Next: Appendix Up: Indexing in RELFUN Previous: A Code Generator

Summary: Heuristic Classification

There is a more global sense (than that of section 14.1) in which this paper combines heuristics and classification, providing a good scheme for this summary section.

In [Cla85] heuristic classification has been identified as a widespread problem solving method. Heuristic classification is comprised of three main phases:

  1. abstraction from a concrete, particular problem description to a problem class,
  2. heuristic match of a principal solution (method) to the problem class, and
  3. refinement of the principal solution to a concrete solution for the concrete problem.

These phases can be correlated with the phases in our indexing scheme:

  1. abstraction from a RELFUN procedure resulting in the relevant head information, the argument columns (see section 11 and the function (``make index type head'' in appendix D),
  2. applying the MV-Algorithm (or one of the other heuristics) resulting in a DAX, and
  3. using the code generator to produce the concrete solution, the WAM code.

Michael Sintek -