Here* is a a paper that was accepted for publication at Onward! 2012 related to the crazy PL idea that I posted about a few months ago and a follow up of my last Onward! paper.
Programmers demand more extensive class libraries so they can reuse more code and write less of their own. However, these libraries are often so large that programmers get lost in deep hierarchies of classes and their members that are very broad in number. Yet language designers continue to focus on computation, leaving tools to solve library exploration problems without much help from the language.
This paper applies language design to improve IDE code completion that enables in-situ library exploration. Inference tackles depth by listing completions as long as the program can be â€œfixedâ€ to support their selection; e.g. â€œpressedâ€ can be listed as a widget completion since a widget can be a button. Influence mitigates breadth by leveraging types as completion selection models; e.g. a pressed event is more likely to be used on a button than a mouse event. We apply this design to YinYang, a language for programming simulations on tablets using touch-based menus.
* Hopefully skydrive behaves better than last time.