The other day, over a very pleasant lunch in the restaurant of Oxford's recently renovated Ashmolean Museum, Oege de Moor gave me a problem about rectangles. The problem is explained more fully later, but roughly speaking one is given a finite set of rectangles RS and a rectangle R completely covered by RS. The task is to construct a single rectangle covering R among the elements of a larger set of rectangles associated with RS, called the saturation of RS. The saturation of RS is the closure of RS under so-called consensus operations, a term coined in (Quine, 1959), in which two rectangles are combined in two distinct ways to form new rectangles. The rectangle problem is a simplified version of containment-checking, a crucial component in a type inference algorithm for Datalog programs (Schäfer & de Moor, 2010). 19 In the Schäfer-de Moor algorithm the problem is generalised to cubes in n-space rather than rectangles in two-space, the components of each cube are given by propositional formulae rather than by intervals on the real line, and certain equality and inhabitation constraints are taken into account. Oege felt that the central proof, Lemma 15 in (Schäfer & de Moor, 2010), deserved to be simplified so he posed the rectangle problem as a special case. This pearl was composed in response to the challenge.

Suppose you are given a number of points in a plane and want to have those lines that each contain a large number of the given points. The Hough transform is a computerized procedure for that task. It was invented by Paul Hough (1962), originally to find the trajectories of subatomic particles in a bubble chamber, and it has even been patented. Nowadays, adaptations of the Hough transform are used, among others, for identification of transformed instances of a predefined figure, instead of just a line, in a digital picture. There are plenty of explanations on the Internet (use search key “Hough transform” and “generalized Hough transform”), some with nice applets to demonstrate the working (add search key “applet” or “demo”). Recently, Hart (2009) has looked back at the invention. We show how the original procedure could have been derived.

Discrete Interval Encoding Trees are data structures for the representation of fat, i.e. densely populated sets over a discrete linear order. In this paper, we introduce algorithms for set-theoretic operations like intersection, union, etc. on sets represented as balanced diets. We empirically analyse their performance and show that these algorithms can outperform previously known algorithms on sets, such as the ones implemented in OCaml's standard library.

Behavioral type and effect systems regulate properties such as adherence to object and communication protocols, dynamic security policies, avoidance of race conditions, and many others. Typically, each system is based on some specific syntax of constraints, and is checked with an ad hoc solver. Instead, we advocate types refined with first-order logic formulas as a basis for behavioral type systems, and general purpose automated theorem provers as an effective means of checking programs. To illustrate this approach, we define a triple of security-related type systems: for role-based access control, for stack inspection, and for history-based access control. The three are all instances of a refined state monad. Our semantics allows a precise comparison of the similarities and differences of these mechanisms. In our examples, the benefit of behavioral type-checking is to rule out the possibility of unexpected security exceptions, a common problem with code-based access control.

The worker/wrapper transformation is a general way of changing the type of a recursive definition, usually applied with an eye to increasing algorithmic efficiency. This note identifies an infelicity in the program transformations presented by Gill & Hutton (The worker/wrapper transformation, J. Funct. Program., vol. 19, 2009, pp. 227–251) and proposes a new totally correct worker/wrapper fusion rule.