An efficient method for handling operator overloading in Ada

Abstract

The coupling of type derivation with operator overloading occupies a substantial portion of the type-checking activities of an Ada compiler. It is thus important to handle overloaded operators as efficiently as possible. Let us recall that the introduction of any type derived from a numeric type, say: type T is range 1 .. 10; (1) creates, at the point of declaration of type T, a new instance of each of the predefined operators that apply to the root type of T, namely INTEGER. In other words, declarations of the form function "+"(x, y : T'BASE) return T'BASE; (2) are understood to appear immediately after the declaration (1). (Here T'BASE refers to the anonymous type, derived from integer, of which type T is a subtype). This means that an innocent declaration such as (1) introduces implicitly 16 new operator declarations (10 arithmetic operators and 6 comparison operators). The visibility of operator designators such as "+" is described in [RM] in terms of these implicit declarations. In what follows we refer to the operators thus introduced as derived operators, in contrast with user-defined operators, which are introduced by explicit declarations and are given user-defined bodies.