Exception handling in scientific computing

This paper describes some simple programming language facilities for raising and handling exceptions, and demonstrates, with a number of examples, their effectiveness in a scientific computing environment. Only a small number of predefined exceptions need to be introduced into the language, but any number of user-defined exceptions can he introduced by the programmer. A key feature is that handlers are attached to operators that can raise exceptions (including user-defined functions and procedures, as well as built-in operators), rather than to expressions or statements. The programmer has a great deal of freedom in specifying what action is to be taken in case an exception arises. The implementation and the applications are mainly concerned with scientific computing, but we believe the facilities are applicable in other areas as well.