Re: [ANN] Lua 5.3.0 (work1) now available

[Owen Shepherd <owen.shepherd@...>]

I completely agree with making it well defined, but wonder if rounding towards -inf is the most expected behavior (or perhaps most consistent)?

My personal inclination would be round towards zero, because it would be the most consistent behavior; that is, for any integral value i, tointeger(i / n) == i // n

Roberto Ierusalimschy wrote:
> Because we expect conversions from floats to ints to be infrequent
> in Lua with integers, we took the "luxury" to make them well behaved
> operations. (Always rounds towards minus infinte, raise an error if
> number is out of range. We also removed several system-dependent tricks
> that Lua used to speed up those conversions.)