On his blog, Mike Lively has a little reminder for developers out there about working with floating point numbers in your applications, specifically their accuracy.

I know many of you all know pretty well that floating point precision and computers don't play nicely with each other. [...] I was working with a piece of code today at the office that was throwing an error saying two values weren't zeroing out when they clearly should have been.

In his code he shows the simple bit that he was using to evaluate if the result of subtraction would come to a certain value. In the comments, others share similar experiences and one even recommends another option to make things a bit more accurate - the bcmath library.