On Thursday, 25 June 2015 at 09:24:28 UTC, DLearner wrote:
A signed decimal datatype to support writing userland
accounting programs (major business application area), without
loss of accuracy (including losses caused by not all 'decimal
decimals' having exact 'binary decimal' representation).
Syntax:
dec(a,b) foo;
Where a is an integer 1<=a<=99 (total no of digits, including
any decimal digits).
b is also an integer 0 <= b <= a (no. of decimal digits);
b = 0 implies b can be omitted e.g. 'dec(4,0) foo;' can also be
written 'dec(4) foo;'.
Basic arithmetic supported.
Normal assignment rule is truncation/padding, but a 'round(d)'
function to round to d decimal places provided.
Also provide a formatting function to convert dec type to
printable character string.
Not very interesting from a theoretical viewpoint (just COBOL
or PL/I datatype),
but (I think) of practical value in encouraging D adoption.
Decimal data type is something that can and should be done in
library, not in the language itself. There is already a proposed
implementation, you too can review it:
http://wiki.dlang.org/Review_Queue
https://github.com/andersonpd/eris/tree/master/decimal
