Bcmul reporting 0

Question

I have a simple piece of code as below.

$amount = 447274.44882;
$rate = 0.00001;

echo floatNumber(bcmul($amount, $rate, 8), 8);

This outputs 0.00000000 when it should be 4.47274449. If I change the rate to 0.0001 then it outputs the correct number, anything higher than 4 decimals and it reports 0.

Am I doing something wrong or is this a known limitation or something? Seems quite a big one if that's the case.

Even without seeing what `floatNumber()` does: using bcmath with floats instead of strings completely beats the purpose. — Álvaro González, Feb 06 '14 at 16:53
try using sprintf http://in1.php.net/sprintf here is almost similar situation like you http://stackoverflow.com/questions/21598667/php-the-awk-subtraction-giving-exponential-values/21599099#21599099 — Abhik Chakraborty, Feb 06 '14 at 16:53
It does? I was hoping it would solve my rounding errors. I'll try it with strings and see if it works. — Linkandzelda, Feb 06 '14 at 16:54
I changed it to echo sprintf('%f', (bcmul($amount, $rate, 8))); and its still 0, also tried casting to string and that was 0 too. — Linkandzelda, Feb 06 '14 at 16:57
But physically changing it to the number in a string wrapped with '' works... — Linkandzelda, Feb 06 '14 at 16:58
@Abhik But that results in rounding errors, which is what I'm trying to stop from happening. Maybe not with this example but a log of other different combinations. — Linkandzelda, Feb 06 '14 at 17:04
Are you aware that `bcmul()` doesn't round? `bcmul('20.01', '1.444', 1)` (28.89...) produces `28.8`. — Álvaro González, Feb 06 '14 at 17:16

Álvaro González · Accepted Answer · 2014-02-06T17:21:02.800

If you cast 0.00001 to string using the default settings (and that's what will happen if you feed bcmul with floats since it expects strings) you'll get this:

var_dump( (string)0.00001 );

string(6) "1.0E-5"

It isn't clearly documented but bcmath functions apparently ~~return~~ cast to zero when faced to invalid input:

var_dump( bcadd('Hello', 'world!', 8) );
var_dump( bcadd('33', 'Foo', 8) );
var_dump( bcdiv('33', 'Foo', 8) );

string(10) "0.00000000"
string(11) "33.00000000"
Warning: bcdiv(): Division by zero
NULL

The whole idea of arbitrary precision libraries is to overcome the limitations of base 2 arithmetic and fixed size storage. Thus you'd need this:

var_dump( bcmul('447274.44882', '0.00001', 8) );

string(10) "4.47274448"

This is great to do math with 100-digit numbers but not particularly useful for simple rounding. In fact, the extension doesn't round at all—it merely truncates:

var_dump( bcmul('20.01', '1.444', 3) );
var_dump( bcmul('20.01', '1.444', 2) );
var_dump( bcmul('20.01', '1.444', 1) );
var_dump( bcmul('20.01', '1.444', 0) );

string(6) "28.894"
string(5) "28.89"
string(4) "28.8"
string(2) "28"

This answer is very elaborate, thanks. I still don't know if using BCmul even when casting between strings will result in inaccurate calculations with large 8 digit floating point numbers. — Linkandzelda, Feb 06 '14 at 17:28
You *can* get inaccurate calculations due to how IEEE floating point precision works. But most of the times it happens in real life it's because of rounding in the wrong place (or not rounding at all). And you can always use *integers* (e.g., cents instead of euros). — Álvaro González, Feb 06 '14 at 17:39

Bcmul reporting 0

1 Answers1