Hey look, it's my first official MacOS app on the @AppStore.

apps.apple.com/us/app/fof/id1…
WTF is this you may ask?

Fair question.

It scratched a *very* small & *very* specific itch we^H^H I had at @twitter.
Over the last few years the #TwitterVMTeam has been using machine learning to automatically tune application JVMs:

youtube.com/watch?v=MODume…
Ian Brown @igb
A challenge in doing a demo of this work is creating an application w/ multiple parameters influencing performance in deterministic ways.
View on Twitter ↗
0 0

Replies

In order to model multiple functions in code with interesting and varying outputs for different input ranges...
...a developer would have to do a lot of math and have an intuitive sense of what function would produce what output.
Personally I find it easier to work backwards from the desired output by drawing the expected graph of output for input X...
...and then generating a function that interpolates between recorded points on the graph for the value of Y for any given X.
So that's what this app does: you draw a graph and then generate code in Java, Python, Scala, or JavaScript...
...for a function that will model the drawn curve with reasonable fidelity.
This makes it easy to add deterministic and programmable functionality to an application, demo, or test.
Anyway, in the unlikely event you have the same problem, this app might help you out. If you do find it useful let me know!