Bug 9717 – `std.math.round` rounds away from zero instead of to the nearest even integer

Rounding to the nearest even integer is a default rounding mode and is an expected, useful (e.g. for mesurements rounding), and documented behaviour for `std.math.round`. But instead it currently rounds away from zero (it even uses C's `roundl`).

Do you have an example code for this bug? I did this: void main() { writeln(std.math.round(1.4)); } and the output is 1, as expected.

(In reply to comment #1) > Do you have an example code for this bug? I did this: > > void main() { writeln(std.math.round(1.4)); } > > and the output is 1, as expected. I believe the "nearest even" is for the .5 tie-breakers: EG: round(0.5) == 0 round(1.5) == 2 round(2.5) == 2 It sounds strange at first, but it keeps things "sign agnostic" and nor shifts the average. http://en.wikipedia.org/wiki/Rounding#Round_half_to_even

That said, "nearest half" is one scheme among many other. I'm not sure about "default".

(In reply to comment #3) > That said, "nearest half" is one scheme among many other. I'm not sure about > "default". I mean the function is documented to behave this way: http://dlang.org/phobos/std_math.html#round Also I haven't seen examples where you have to use another theme but I saw ones where you have to use "half to even" theme.

Both the C function[0] and the CRuntime_Microsoft version [1] actually explicitly do rounding away from zero rather than to even. There are other functions in std.math which round using the current rounding mode, which defaults to even, so this is most likely just a documentation error. [0]: www.codecogs.com/library/computing/c/math.h/round.php#sec1 [1]: in std/math.d

Commits pushed to master at https://github.com/D-Programming-Language/phobos https://github.com/D-Programming-Language/phobos/commit/91c38b4bfda60278924140e7bded63f56903a15e fix issue 9717 - `std.math.round` rounds away from zero instead of to the nearest even integer https://github.com/D-Programming-Language/phobos/commit/0c39d2d341eac3f3c81f6608dedd25ee2689d576 Merge pull request #3849 from Infiltrator/patch-1 fix issue 9717 - `std.math.round` rounds away from zero instead of to the nearest even integer

This issue isn't fixed. It's state is WONTFIX. I'm strictly against documenting current useless specification as people either don't care about rounding or prefer bank rounding method as it is less biased.

(In reply to Denis Shelomovskij from comment #8) > This issue isn't fixed. It's state is WONTFIX. > I'm strictly against documenting current useless specification as people > either don't care about rounding or prefer bank rounding method as it is > less biased. There's rint/lrint?

(In reply to Iain Buclaw from comment #9) > (In reply to Denis Shelomovskij from comment #8) > > This issue isn't fixed. It's state is WONTFIX. > > I'm strictly against documenting current useless specification as people > > either don't care about rounding or prefer bank rounding method as it is > > less biased. > > There's rint/lrint? Yes, there is a way to do it: ensure default rounding mode is set and call `rint`. But there is still two function calls (i.e. one `with(FloatingPointControl) assert(rounding == roundToNearest)`) if you control rounding mode and even more calls if you don't. My point is: why are we using C behavior for `round` when it is not a generally wanted behaviour? It's like keeping C issues in D and I don't see any point of it.

Commits pushed to stable at https://github.com/D-Programming-Language/phobos https://github.com/D-Programming-Language/phobos/commit/91c38b4bfda60278924140e7bded63f56903a15e fix issue 9717 - `std.math.round` rounds away from zero instead of to the nearest even integer https://github.com/D-Programming-Language/phobos/commit/0c39d2d341eac3f3c81f6608dedd25ee2689d576 Merge pull request #3849 from Infiltrator/patch-1