Accepted answer

Crossfilter works best with a flat array of observations. It looks like your data might best be represented by observations (rows) of {machine, x, y}

It's hard to tell with the example you've given, because it is very deeply nested but many of the levels are unused, exactly what the format is. I'd advise you to either flatten it out when you're producing it, or if you have no control over that, loop over it and produce an array of just the deepest {x,y} objects, but annotate them with the machine id from the level above that.

Related Query

More Query from same tag