Practical type inference based on success typings

Lindahl, T; Sagonas, K

dc.contributor.author	Lindahl, T	en
dc.contributor.author	Sagonas, K	en
dc.date.accessioned	2014-03-01T02:44:09Z
dc.date.available	2014-03-01T02:44:09Z
dc.date.issued	2006	en
dc.identifier.uri	https://dspace.lib.ntua.gr/xmlui/handle/123456789/31716
dc.subject	Constraint-based type inference	en
dc.subject	Erlang	en
dc.subject	Subtyping	en
dc.subject	Success typings	en
dc.subject.other	Algorithms	en
dc.subject.other	Computer programming	en
dc.subject.other	Functions	en
dc.subject.other	Image reconstruction	en
dc.subject.other	Interfaces (computer)	en
dc.subject.other	Bottom-up type inference	en
dc.subject.other	Inference algorithm	en
dc.subject.other	Program maintenance	en
dc.subject.other	Static type checks	en
dc.subject.other	Program compilers	en
dc.title	Practical type inference based on success typings	en
heal.type	conferenceItem	en
heal.identifier.primary	10.1145/1140335.1140356	en
heal.identifier.secondary	http://dx.doi.org/10.1145/1140335.1140356	en
heal.publicationDate	2006	en
heal.abstract	In languages where the compiler performs no static type checks, many programs never go wrong, but the intended use of functions and component interfaces is often undocumented or appears only in the form of comments which cannot always be trusted. This often makes program maintenance problematic. We show that it is possible to reconstruct a significant portion of the type information which is implicit in a program, automatically annotate function interfaces, and detect definite type clashes without fundamental changes to the philosophy of the language or imposing a type system which unnecessarily rejects perfectly reasonable programs. To do so, we introduce the notion of success typings of functions. Unlike most static type systems, success typings incorporate subtyping and never disallow a use of a function that will not result in a type clash during runtime. Unlike most soft typing systems that have previously been proposed, success typings allow for compositional, bottom-up type inference which appears to scale well in practice. Moreover, by taking control-flow into account and exploiting properties of the language such as its module system, success typings can be refined and become accurate and precise. We demonstrate the power and practicality of the approach by applying it to Erlang. We report on our experiences from employing the type inference algorithm, without any guidance, on programs of significant size. Copyright © 2006 ACM.	en
heal.journalName	PPDP'06 - Proceedings of the Eight ACM SIGPLAN Symposium on Principles and Practice of Declarative Programming	en
dc.identifier.doi	10.1145/1140335.1140356	en
dc.identifier.volume	2006	en
dc.identifier.spage	167	en
dc.identifier.epage	178	en