Is there any performance gain in indexing a boolean field?

Question

I\'m just about to write a query that includes a WHERE isok=1. As the name implies, isok is a boolean field (actually a TINYINT(1) UNSIGNED

Answer 1

It depends on the distribution of the data.

Imagine I had a book with 1000 closely typed pages, and the only words in my book were 'yes' and 'no' repeated over and over and distributed randomly. If I was asked to circle all the instances of 'yes', would an index in the back of the book help? It depends.

If there was a half-and-half random distribution of yes's and no's, then looking up in the index wouldn't help. The index would make the book a lot bigger, and anyway I'd be quicker just to start from the front and work my way through each page looking for all the instances of 'yes' and circling them, rather than looking up each item in the index and then taking the reference from the index entry to the page that it refers to.

But if there were, say, just ten instances of 'yes' in my thousand-page book and everything else was just millions of no's, then an index would save me loads of time in finding those ten instances of 'yes' and circling them.

It's the same in databases. If it's a 50:50 distribution, then an index isn't going to help - the database engine is better off just ploughing through the data from start to finish (full table scan), and the index would just make the database bigger, and slower to write and update. But if it is something like a 4000:1 distribution (as per oucil in this thread), then an index seek can speed it up hugely, if it is the 1 in 4000 items that you are looking for.