Optimized search

[kronick]

Addresses #31 ("Improve function performance for an increasing number of stops"). Full performance results are in that ticket.

I wasn't sure if including profile.js in /tests/ is the right place for it-- just let me know if I should do something else there.

cc/ @mourner

[mourner]

This looks great apart from a minor remark. 🚢

[mourner]

The edge cases are the same code in two places, so we can move it under the binarySearchForIndex function to remove repetition.

[kronick]

@mourner Good point-- it is repetitive. I can't see a clear way right now to clean it up without adding additional conditionals in evaluateExponentialFunction to signal that we want to clamp the output and not interpolate if the edge cases are hit. The first thing I tried makes performance notably slower on case where we have few stops. Maybe you can offer a suggestion?

I tried this in evaluateExponentialFunction:

// Let binarySearchForIndex() handle the edge cases
var index = binarySearchForIndex(parameters.stops, input);

// But make sure not to run interpolation on edge cases
if (index >= parameters.stops.length - 1)
  return parameters.stops[parameters.stops.length - 1][1];
if (input <= parameters.stops[0][0])
  return parameters.stops[0][1];

return interpolate(
        input,
        base,
        parameters.stops[index][0],
        parameters.stops[index + 1][0],
        parameters.stops[index][1],
        parameters.stops[index + 1][1]
);

and performance went down to:

  >>> Evaluating exponential function for 1000000 iterations with 10 stops
  Only include values within the domain of stops:
  Time: 182.110ms
  Include values outside the domain of stops:
  Time: 97.764ms

vs about 156ms and 80ms before the change.

[mourner]

OK, let's follow up with other PRs if we find any ways to improve.