Separation logic

Separation logic is an extension of Hoare's logic oriented to reasoning about mutable data structures (or, programs with dynamically allocated pointers). It enables more compact proofs and specs of imperative programs than before because of its support for local reasoning, where specifications and proofs concentrate on the portion of memory used by a program component, and not the entire global state of the system.

Assertions. Separation logic assertions describe heaps. Besides the normal logical connectives (conjunction, disjunction, implication, etc), SL introduces the separating conjunction ($*$) of two assertions. A heap satisfies the separating conjunction of two assertions if it can be divided into two disjoint parts satisfying the two assertions. Formally, $h \models P_1 * P_2$ if and only if there exist (disjoint) $h_1$ and $h_2$ such that $h = h_1 \uplus h_2$ and $h_1 \models P_1$ and $h_2 \models P_2$.

Proof rules. SL inherits the standard rules and axioms of Hoare logic, has its own proof rules for heap-manipulating commands, and the all-important frame rule.

$$\frac{ \{P\}~C\{Q\} }{ \{P*F\}~C\{Q*F\} }\quad \mathbf{fv}(F)\cap\mathbf{mod}(C) = \emptyset$$

This says that a program that runs correctly with a heap satifying its precondition also runs correctly with a bigger initial heap. Further, it is guaranteed not to touch the additional portion heap.

The frame rule enables local reasoning: the specification of a program need only mention the portion of the heap used by the program.

Introductory material

Extensions of Separation Logic

There are extensions of separation logic:

Tool support

Separation logic has also been extensively used in semi-automatic and automatic program analyses (often known as shape analyses) for proving functional correctness of programs or simply the absence of certain kinds of runtime errors and/or memory leaks.

For more information, see ToolSupport.

Further reading

TODO