`rem(x, y, r)` - Missing implementation for rounding mode `RoundNearestTiesUp` and `RoundNearestTiesAway`

[lvlte]

Integer, float and rational types don't have an implementation of rem(x, y, r) with RoundNearestTiesUp and RoundNearestTiesAway rounding modes, eg. with RoundNearestTiesUp :

julia> rem(5.0, 2.0, RoundNearestTiesUp)
ERROR: MethodError: no method matching rem(::Float64, ::Float64, ::RoundingMode{:NearestTiesUp})
...

julia> rem(5, 2, RoundNearestTiesUp)
ERROR: MethodError: no method matching rem(::Int64, ::Int64, ::RoundingMode{:NearestTiesUp})
...

julia> rem(5//1, 2//1, RoundNearestTiesUp)
ERROR: MethodError: no method matching rem(::Rational{Int64}, ::Rational{Int64}, ::RoundingMode{:NearestTiesUp})
...

[anshul23102]

I’d like to take this up 👍

The lack of rem(x, y, r) implementations for RoundNearestTiesUp and RoundNearestTiesAway across Integer, AbstractFloat, and Rational types looks like a clear gap in rounding-mode coverage. I plan to:

• Investigate the expected mathematical semantics of rem under these rounding modes (especially tie-breaking behavior).
• Check consistency with existing div / fld / cld rounding implementations.
• Add appropriate method definitions (or shared internal logic) for these numeric types.
• Include tests covering tie cases to ensure correctness and alignment with Julia’s rounding rules.

I’ll update this thread once I have a concrete proposal or WIP PR.

[lvlte]

Hi @anshul23102,

I'm already on it ! I have created the issue just to have a reference for a PR that is coming. Sorry I should have mentioned it in the description.

Currently I think I have addressed the issue for the Integer and AbstractFloat types, I'm currently running some tests.

So what you can do is to focus on the implementation for Rational ?

[anshul23102]

Thanks for the heads-up, and no worries at all 🙂

Great to hear you’ve already covered Integer and AbstractFloat. I’m happy to focus on the Rational implementation then.

I’ll look into:

• Defining rem(x::Rational, y::Rational, r::RoundingMode) semantics consistent with the div/rem behavior and tie-breaking rules.
• Reusing existing logic where possible (e.g. via div(x, y, r) and x - y*div(...)).
• Adding targeted tests, especially for tie cases, to match RoundNearestTiesUp and RoundNearestTiesAway.

[nsajko]

xref issue #29939