Deal with decimals using PreciseMoney

[mathiasverraes]

idea from http://joda-money.sourceforge.net/userguide.html

There are two main classes which are both based on BigDecimal:
Money - an amount in a single currency where the decimal places is determined by the currency
BigMoney - an amount in a single currency where there is no restriction on the scale

For example, the currencies of US dollars, Euros and British Pounds all use 2 decimal places, thus a Money instance for any of these currencies will have 2 decimal places. By contrast, the Japanese Yen has 0 decimal places, and thus any Money instance with that currency will have 0 decimal places.

Instances of BigMoney may have any scale that can be supported by BigDecimal.
Conversion between a Money and a BigMoney can be performed using the methods toBigMoney() and toMoney(). The latter optionally takes a rounding mode to handle the case where information must be lost.

[mathiasverraes]

using bcmath?

[sa-tasche]

using gmp?

[camspiers]

Related to this, there is a fairly big issue with integer overflow in this library. Although I really like your library Mathias, I choose to use https://github.com/sebastianbergmann/money for various reasons, but I still want to help this library.

If you have a look at my recent pull requests:

• Provide new functionality to ensure that internal operations don't disguise imprecision sebastianbergmann/money#9
• Throw more informative exceptions when addition and subtract cause overflow sebastianbergmann/money#13

This library suffers from the same issues, either silently (and wildly) losing imprecision, or actually failing but not really given appropriate errors.

Regards

[mathiasverraes]

see https://github.com/mathiasverraes/money/pull/55/files#r16972250

[josecelano]

Which BigDecimal class are you planning to use?
I am working on an implementation using https://github.com/Litipk/php-bignumbers

[josecelano]

@mathiasverraes I have done a test implementing Money class using a BigDecimal library (https://github.com/Litipk/php-bignumbers) I have only changed amount type from integer to Decimal.

Here is the test:
https://github.com/josecelano/money/blob/master/lib/Money/BigMoney.php

I call it BigMoney but I only want to use your libary with big integers (I do not need to scale over the currency default decimal places). Then I suppose it is a new implementation of Money. Althougth I think it could be easily changed to implement your BigMoney.

Just after the commit I have found this new library: https://github.com/keiosweb I suppose I can use it for all possible values (int, big int and more presicion). Do you think it's better to have only one class rather than 2 as you propose? I think it is a good idea to have two but the first one (Money) should support big amounts. I suppose there could be countries with hyperinflation where you can easily overflow max integer in 32 bits systems.

[sagikazarmark]

@josecelano big integer values are supported now on the nextrelease branch.

[josecelano]

OK thanks @sagikazarmark but I do not see any big integer in that brancg. It seems internal amount attribute is still an integer.

[sagikazarmark]

Integer support is implemented as Calculators. If you have gmp or bcmath installed, it will automatically use those which allows you to represent big integers.

[josecelano]

@sagikazarmark I have not had time to test this new verion but there is something I still do not understand, you are using big integer functions for calculations but you are still using an integer to store the amount attribute, then you still have the integer max size limit. For example if I want to add two Money objects and the result is greater than max integer value (in 32 or 64 bit systems), then the result does not fit in an integer.

GMP calculator add method returns a string (gmp_strval) but Money constructor expects an integer.

[sagikazarmark]

@josecelano that's what the docblock says, but look at the validation:

https://github.com/mathiasverraes/money/blob/nextrelease/src/Money.php#L57

There is a PR which fixes that docblock as well.

[frederikbosch]

@josecelano This is fixed in #133.

[TNAJanssen]

What is the status of this issue?