Quinton Miller
Quinton Miller
I was thinking about how to bridge the differences between our rounding modes and MPFR's. MPFR defines seven: * `MPFR_RNDN`, equivalent to `RoundingMode::TIES_EVEN` * `MPFR_RNDZ`, equivalent to `RoundingMode::TO_ZERO` * `MPFR_RNDU`,...
You can probably make `to` an external parameter name and `limit` the internal one to get the best of both worlds.
But giving different external and internal parameter names doesn't imply that the docs must use the former. For instance, [`Steppable`](https://crystal-lang.org/api/master/Steppable.html#step%28%2A%2Ctolimit%3Dnil%2Cbystep%2Cexclusive%3ABool%3Dfalse%2C%26%29%3ANil-instance-method) does this: > `def step(*, to limit = nil, by...
It seems odd that `GC.malloc_atomic` crashes here.
There is also [Dragonbox](https://github.com/jk-jeon/dragonbox) now.
Dragonbox implements only the shortest round-trip conversion; while this is good enough for `Float#to_s`, it is unsuitable for `sprintf` `%f` / `%e` / `%g` because double rounding would occur. Instead...
We have other similarly sized lookup tables (Unicode properties) so that is probably not an issue for now. But I will definitely keep an eye out for that Ryu Printf...
Expanding the subtyping relationship manually, we could probably write `T.union_types.all? { |t| t
As I mentioned somewhere else (in the `Int128` literals issue I think), the parser does not have access to the `Crystal::Program` because it doesn't need one, so the warnings have...
I think it should be the opposite: the parser itself should be able to report warnings even without a `Crystal::Program`.