Copyright 2012, Ben Sigelman ([email protected])
MIT-licensed (http://opensource.org/licenses/MIT)

'ftv' is the Fast Timeseries Visualizer.

In its current form, ftv is a proof-of-concept. Many necessary features are
missing, but what's present demonstrates how fast client-side timeseries
visualization can be in a modern browser. The following techniques help to keep
ftv speedy:
 - The bulk of the data (the timestamps and points themselves) are never copied
   within the javascript layer. When loading data from a server (like the
   demonstration server, written in python), this means that the data comes
   back in a specific binary format that's friendly for the javascript DataView
   API. The javascript layer then establishes pointers into that data
   structure, rather than copying it or, worse still, parsing an analogous JSON
   representation.
 - The HTML5 Canvas calling patterns are friendly to hardware acceleration
   (though there's still more work that could be done).
 - Interactive updates to the canvas do not require a full redraw of the
   "background", nor do they require an additional transparent canvas overlay;
   this seems to be the more efficient approach in the implementations tried so
   far.
 - FTV doesn't have lots of features. It renders timeseries, but it does not do
   statistical aggregations of those timeseries, et cetera. This, of course,
   makes it easier to keep the implementation simple and efficient.