Functional Programming Community Research Articles

Functional Programming in Financial Markets (Experience Report)

We present a case-study of using functional programming in the real world at a very large scale. At Standard Chartered Bank, Haskell is used in a core software library supporting the entire Markets division -- a business line with 3 billion USD operating income in 2023. Typed functional programming is used across the entire tech stack, including foundational APIs and CLIs for deal valuation and risk analysis, server-side components for long-running batches or sub-second RESTful services, and end-user GUIs. Thousands of users across Markets interact with software built using functional programming, and over one hundred write functional code. In this experience report we focus on how we leverage functional programming to orchestrate type-driven large-scale pricing workflows. The same API can be used to price one trade locally, or millions of trades across thousands of cloud nodes. Different parts of the computation can be run and inspected individually, and recomputing one part triggers recalculation of the dependent parts only. We build upon decades of research and experience in the functional programming community, relying on concepts such as monads, lenses, datatype generics, and closure serialisation. We conclude that the use of functional programming is one of the main drivers of the success of our project, and we see no significant downsides from it.

Efficient Dual-Numbers Reverse AD via Well-Known Program Transformations

Where dual-numbers forward-mode automatic differentiation (AD) pairs each scalar value with its tangent value, dual-numbers reverse-mode AD attempts to achieve reverse AD using a similarly simple idea: by pairing each scalar value with a backpropagator function. Its correctness and efficiency on higher-order input languages have been analysed by Brunel, Mazza and Pagani, but this analysis used a custom operational semantics for which it is unclear whether it can be implemented efficiently. We take inspiration from their use of linear factoring to optimise dual-numbers reverse-mode AD to an algorithm that has the correct complexity and enjoys an efficient implementation in a standard functional language with support for mutable arrays, such as Haskell. Aside from the linear factoring ingredient, our optimisation steps consist of well-known ideas from the functional programming community. We demonstrate the use of our technique by providing a practical implementation that differentiates most of Haskell98.

Generating circuits with generators

The most widely used languages and methods used for designing digital hardware fall into two rough categories. One of them, register transfer level (RTL), requires specifying each and every component in the designed circuit. This gives the designer full control, but burdens the designer with many trivial details. The other, the high-level synthesis (HLS) method, allows the designer to abstract the details of hardware away and focus on the problem being solved. This method however cannot be used for a class of hardware design problems because the circuit's clock is also abstracted away. We present YieldFSM, a hardware description language that uses the generator abstraction to represent clock-level timing in a digital circuit. It represents a middle ground between the RTL and HLS approaches: the abstraction level is higher than in RTL, but thanks to explicit information about clock-level timing, it can be used in applications where RTL is traditionally used. We also present the YieldFSM compiler, which uses methods developed by the functional programming community -- including continuation-passsing style translation and defunctionalization -- to translate YieldFSM programs to Mealy machines. It is implemented using Template Haskell and the Clash functional hardware description language. We show that this approach leads to short and conceptually simple hardware descriptions.

Tableless Calculation of Circular Functions on Dyadic Rationals

I would like to tell a story. A story about a beautiful mathematical relationship that elucidates the computational view on the classic subject of trigonometry. All stories need a language, and for this particular story an algorithmic language ought to do well. What makes a language algorithmic? From our perspective as the functional programming community, an algorithmic language provides means to express computation in terms of functions, with no implementation-imposed limitations. We develop a new algorithm for the computation of trigonometric functions on dyadic rationals, together with the language used to express it, in Scheme. We provide a mechanically-derived algorithm for the computation of the inverses of our target functions. We address efficiency and accuracy concerns that pertain to the implementation of the proposed algorithm either in hardware or software.

Transposing G to C♯: Expressivity of generalized algebraic data types in an object-oriented language

Generalized algebraic datatypes (GADTs) are a hot topic in the functional programming community. Previously we showed that object-oriented languages such as C♯ and Java can express GADT declarations using Generics, but only some GADT programs. The addition of equational constraints on type parameters recovers expressivity. We now study this expressivity gap in more depth by extending an earlier translation from System F to C♯ to handle GADTs. Our efforts reveal some surprising limitations of Generics and provide further justification for equational constraints.

