Implement OpAssign for Ratio #12
Conversation
I think it should though, because un-reduced ratios make overflow (#13) even worse. To call I notice that you added integer support, |
|
Thanks! Although it costs a few more instructions, |
|
Can you merge or rebase your branch? I think your changes are colliding with #11. |
src/lib.rs
Outdated
| @@ -1253,17 +1449,6 @@ mod test { | |||
| fn test_div_0() { | |||
| let _a = _1 / _0; | |||
| } | |||
|
|
|||
| #[test] | |||
| fn test_checked_failures() { | |||
There was a problem hiding this comment.
It seems this test was accidentally removed in your merge?
There was a problem hiding this comment.
Wow! That was bad of me :(
|
Now you pulled in a (If you have trouble with this, I may be able to push such an update to your PR myself.) |
|
I rewrote history to make things cleaner. |
|
Great, thanks! bors r+ |
12: Implement OpAssign for Ratio r=cuviper a=c410-f3r Copied shamelessly from [#274](rust-num/num#274) but fixes #5. There is only one problem to solve before anything. Since the implementation doesn't use the `reduce` function, it is possible to exist multiple types of zeros (`0/4`, `0/21`, `0/66`) instead of the default `0/1`, which implies in test errors like `test(_1_2, _NEG1_2, _0);`, therefore, should I rewrite the test suite to be more "zero" friendly?
Build succeeded |
Copied shamelessly from #274 but fixes #5.
There is only one problem to solve before anything. Since the implementation doesn't use the
reducefunction, it is possible to exist multiple types of zeros (0/4,0/21,0/66) instead of the default0/1, which implies in test errors liketest(_1_2, _NEG1_2, _0);, therefore, should I rewrite the test suite to be more "zero" friendly?