Use Prolog as Host Language?

The last example shows that Prolog syntax (and semantics) and finite domains go quite well together. Actually, it is more than that: due to the incremental nature of constraint programming (prototyping and building an application incrementally is easy and natural), the availability of interactive interpreters for CLP languages (inherited from Prolog) is a plus, as experimentation and debugging are parts inherent to the development of a program.

Also, the built-in backtracking of logic programming allows the easy customization of search procedures for the cases in which standard CLP procedures are not good enough: this may happen when there are hints as to what is the best direction to search in. Small examples might not show that, due to small search times, but large examples often make the difference apparent.

Some interesting characteristics from Prolog are also inherited, which are not found in other languages:

• A built-in database, which can be used (with caution) to implement global variables, but whose main strength is in saving intermediate results which do not need to be recomputed (lemmas) and, in the extreme, to generate and change program code dynamically.
• Meta-programming facilities, which allow the program to be managed as if it were data, examine its code while running, calling goals and collecting the solutions produced on backtracking, and other goods which are only available to logic programming.
• Easy definition of meta-languages and easy developing of interpreters for those languages. This allows the user to create a high-level language suited for her/his needs, with which developing the final application will be easier, and to code an interpreter for such a language.

Of course, there are some disadvantages in using Prolog as host language, mainly concerned with the difference of the logic programming paradigm with respect to other paradigms:

• It might be not as well accepted as other languages: Prolog is sometimes not part of standard curriculum in Computer Science, and therefore some training is usually needed, and some programmers might be reluctant to undertake learning a new paradigm.
• There are notable differences with respect to conventional languages in the way data structures are handled: but these differences, in the end, favor the programmer, for they turn out to be easier to work with and to define, and more secure in what respect errors caused by illegal memory accesses, etc. Control is also quite different: the embedded search, once understood, is a very powerful way of programming.

Last, but not least, there are different products which implement the CLP paradigm. Depending on the problem some of them might be more adequate than others. But very probably the final application in an industrial environment will have to interact with other programs, so the possibility of having an interface (other than a raw text file) is a point to take into account. Fortunately this is the case for all commercially available Prolog and CLP systems.