Performance challenges in modular parallel programs

Over the past decade, many programming languages and systems for parallel-computing have been developed, including Cilk, Fork/Join Java, Habanero Java, Parallel Haskell, Parallel ML, and X10. Although these systems raise the level of abstraction at which parallel code are written, performance continues to require the programmer to perform extensive optimizations and tuning, often by taking various architectural details into account. One such key optimization is granularity control, which requires the programmer to determine when and how parallel tasks should be sequentialized. In this paper, we briefly describe some of the challenges associated with automatic granularity control when trying to achieve portable performance for parallel programs with arbitrary nesting of parallel constructs. We consider a result from the functional-programming community, whose starting point is to consider an "oracle" that can predict the work of parallel codes, and thereby control granularity. We discuss the challenges in implementing such an oracle and proving that it has the desired theoretical properties under the nested-parallel programming model.

Lewis meets Brouwer: Constructive strict implication

C.I. Lewis invented modern modal logic as a theory of “strict implication” ⥽. Over the classical propositional calculus one can as well work with the unary box connective. Intuitionistically, however, the strict implication has greater expressive power than □ and allows to make distinctions invisible in the ordinary syntax. In particular, the logic determined by the most popular semantics of intuitionistic K becomes a proper extension of the minimal normal logic of the binary connective. Even an extension of this minimal logic with the “strength” axiom, classically near-trivial, preserves the distinction between the binary and the unary setting. In fact, this distinction has been discovered by the functional programming community in their study of “arrows” as contrasted with “idioms”. Our particular focus is on arithmetical interpretations of intuitionistic ⥽ in terms of preservativity in extensions of HA, i.e., Heyting’s Arithmetic.

The algebra of recursive graph transformation language UnCAL: complete axiomatisation and iteration categorical semantics

The aim of this paper is to provide mathematical foundations of a graph transformation language, called UnCAL, using categorical semantics of type theory and fixed points. About 20 years ago, Bunemanet al. developed a graph database query language UnQL on the top of a functional meta-language UnCAL for describing and manipulating graphs. Recently, the functional programming community has shown renewed interest in UnCAL, because it provides an efficient graph transformation language which is useful for various applications, such as bidirectional computation.In order to make UnCAL more flexible and fruitful for further extensions and applications, in this paper, we give a more conceptual understanding of UnCAL using categorical semantics. Our general interest of this paper is to clarify what is the algebra of UnCAL. Thus, we give an equational axiomatisation and categorical semantics of UnCAL, both of which are new. We show that the axiomatisation is complete for the original bisimulation semantics of UnCAL. Moreover, we provide a clean characterisation of the computation mechanism of UnCAL called ‘structural recursion on graphs’ using our categorical semantics. We show a concrete model of UnCAL given by the λG-calculus, which shows an interesting connection to lazy functional programming.

Iteration Algebras for UnQL Graphs and Completeness for Bisimulation

This paper shows an application of Bloom and Esik's iteration algebras to model graph data in a graph database query language. About twenty years ago, Buneman et al. developed a graph database query language UnQL on the top of a functional meta-language UnCAL for describing and manipulating graphs. Recently, the functional programming community has shown renewed interest in UnCAL, because it provides an efficient graph transformation language which is useful for various applications, such as bidirectional computation. However, no mathematical semantics of UnQL/UnCAL graphs has been developed. In this paper, we give an equational axiomatisation and algebraic semantics of UnCAL graphs. The main result of this paper is to prove that completeness of our equational axioms for UnCAL for the original bisimulation of UnCAL graphs via iteration algebras. Another benefit of algebraic semantics is a clean characterisation of structural recursion on graphs using free iteration algebra.

Practical probabilistic programming with monads

The machine learning community has recently shown a lot of interest in practical probabilistic programming systems that target the problem of Bayesian inference. Such systems come in different forms, but they all express probabilistic models as computational processes using syntax resembling programming languages. In the functional programming community monads are known to offer a convenient and elegant abstraction for programming with probability distributions, but their use is often limited to very simple inference problems. We show that it is possible to use the monad abstraction to construct probabilistic models for machine learning, while still offering good performance of inference in challenging models. We use a GADT as an underlying representation of a probability distribution and apply Sequential Monte Carlo-based methods to achieve efficient inference. We define a formal semantics via measure theory. We demonstrate a clean and elegant implementation that achieves performance comparable with Anglican, a state-of-the-art probabilistic programming system.

