On 10/17/2011 9:30 AM, Black, Michael (IS) wrote:
According to this benchmark the break-even point is at 40% nulls. I asssume you have a different test?
I did mention "rule of thumb". Specific cases may vary. I must admit I'm too lazy to build tests for someone else's problem.
The fact that all NULL values are clustered together in rows with sequential rowids might have skewed the results in your test. Better locality of reference, fewer pages to read from disk, improved cache utilization.
-- Igor Tandetnik _______________________________________________ sqlite-users mailing list sqlite-users@sqlite.org http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users