The purpose of this paper is to compare and contrast the modeling capabilities of seven algebraic modeling languages (ML) available today, namely, AMPL, GAMS, LINGO, LPL, MPL, PC-PROG and XPRESS-LP. In general, these MLs do an excellent job of providing an interface with which the modeler can specify an algebraically formatted linear program (LP). That is, each ML provides a substantial improvement in time and convenience over the matrix generator/report writers of the last few decades. Further, each of the MLs provides: (1) significant flexibility in model specification, instantiation and modification, (2) effective and efficient conversion from algebraic to solver format, and (3) an understandable and, for the most part, self-documenting model representation. In addition, each of the MLs is constantly being updated and upgraded to provide additional capabilities sought by practitioners and users. However, as shown in the fifteen tables provided in the body of this paper, each ML has its own set of competitive advantages. For example, the most integrated environments (i.e. those integrating the modeling language with a full-screen editor, data import capabilities and a solver) are provided by LINGO and PC-PROG. The most user-friendly interfaces are provided by MPL and PC-PROG, both of which provide window-based interfaces to create models and pop-up windows to display error messages; MPL also uses pull-down menus to specify various operations, whereas PC-PROG uses function keys for operational control. Package costs are led by a current (March, 1991) introductory offer from LINGO. Modeling effectiveness, especially with respect to flexibility in specifying arithmetic statements, is led by GAMS and LPL. Model compactness, as measured by the number of lines required to specify a model, is led by AMPL, LPL, MPL and PC-PROG; LPL, MPL and PC-PROG also provide context sensitive editors which automatically position the cursor where the error was detected. And finally, the most comprehensive user documentation is provided by GAMS, whereas GAMS, LINGO and LPL provide extensive libraries of sample models for those users who “learn by example”.
Read full abstract