One of the coolest things about this gist is how little code is used for the animation; basically just these lines: d3.selectAll(nodes).transition()

One of the coolest things about this gist is how little code is used for the animation; basically just these lines: d3.selectAll(nodes).transition() .duration(2500) .delay(function() { return this.depth * 50; }) .ease("quad-in-out") .tween("position", function() { var d = this, i = d3.interpolate([d[0], d[1]], [d.y, d.x]); return function(t) { var p = i(t); d.t = t; d[0] = p[0]; d[1] = p[1]; }; }); The tree layout auto-magically gives every node a depth property, we delay the transition by an amount proportional to the depth, and then over the course of two-and-a-half seconds tween the line segment into the new position. Simple and effective. The hard part is generating the maze.