On Thu, 9 Jul 2015 22:28:04 +0100 Simon Slavin <slavins at bigfraud.org> wrote:
> If you're going to do it properly you have a keywords column which > contains strings like > > , > ,animal, > ,animal,predator, > ,animal,predator,bird, Why do it that way? I would recommend a schema like Keith's, with one keyword per row in a separate table. The OP mentioned that searching 250,000 rows took 2 seconds. Because a LIKE clause can't use an index, every query of a column such as you suggest will require a linear scan, plus pattern-matching overhead. A normalized plan like Keith's permits indexing by keyword, which should be blazing fast when testing for equality to string constants. --jkl