesBy(function(n) {
  return this.degree(n.id) > 0;
}, 'non-isolates');

For each edge of the graph, the edgesBy processor sets the attribute hidden to false if the predicate is true for the edge. For instance:

// Only edges of size above one should be visible:
filter.edgesBy(function(e) {
  return e.size > 1;
}, 'edge-size-above-one');

For each neighbor node of a specified node, the neighborsOf processor sets the attribute hidden to true if it is not directly connected to the node. It also sets the hidden attribute of edges to true if one of the edge's extremities is hidden. For instance:

// Only neighbors of the node 'n0' should be visible:
filter.neighborsOf('n0');

Processors instanciated with a predicate are called filters. Filters are not applied until the apply method is called.

Filters chain

Combining filters is easy! Declare one filter after another, then call the apply method to execute them on the graph in that order. For instance:

// graph = {
//   nodes: [{id:'n0'}, {id:'n1'}, {id:'n2'}, {id:'n3'}], 
//   edges: [
//     {id:'e0', source:'n0', target:'n1', size:1},
//     {id:'e1', source:'n1', target:'n2', size:0.5},
//     {id:'e2', source:'n1', target:'n2'}]
// }
filter
    .nodesBy(function(n) {
      return this.degree(n.id) > 0;
    })
    .edgesBy(function(e) {
      return e.size >= 1;
    })
    .apply();
// n3.hidden == true
// e1.hidden == true
// e2.hidden == true

Combined filters work like if there was an 'AND' operator between them. Be careful not to create mutually exclusive filters, for instance:

filter
    .nodesBy(function(n) {
      return n.attributes.animal === 'pony';
    })
    .nodesBy(function(n) {
      return n.attributes.animal !== 'pony';
    })
    .apply();
// all nodes are hidden

Filters are internally stored in an array called the chain.

Undo filters

Undoing filters means to remove them from the chain. Filters can be undone easily. Choose which filter(s) to undo, or undo all of them at once.

Filters can be associated with keys at declaration, where keys are any string you give. For instance, the following filter has the key node-animal:

filter.nodesBy(function(n) {
  return n.attributes.animal === 'pony';
}, 'node-animal');

Manually undo this filter as follows:

filter
  .undo('node-animal')
  .apply(); // we want it applied now

Multiple filters can be undone at once, for instance:

filter.undo('node-animal', 'edge-size', 'high-node-degree');
// don't forget to call `apply()` anytime!

Alternative syntax:

var a = ['node-animal', 'edge-size', 'high-node-degree'];
filter.undo(a);
// don't forget to call `apply()` anytime!

Finally, undo all filters (with or without keys) as follows:

filter.undo();
// don't forget to call `apply()` anytime!

Warning: you can't declare two filters with the same key, or it will throw an exception.

Export the chain

The exported chain is an array of objects. Each object represents a filter by a triplet (?key, processor, predicate). The processor value is the internal name of the processor: filter.processors.nodes, filter.processors.edges, filter.processors.neighbors. The predicate value is a copy of the predicate function. Dump the chain using the export method as follows:

var chain = filter.export();
// chain == [
//  {
//     key: '...', 
//     processor: '...', 
//     predicate: function() {...}
//   }, ...
// ]

Import a chain

You can load a filters chain using the import method:

var chain = [
  {
    key: 'my-filter',
    predicate: function(n) { return this.degree(n.id) > 0; },
    processor: 'filter.processors.nodes'
  }
];
filter
  .import(chain)
  .apply();