article Free Access Share on Incremental evaluation for a general class of circular attribute grammars Authors: J. A. Walz Cornell Univ., Ithaca, NY Cornell Univ., Ithaca, NYView Profile , G. F. Johnson Univ. of Maryland, College Park, MD Univ. of Maryland, College Park, MDView Profile Authors Info & Claims ACM SIGPLAN NoticesVolume 23Issue 7July 1988 pp 209–221https://doi.org/10.1145/960116.54011Online:01 June 1988Publication History 6citation247DownloadsMetricsTotal Citations6Total Downloads247Last 12 Months4Last 6 weeks1 Get Citation AlertsNew Citation Alert added!This alert has been successfully added and will be sent to:You will be notified whenever a record that you have chosen has been cited.To manage your alert preferences, click on the button below.Manage my AlertsNew Citation Alert!Please log in to your account Save to BinderSave to BinderCreate a New BinderNameCancelCreateExport CitationPublisher SiteeReaderPDF

A crucial aspect of a program intended for general use is its behavior in the presence of erroneous inputs. For instance, much attention has been devoted to the problems of error detection, reporting, and correction in compilers1,2. As programming languages and systems based in one way or another on unification3 become more common, it becomes increasingly important to develop a theory of error detection and correction for unification-based systems. We report here on a language-based editor for a variant of ML4 that uses a novel approach to the isolation of likely causes of user errors. As the user edits and manipulates his or her program, unification is incrementally applied to determine its type correctness. If a type inconsistency arises, a maximum flow technique is applied to the set of type equations to determine the most likely source of error. In this way a determination can be made as to the relative strengths with which the set of type equations asserts multiple contradictory hypotheses. In a language such as ML, the type of an object is inferred from patterns of usage. Often it is the case that most uses of a given object are mutually consistent, whereas one or a very small number of uses conflict with the general usage pattern. In MOE (ML-Oriented Editor), if it is possible to discern that such a situation has arisen, likely errors are highlighted at high intensity and all other program components that contributed to the inferred type of the object are highlighted at a lower intensity. In providing error information to the user, two principles are observed:(1) Error indications should be complete but parsimonious; the user should see highlighted on the screen everything that contributed directly to an error, but nothing. more,(2) The user's attention should be drawn to what appear to be the anomalies that are responsible for errors.Language-based editors permit a new level of quality in the process of helping users when inputs are, for one reason or another, invalid. Unification-based type inference in language-based editors appears to have been first considered by Meertens 5. Snelting and Bahlke have more recently also explored this approach.