|
|
Posted by Anthony Tippett on 04/05/05 04:14
Thanks to everyone that helped. After further googling, bug 9288 had a
good explanation of what was going on ( which is not a bug)
http://bugs.php.net/bug.php?id=9288&edit=3
I'll just include the answer for anyone that comes upon this tread on a
search engine.
[15 Feb 2001 2:55pm CET] hholzgra at php rot dot fot net
short answer:
never use floats or doubles for financial data!
longer answer:
the internal number format in modern computers
is binary (base 2) and not decimal (base 10) for performance
and complexity reasons
while it is possible to convert decimal numbers into binaries
and back this does not hold true for fractions
something like 0.3 (decimal) would be a periodic binary
fraction like 10/3 is 0.33333333333333333333...
in decimal
this leads to loss of precision when calculation with
decimal fractions as you have when storing currency values
solution:
if you just summ up values then you should store values in
the smalest unit your currency has (pennies?) instead of
what you are used to (Pounds?) to totally avoid fractions
if you cannot avoid fractions (like when dealing with
percentage calculations or currency conversions) you
should just be aware of the (usually very small) internal
conversion differences
(0.000000000000000027755575615629 in your example)
or use the bcmath extension, although for monetary
values you should go perfectly fine with using round(...,2)
on your final results
Anthony Tippett wrote:
> btw, thanks for your response.
>
> I'm not sure as if I understand why. It's not like I'm using a very
> precise number when dealing with the hundreths place.
>
> Even without the multiplication the number gets messed up.
>
> eg.
>
> $a = 17.00;
> $a+= 1.10;
> $a+= 0.32;
> $a+= 0.07;
>
>
> print $a."<br>"; // 18.49
>
> var_dump($a); // float(18.49)
> var_dump($a-18.49); // float(3.5527136788005E-15)
>
> I'm just trying to add money amounts? Can I not rely on floats to do this?
>
>
>
> Richard Lynch wrote:
>
>>Floats are NEVER going to be coming out "even" reliably.
>>
>>You'll have to check if the difference is less than X for whatever number
>>X you like.
>>
>>Or you can look at something like BC_MATH where precision can be carried
>>out as far as you like...
>>
>>But what you are seeing is to be expected.
>>
>>That's just the way computers work, basically.
>>
>>On Mon, April 4, 2005 5:07 pm, Anthony Tippett said:
>>
>>
>>>Ok i've narrowed it down a little bit but still can't figure it out..
>>>
>>>Here's the code and what I get for the output. Does anyone know what's
>>>going on? Can someone else run it on their computer and see if they get
>>>the same results?
>>><?php
>>>
>>>$a = "17.00" * "1";
>>>$a+= "1.10" * "1";
>>>$a+= "0.32" * "1";
>>>$a+= "0.07" * "1";
>>>
>>>print $a."<br>"; // 18.49
>>>
>>>var_dump($a); // float(18.49)
>>>var_dump($a-18.49); // float(3.5527136788005E-15)
>>>?>
>>>
>>>
>>>Anthony Tippett wrote:
>>>
>>>
>>>>I'm having trouble figuring out why subtraction of two floats are giving
>>>>me a very small number. I'm thinking it has something to do with the
>>>>internals of type casting, but i'm not sure. If anyone has seen this or
>>>>can give me some suggestions, please.
>>>>
>>>>I have 2 variables that go through a while loop and are
>>>>added/subtracted/ multipled. InvAmt and InvPay are shown as
>>>>floats but when they are subtracted, they give me a really small
>>>>number
>>>>
>>>>// code
>>>>var_dump($InvAmt);
>>>>var_dump($InvPay);
>>>>var_dump($InvAmt-$InvPay);
>>>>
>>>>// output
>>>>float(18.49)
>>>>float(18.49)
>>>>float(2.1316282072803E-14)
>>>>
>>>
>>>--
>>>PHP General Mailing List (http://www.php.net/)
>>>To unsubscribe, visit: http://www.php.net/unsub.php
>>>
>>>
>>
>>
>>
>
[Back to original message]
|