Order Preserving Key Compression

Order preserving data compression can improve sorting and searching performance, and hence the performance of database systems. We describe a new parsing (tokenization) technique that can be effectively applied to "keys", producing substantial compression. It can be applied to both the uncompressed and compressed strings, permiting a more flexible choice of variable length dictionary entries and their compressed forms. The key notion is that the way that entries partition the space of strings determines how the entries are to be parsed. The result is economy in the number of encodings that are needed in order to preserve order and great flexibility in how they are encoded. We illustrate one use of order preserving compression, padding character compression for multifield keys. This demonstrates the dramatic gains possible with the new technique, some of which cannot be achieved by any other method.