Column oriented compilation of variant tables

The objective of this work is to improve variant table evaluation in a configurator by compiling/compressing the tables individually to reduce both processing time and space. The main result is a proposed simple heuristic for decomposing the variant table into subtables and a representation linking these subtables in a directed acyclic graph (DAG). The size of the compression obtained by this heuristic for examples used in [2, 10] is comparable to that achieved there. However, a formal analysis of complexity has not yet been completed. A prototype implemented in Java exists. Objectives in designing it were to keep it completely decoupled from any particular configurator, while using little machinery in order to keep software maintenance costs low. Testing both on abstract examples and on tables that resemble real customer data is ongoing and looks promising. Non-atomic table cells (such as real intervals, or value sets) are supported. My approach to negative variant tables [8] has been incorporated into the implementation.