Finite maps are ubiquitous in many applications, but perhaps nowhere more so than in compilers and other language processors. In these applications, three operations on finite maps dominate all others: looking up the value associated with a key, inserting a new binding, and merging two finite maps. Most implementations of finite maps in functional languages are based on balanced binary search trees, which perform well on the first two, but poorly on the third. We describe an implementation of finite maps with integer keys that performs well in practice on all three operations. This data structure is not new -- indeed, it is thirty years old this year -- but it deserves to be more widely known.

Edison is a library of functional data structures implemented in Haskell. It supports three main families of abstractions: sequences, collections (e.g., sets and priority queues), and associative collections (e.g., nite maps). This paper summarizes the design of Edison, with particular attention to how that design is inuenced by details of Haskell. 1 Introduction There is a growing recognition that a useful set of libraries is at least as important to the acceptance of a programming language as the design of the language itself. A library of fundamental data structures such as queues, sets, and nite maps is particularly important in this regard. However, high-quality examples of such libraries, such as the STL [14] in C++ or the the collection classes [3] in Smalltalk, are rare. Edison is a library of ecient data structures suitable for implementation and use in functional programming languages. It is named after Thomas Alva Edison and for the mnemonic value of EDiSon (Ecient Data ...