Floating-Point Pitfalls

Floats are not exact. This surprises beginners, but it happens in every programming language, because computers store numbers in binary.

1
2
3

<?php
var_dump(0.1 + 0.2);
var_dump(0.1 + 0.2 === 0.3);

1
2

float(0.30000000000000004)
bool(false)

Just as 1/3 cannot be written exactly in decimal (0.3333…), 0.1 cannot be written exactly in binary.

Echo hides the problem

echo rounds to a precision of 14 digits, so the problem is not visible.

1
2

<?php
echo 0.1 + 0.2;

1

0.3

Do not trust what echo shows. The stored number is slightly different.

A famous trap

1
2
3
4
5

<?php
$value = (0.1 + 0.7) * 10;

echo (int) $value . "\n";
var_dump($value);

1
2

7
float(7.999999999999999)

You expected 8. The cast cut the number off at 7. When you must turn a float into a whole number, round it first.

1
2
3
4

<?php
$value = (0.1 + 0.7) * 10;

echo (int) round($value);

1

8

Comparing floats

Never compare floats with ==. Check whether they are close enough.

1
2
3
4
5
6
7

<?php
function almostEqual(float $a, float $b, float $epsilon = 1e-9): bool
{
    return abs($a - $b) < $epsilon;
}

var_dump(almostEqual(0.1 + 0.2, 0.3));

1

bool(true)

Money

Do not store money as floats. Two safe options:

1. Integers in the smallest unit.

1
2
3
4
5
6
7

<?php
$price = 1999;      // 19.99 stored as 1999 cents
$quantity = 3;

$total = $price * $quantity;

echo number_format($total / 100, 2);

1

59.97

2. BCMath, a PHP extension that does exact decimal maths on strings.

1
2
3

<?php
echo bcadd("0.1", "0.2", 2) . "\n";
echo bcmul("19.99", "3", 2);

1
2

0.30
59.97

Huge integers

Integers above PHP_INT_MAX become floats and lose digits. For exact big numbers, use bcmath or gmp.