A more functionrich number (than Bigdecimal) holding java library

• The operation should try and keep the answer of the same type as the original inputs, thus if the holder started of with a Double then the result should also be a double.

• If inputs contain holders of different number types then it should auto widden etc.

• There should be a single type of Holder with getters to return the result as an integer, double, bigdecimal, big integer etc with exceptions thrown if conversion fails.

• Code should look like Bigdecimal.

• Ideally immutable

• Code shouldnt know or care whats inside the holder, operations just work until conversion is needed at a later stage.

• Unfortunately BigDecimal is not really complete for my needs its missing many key functions eg: sine, log and most of the static helpers on Math.

ApacheCommonsMath

Axelcb made a suggestion to use the apache commons math library. The main class of use in my case seem to be DFP.

• Examining DFP there is no simple way to pass a BigDecimal/BigInteger and build a DFP. It would be nice if precision would also be a parameter and used during the suck in process.

• none of the functions (like multiply) accept a context w/ a precision and rounding just like BigDecimal. I really dont understand the reasoning for the clunky way precisions and rounding are handled.

• What is a DfpField, what exactly does it do and why are there fields in a DFP ??? Please dont tell me this Calendar again.

• there are no methods to export to a BigDecimal or BigInteger

dfp

• is this the original inspiration for apache commons math DFP class ?

• the first concern about import from a BigDecimal or BigInteger is lacking.

开发者_Python百科

• there are no methods to export to a BigDecimal or BigInteger

You might try the package gnu.math from the kawa language framework which is used to implement the full Scheme "numeric tower". The package should be pretty self-contained and independent from the rest of kawa.

BigDecimal is not final. You can extend it and add the methods (sine, log, etc) to the subclass.

I know these libraries which may be usable for you:

• High Performance Arbitrary Precision Arithmetic Package
• Decimal Floating Point Java class library which is now part of the Commons-Math 2.2 API
• JScience Real Implementation

I think the best way is to write wrapper for BigDecimal and implement all your additional options.