So, if I get this right, being 'data-friendly' sums up to using data in the form [ {val1: 1, val2: 2}, {val1: 2, val2: 3}] and then mapping x: 'val1', y: 'val2' instead of something like {x: [1,2], y: [2,3]}. The former assumes that the input data comes in a csv or a similar format. This makes charting computed data (à la MATLAB or numpy/matplotlib) rather awkward.

You've got it right. Also I have an idea to make an API that joins data processing and plotting like tauCharts.data(#).aggregate(#).plot(#)