Javascript – Fix Node Position in D3 Force Directed Layout

d3-force-directedd3.jsdata-visualizationforce-layoutjavascript

I want some of the nodes in my force-directed layout to ignore all forces and stay in fixed positions based on an attribute of the node, while still being able to be dragged and exert repulsion on other nodes and maintain their link lines.

I thought it would be as simple as this:

force.on("tick", function() {
    vis.selectAll("g.node")
        .attr("transform", function(d) {
            return (d.someAttribute == true) ?
               "translate(" + d.xcoordFromAttribute + "," + d.ycoordFromAttribute +")" :
               "translate(" + d.x + "," + d.y + ")"
        });
  });

I have also tried to manually set the node's x and y attributes each tick, but then the links continue to float out to where the node would be if it was affected by the force.

Obviously I have a basic misunderstanding of how this is supposed to work. How can I fix nodes in a position, while keeping links and still allowing for them to be draggable?

Best Answer

Set d.fixed on the desired nodes to true, and initialize d.x and d.y to the desired position. These nodes will then still be part of the simulation, and you can use the normal display code (e.g., setting a transform attribute); however, because they are marked as fixed, they can only be moved by dragging and not by the simulation.

See the force layout documentation for more details (v3 docs, current docs), and also see how the root node is positioned in this example.

Related Solutions

Javascript – Retrieve the position (X,Y) of an HTML element

The correct approach is to use element.getBoundingClientRect():

var rect = element.getBoundingClientRect();
console.log(rect.top, rect.right, rect.bottom, rect.left);

Internet Explorer has supported this since as long as you are likely to care about and it was finally standardized in CSSOM Views. All other browsers adopted it a long time ago.

Some browsers also return height and width properties, though this is non-standard. If you're worried about older browser compatibility, check this answer's revisions for an optimised degrading implementation.

The values returned by element.getBoundingClientRect() are relative to the viewport. If you need it relative to another element, simply subtract one rectangle from the other:

var bodyRect = document.body.getBoundingClientRect(),
    elemRect = element.getBoundingClientRect(),
    offset   = elemRect.top - bodyRect.top;

alert('Element is ' + offset + ' vertical pixels from <body>');

D3.js – D3 force directed layout with bounding box

There's a bounding box example in my talk on force layouts. The position Verlet integration allows you to define geometric constraints (such as bounding boxes and collision detection) inside the "tick" event listener; simply move the nodes to comply with the constraint and the simulation will adapt accordingly.

That said, gravity is definitely a more flexible way to deal with this problem, since it allows users to drag the graph outside the bounding box temporarily and then the graph will recover. Depend on the size of the graph and the size of the displayed area, you should experiment with different relative strengths of gravity and charge (repulsion) to get your graph to fit.

Best Answer

Related Solutions

Javascript – Retrieve the position (X,Y) of an HTML element

D3.js – D3 force directed layout with bounding box

Related Topic