Structural Induction Principles for Functional Programmers

User defined recursive types are a fundamental feature of modern functional programming languages like Haskell, Clean, and the ML family of languages. Properties of programs defined by recursion on the structure of recursive types are generally proved by structural induction on the type. It is well known in the theorem proving community how to generate structural induction principles from data type declarations. These methods deserve to be better know in the functional programming community. Existing functional programming textbooks gloss over this material. And yet, if functional programmers do not know how to write down the structural induction principle for a new type - how are they supposed to reason about it? In this paper we describe an algorithm to generate structural induction principles from data type declarations. We also discuss how these methods are taught in the functional programming course at the University of Wyoming. A Haskell implementation of the algorithm is included in an appendix.

Species and functors and types, oh my!

The theory of combinatorial species , although invented as a purely mathematical formalism to unify much of combinatorics, can also serve as a powerful and expressive language for talking about data types. With potential applications to automatic test generation, generic programming, and language design, the theory deserves to be much better known in the functional programming community. This paper aims to teach the basic theory of combinatorial species using motivation and examples from the world of functional programming. It also introduces the species library, available on Hackage, which is used to illustrate the concepts introduced and can serve as a platform for continued study and research.

Functional parallel algorithms

Functional programming presents several important advantages in the design, analysis and implementation of parallel algorithms: It discourages iteration and encourages decomposition. It supports persistence and hence easy speculation. It encourages higher-order aggregate operations. It is well suited for defining cost models tied to the programming language rather than the machine. Implementations can avoid false sharing. Implementations can use cheaper weak consistency models. And most importantly, it supports safe deterministic parallelism. In fact functional programming supports a level of abstraction in which parallel algorithms are often as easy to design and analyze as sequential algorithms. The recent widespread advent of parallel machines therefore presents a great opportunity for functional programming languages. However, any changes will require significant education at all levels and involvement of the functional programming community. In this talk I will discuss an approach to designing and analyzing parallel algorithms in a strict functional and fully deterministic setting. Key ideas include a cost model defined in term of analyzing work and span, the use of divide-and-conquer and contraction, the need for arrays (immutable) to achieve asymptotic efficiency, and the power of (deterministic) randomized algorithms. These are all ideas I believe can be taught at any level.

The worker/wrapper transformation

AbstractThe worker/wrapper transformation is a technique for changing the type of a computation, usually with the aim of improving its performance. It has been used by compiler writers for many years, but the technique is little known in the wider functional programming community, and has never been described precisely. In this article we explain, formalise and explore the generality of the worker/wrapper transformation. We also provide a systematic recipe for its use as an equational reasoning technique for improving the performance of programs, and illustrate the power of this recipe using a range of examples.

Spreadsheet functional programming

AbstractThe functional programming community has shown some interest in spreadsheets, but surprisingly no one seems to have considered making a standard spreadsheet, such as Excel, work with a standard functional programming language, such as Haskell. In this paper, we show one way that this can be done. Our hope is that by doing so, we might get spreadsheet programmers to give functional programming a try.

Generalized algebraic data types and object-oriented programming

Generalized algebraic data types (GADTs) have received much attention recently in the functional programming community. They generalize the (type) parameterized algebraic datatypes (PADTs) of ML and Haskell by permitting value constructors to return specific, rather than parametric, type-instantiations of their own datatype. GADTs have a number of applications, including strongly-typed evaluators, generic pretty-printing, generic traversals and queries, and typed LR parsing. We show that existing object-oriented programming languages such as Java and C# can express GADT definitions, and a large class of GADT-manipulating programs, through the use of generics, subclassing, and virtual dispatch. However, some programs can be written only through the use of redundant runtime casts. Moreover, instantiation-specific, yet safe, operations on ordinary PADTs only admit indirect cast-free implementations, via higher-order encodings. We propose a generalization of the type constraint mechanisms of C# and Java to both avoid the need for casts in GADT programs and higher-order contortions in PADT programs; we present a Visitor pattern for GADTs, and describe a refined switch construct as an alternative to virtual dispatch on datatypes. We formalize both extensions and prove type soundness.

A simple implementation technique for priority search queues

This paper presents a new implementation technique for priority search queues. This abstract data type is an amazing blend of finite maps and priority queues. Our implementation supports logarithmic access to a binding with a given key and constant access to a binding with the minimum value. Priority search queues can be used, for instance, to give a simple, purely functional implementation of Dijkstra's single-source shortest-paths algorithm. A non-technical concern of the paper is to foster abstract data types and views. Priority search queues have been largely ignored by the functional programming community and we believe that they deserve to be known better. Views prove their worth both in defining a convenient interface to the abstract data type and in providing a readable implementation.