State Visibility and Communication in Unifying Theories of Programming

Abstract

We explore the interactions between programvariable state visibility and communication behaviour in state-rich CSP-like processes, using the Unifying Theories of Programming (UTP) framework. The key results of this work are: having variable state visible while a process is waiting to communicate, results in an operationally complex theory of behaviour; by contrast, considering state as unobservable during communication waitperiods results in an elegant theory, with much cleaner operational intuitions. The language constructs most affected by this observability choice are those of external choice and parallel composition. We also discuss situations where this state hiding can prevent the adoption of interesting operators that seize control from waiting processes.

Unifying theories of reactive design contracts

“…Definition 3.9); and (2) that the pericondition does not refer to st , as characterised by the existential quantifier. The latter follows as a direct consequence of R3 h [7] (cf. Definition 2.10), which requires that st must not be restricted in an intermediate state.…”

“…Our generalised theory [16] removes the ref and ref variables, which allows us to characterise behavioural semantic models other than failures-divergences. Moreover, we add st, st : Σ to explicitly model state as suggested by [7], where Σ is a state space type. In our previous work [16], we have shown how the UTP theory of reactive processes can be generalised by characterising the trace model with an abstract algebra, called a "trace algebra".…”

“…Finally, we have R3 h , a version of R3 from [7] that introduces the concept of intermediate observations, whilst ensuring that state variables are not included. R3 h states that if a process observes wait to be true, then its predecessor has not yet terminated and thus it should behave like the reactive identity, I I R .…”

“…This operator does not give special treatment to state variables: they are simply identified in intermediate states like other observational variables. As discussed in detail in Section 6, R3 h allows a simpler treatment of state variables, supports additional algebraic laws for assignment and state substitution, and solves the problem with external choice for which it was originally designed [7], though at the cost of losing McEwan's interruption operator [40]. Thus, the use of R3 h instead of R3 is a design decision based on the particular modelling facilities of interest.…”

“…If P terminates, then conversely the behaviours of Q are lost. Crucially, the internal state of P is retained following interruption [7], and is passed on to Q, which is unique to McEwan's work (cf. [53,58,60]).…”