An Operational Semantics for Constraint-Logic Imperative Programming

Functional and Constraint Logic Programming

“…As a result, this allows search to focus on branches relevant to the respective access, thus effectively reducing the state space. Note that these considerations are similar to those regarding the Label reduction rule from [3] that is used for substituting primitive type logic variables with their potential values. Similar to the present case, Label is suggested to be used only as a last resort if no other rule can be applied as its application results in one branch per potential value, which usually are a lot.…”

“…However, if an expression contains unbound variables, they cannot evaluate to a specific value. Therefore, the execution environment treats those variables symbolically and creates a symbolic expression [3]. For instance, after executing Listing 2, y holds the constant value 5 (as expected in Java), whereas z holds the symbolic expression x + 5. int x free; int i = 2, j = 3; int y = i + j; // y == 5. int z = x + y; // z == x + 5.…”

“…However, as soon as a symbolic expression is used as part of a condition that leads to branching (e. g., in an if statement), it is possible that the execution environment cannot decide on a unique outcome, e. g. whether a condition evaluates to true or false. When there is more than one choice, non-determinism is introduced, so that execution may continue with any of the possible branches [3]. The execution environment makes a choice by selecting a branch, thus asserting a particular outcome (e. g., the condition shall be false).…”

“…Since type variables are that different, they are excluded from further considerations in this work, resulting in a definition of reference types that is congruent to that of C# [10]. 3 Class and interface types exhibit an identical structure [5], whereas array types are interpreted differently. Even though array types are interesting as well, this work focuses on class and interface types for now.…”

“…Similar to the present case, Label is suggested to be used only as a last resort if no other rule can be applied as its application results in one branch per potential value, which usually are a lot. If this is done too early during evaluation, this increases the state space unnecessarily [3].…”