Bor-Yuh Evan Chang: Research Statement

Domain Combinators. Recent years have seen explosion in the use of dynamic languages, such as JavaScript, in software systems because of their ease of use for prototyping but also because of their flexibility. This flexibility has enabled numerous advanced and useful frameworks that simply cannot exist in traditional static languages. At the same time, this flexibility has come with a steep cost in tool support: current static analysis techniques designed primarily for traditional static languages are ineffective. Collaborating with researchers at ENS Paris, we have designed a number of abstract domain combinators that directly target dynamic language features in the Jsana project, such as sets and dictionaries [8–12]. This line of work asserts that dynamic languages are not beyond the scope of analysis techniques but require a rethinking of the way in which we apply them. I have also contribute to the design of several abstract domain combinators for reasoning about the heap of low-level C code. This line of work has addressed reasoning about multiple views to the same memory as with C-style unions [14], multiple inductive summaries of a recursive data structure [17, 18], and for unstructured sharing [15]. It has also addressed a fundamental problem in simultaneously reasoning about recursion in memory and recursion in execution [16] and was published at POPL 2011—a top, highly-selective venue (23%). A synthesis of this line of work was solicited as an invited paper [5].