CODEWITHSUNDEEP
mathhaskell
MathematicalOrchid
MathematicalOrchid

Reputation: 62848

Arbitrary precision for non-integers

Haskell has a built-in Integer type, which handles arbitrary-precision integers. There is also Rational, which is an arbitrary-precision fraction. But arithmetic on such things requires finding a common denominator, and then cancelling the result down to least terms.

What if I wanted to do floating-point arithmetic with (say) 100 bits of precision in the mantissa? How would I do that?

I see there's a Data.Fixed module, but that seems to provide a handful of custom-written types with fixed precision. What I want is something where I can dynamically increase or decrease precision at run-time, according to how much accuracy is required for each task.

PS. I'm not looking for decimal arithmetic, although I suppose it would be interesting to know whether that's available somewhere...

Upvotes: 8

Views: 362

Answers (2)

augustss
augustss

Reputation: 23014

Try Data.Number.CReal from the numbers package. It gives you the precision you ask for when converting to a string.

Upvotes: 4

user3374972
user3374972

Reputation: 1

How about http://hackage.haskell.org/package/numbers-2009.8.9/docs/Data-Number-BigFloat.html ? In this BigFloat data type you choose precision on your own. I know it's decimal, but maybe you can easily change it to binary

Upvotes: 0

Related Questions