I was discussing inefficient indexes with a colleague today. We have a table with many indexes on inefficient columns. It has not become an issue (at least not an obvious one) since we seldom delete more than one record at a time.
There apparently was no way to avoid indexing these columns as they are used heavily in various queries. Today I had a brainstorm. I read recently that some types of SQL Server indexes add a hidden "uniquifier" to each table and include it in their indexes behind the scenes so even an index on a mostly null column will not cause problems. The obvious parallel in RBASE would be to change all these inefficient indexes I mentioned to multi column indexes with the table primary key as the second column in each. My tests indicate that this does not seem to affect performance negatively. My BIG concern is whether having the primary key included in so many multi-column indexes would cause problems in its own right. If it would not, it would be a sweet solution. Has anyone tried such a scheme? Any feedback, especially from Razzak, would be appreciated. Dennis McGrath
