Assuming that performance is not an issue, and knowing that I only want to use ONE type for all the numbers I have in my program, which type will be the correct one to use? I think that BigDecimal is probably the best option, but I'm not sure of this.
Asked
Active
Viewed 181 times
1
vainolo
- 6,907
- 4
- 24
- 47
-
3I am afraid this is a primarily opinion based question, and probably too broad to answer. BigDecimal is robust enough to handle almost anything your likely to throw at it, but we have no idea what your application does or needs to do. – Rudi Kershaw Feb 28 '14 at 14:06
-
you answeared yourself - BigDecimal – Leos Literak Feb 28 '14 at 14:07
-
Maybe you functions should be generic enough to work with any kind of number instead of working with "the" number type that doesn't exist. BigDecimal can't even represent 1/3 correctly, let alone pi, sqrt(-1), a quaternion or most surreal numbers. – Niklas B. Feb 28 '14 at 14:09
-
Why do you need only **one** type for representing numbers? – rpax Feb 28 '14 at 14:12
-
3String. Can represent everything. :-) – Seelenvirtuose Feb 28 '14 at 14:22
-
@Seelenvirtuose that's very hardcore, don't you think? +1 – rpax Feb 28 '14 at 14:24
-
1Write the code to accept `java.lang.Number` as input. Then the caller who knows the number can choose the optimal number type for the value, which might be `BigDecimal` but doesn’t have to. – Holger Feb 28 '14 at 15:39
2 Answers
7
It depends on how you want to use the numbers.
If you are unlikely to need floating point and the numbers aren't going to get too big use int as the most efficient use of the CPU.
If the numbers might get too big for int use long. Uses twice the memory but still cpu efficient.
If too big for long use BigInteger. Uses the optimum number of bytes for the number. Breaks when number > about 2^65535 so works for most scenarios. Slower than int or long but still seriously fast.
If too big for BigInteger you're on your own. Perhaps one day ...
If you need float maths float or double are good so long as you are working in their range.
For big float maths or a more predictable calculation than double or float, use BigDecimal. Ha sits limitations like BigInteger but it is rare to be bitten by them.
Beyond that if you are working solely in rationals then you may find a rational number class useful and more accurate than BigDecimal. This looks ok.
If you are likely to move into irrational territory you may find some of the continued-fractions classes of use. Here is an example.
OldCurmudgeon
- 64,482
- 16
- 119
- 213
1
You are correct. If you only want to use one type (which excludes BigInteger ),
you should use BigDecimal.
Here is an explanation : The need of BigDecimal
And you can also check this answer: Double vs. BigDecimal?
