Polyvariant expansion is a binding-time-improving transformation for offline partial evaluation. We show how to achieve it automatically for a higher-order functional language using the interpretive approach. We have designed and implemented an interpreter that statically propagates binding times. When specialized with respect to a source program, it performs polyvariant expansion. Extending the interpreter to an online specializer allows us to generate a binding-time-polyvariant compiler generator. Maintaining the binding times affords an abstract interpretation equivalent to a constraint-based binding-time analysis.

Attribute grammars are well-designed to construct complex algorithms by composing several ones together. Actually, there exists a powerful transformation called descriptional composition which highly simplifies the composition of two attribute grammars by removing useless intermediate constructions. However, most of non-linear algorithms can not be expressed with attribute grammars. Thus, many compositions can not be simplified by the descriptional composition. In this paper, we present Equational Semantics, a formalism largely inspired by attribute grammars but where nonlinear algorithms can be encoded. More precisely, instead of being restricted to one input static tree as it is the case for attribute grammars, an algorithm encoded with Equational Semantics may use dynamically constructed trees. This formalism consists in an very poor abstract syntax. We present its semantics and some of its transformations such as partial evaluation and descriptional composition (also called defores...