I still remember tracing down an odd bug involving taxes and contract calculations that resulted from a programmer I thankfully couldn't identify using a single precision float to do calculations that involved unit of measure conversions that then had calculations involving contract terms and weights. The original unit stored was a decimal with 30 total digits and 10 decimal digits. When I finally found the part of the code where decimals suddenly because floats, I was inclined to do violence given it was 3AM and I was chasing a couple of hundred thousand pennies in various places. It was deep in the middle of a stored procedure so the standard excuses don't apply.

I swear there are not enough classes about numbers.