An abstract domain for objects in dynamic programming languages

Dynamic languages, such as JavaScript, PHP, Python or Ruby, provide a memory model for objects data structures allowing programmers to dynamically create, manipulate, and delete objects’ properties. Moreover, in dynamic languages it is possible to access and update properties by using strings: this represents a hard challenge for static analysis. In this paper, we exploit the finite state automata abstract domain, approximating strings, in order to define a novel abstract domain for objects. We design an abstract interpreter useful to analyze objects in a toy language, inspired by real-word dynamic programming languages. We then show, by means of minimal yet expressive examples, the precision of the proposed abstract domain.