WL#407: Boyer-Moore optimization of LIKE for POS and INSTR functions

Affects: Server-7.1   —   Status: Un-Assigned

Boyer-Moore (sp?) optimisation of LIKE that should be
enabled also by the POS and INSTR functions.

Monty has done a good optimisation of LIKE. If you have a long TEXT field
called MYFIELD and need to look for the records with occurrences of the
text string "my value", Monty implements the screening test


as follows: 

1. He lists the eight characters " aelmuvy" as the only
   ones existing in the string he looks for

2. He looks only at positions 8, 16, 24 (multiples of the
   search string length) to see whether that letter is
   one of " aelmuvy". 

3. If not, he just skips forward at eightfold speed. 

4. Otherwise, he looks whether the other parts of the string 
   would match.

Now, Monty told me we could add a feature to LIKE to tell the end user at
which position the text was found. "But we have POS and INPUT for that", I
said. Then he remembers he didn't implement Boyer-Moore optimisation for
POS and INSTR, which is the task.