Principles of Abstract Interpretation

Abstract

Some years ago, the author of the reviewed book and the author of the review shared a ride in the shuttle from Grenoble to Lyon Airport. The author-to-be told the reviewer-to-be about the state of his book project. The existing draft had 1,000 pages. The reviewer explained to the author that in his deep insight into the nature of things and his long-term experience with textbooks there were books that improve the world and there were books that are being read. The author asked for confirmation whether the reviewer felt that the coming book would belong to those books that improve the world. Let me skip how I got myself out of this difficult situation. When asked whether a book will be read, the question is by whom. Citing the author, this book is intended for readers interested in the theory of abstract interpretation, the understanding of formal methods, and the design of verifiers and static analyzers. And my answer is, it is a must read for these groups of people. To make one thing clear from the beginning: This reviewer need not be convinced of the value of Abstract Interpretation as his greatest scientific achievements [5, 6, 8, 10] are based on the foundational work on Abstract Interpretation by Patrick and Radhia Cousot. Static analyses are distinct from . . .model checking, which verifies the correctness of a separate external specification of a program [9]. In model checking, a user supplies the program to be verified and the logical expression or the automaton against which the program is to be checked at the same time. In static analysis there are distinct times, a time when an abstract interpreter is designed with certain facts in mind to be extracted from a class of programs, and there is a time when the abstract interpreter is applied by programmers or verification engineers to extract these type of facts from particular programs. This enables a fruitful division of work. The first phase, the design of non-trivial abstract interpreters needs highly competent specialists, while the second phase is easier, although sometimes also non-trivial. The designer needs to identify abstract domains to