How to Use a Library?

Containers are well-known data structures in computer science, and so are search routines. What is it, then, that makes the Standard Template Library [2,9,14] —a library of containers and search routines—a source of inspiration for programmers of various kinds? Today, there’s winSTL, comSTL, STLport, parallel STL; there are numerous STL-minded libraries (MTL, BTL, VTL, GTL, CSTL, for Matrix, Bioinformatics, View, Graph, and the Copenhagen Standard Template Library, to name but a few), and an even higher number of less acronym-conscious projects for implementing or extending STL, or expressing it in various high-level programming languages. Worthwhile also noting separately, the Boost library initiative in C++ with some 1800 developers [3]. The fascination radiating from STL and the libraries inspired by it, lies in the goal of standardization. Once a standardized computational domain exists, so the motivation of STL-like generic libraries, software engineers embrace the opportunity to resort to library routines instead of hand-crafting code, and the (meanwhile proverbial) problem of “reinventing the wheel” disappears in a natural way. However, using a library is more difficult than using a stand-alone, non-generic program. In fact, exactly those features of a library that make reuse possible, thus decrease software maintenance cost, are the ones that contribute to the complexity of library usage, thus might in the end increase its costs. In this paper, we discuss the two library composition problems we consider to be the most pressing ones in practice: the question of optimal and of consistent compositions of library components. We prepare the discussion by elaborating on the idea of standardization and its impact on the design of libraries.