I thought that I might use the foreign key to refer to several different tables (reviews, but maybe also articles and images tables that I hadn’t created yet), and then differentiate between them using a type field. Perhaps the other tables (articles, images etc) would not have numeric primary keys, so I hedged my bets and went with a varchar.

Not a great idea in hindsight, but I didn’t realise it would have such a terrible impact on performance. Basically direct lookups were fine, but large joins weren’t using indices that were available. So nasty full or partial table scans were happening, rather than using nice efficient index lookups. Once I switched the foreign key to an integer, the indices came into play, and performance was transformed.

None of this was obvious initially, as a few table scans don’t matter too much with a few thousand rows, but once you’re approaching 200,000 rows, it becomes more of an issue.

lol - how could you have flexibility beyond integer id’s if the primary key is an integer

Oh well… - most things you wrote in the past can be improved if you look back over them!!

Ian