If dealing with money in a float is bad, then why does money_format() do it?

问题

I've been waffling on how to deal with currency display and math in PHP, and for a long time have been storing it in MySQL using the DECIMAL type, and using money_format() to format it for display on the web page. However, today I looked at the actual prototype:

string money_format ( string $format , float $number )

I'm a little confused now. All I've been told is, avoid floats for money! But here it is, the fundamental formatting function (say that five times fast), casting the input to a float. number_format() does the same.

So my questions are:

Unless I'm dealing with fractional cents or trillions of dollars (and I'm dealing with neither), should I be concerned at all about displaying and storing (but never doing math on) currency that's been cast to a float? Will I ever come close to the area of having floating point inaccuracies change my figures?
If the answer to #1 is that I should indeed be concerned, then why is money_format() built this way?

回答1:

Unless I'm dealing with fractional cents or trillions of dollars (and I'm dealing with neither), should I be concerned at all about displaying and storing (but never doing math on) currency that's been cast to a float? Will I ever come close to the area of having floating point inaccuracies change my figures?

For pure rounding/display purposes, you're safe as long as the absolute floating-point representation error is less than $0.005 (so that rounding to the nearest cent is correct).

With IEEE 754 single-precision, you're safe up to $131,072.00. ($131,072.01 is represented as 131072.015625, which incorrectly rounds up.)

Double precision (which PHP's float uses) doesn't fail until $70,368,744,177,664.01 (which also has .015625 for the cents). You have nothing to worry about.