Int64 doesn't handle lower bound properly

[cakey]

d := new(decimal.Big).SetRat(new(big.Rat).SetInt64((math.MinInt64)))
fmt.Println(d.Int64())

d2 := new(decimal.Big).SetRat(new(big.Rat).SetInt64((math.MinInt64 + 300)))
fmt.Println(d2.Int64())

d3 := new(decimal.Big).SetRat(new(big.Rat).SetInt64((math.MinInt64 + 350)))
fmt.Println(d3.Int64())

> 0 false
> 0 false
> -9223372036854775000 true

Unless I'm misunderstanding, all 3 should return true?

It looks like it's triggering the u > math.MaxInt64 branch.

I imagine it might be something related to .compact being uint64, but here we care about int64, but I haven't dived in too deep.

I realised it might have something to do with the set precision - if i set precision high enough, then math.MinInt64+1 works correctly, but never math.MinInt64.

Is this intended behaviour?

[ericlagergren]

You're correct about precision. Your call to SetRat rounds down based on z's Context. (Essentially, SetRat is computed as z = a / b where a and b are the numerator and denominator of x, respectively.)

You're also correct that branch doesn't take into consideration that the negative half of a 64-bit integer is +1 larger than the positive half.

[ericlagergren]

Okay. That's fixed. I'm gonna push up a 3.1 which has a handful of bug fixes once Travis CI green lights me. Thanks for reporting them. Your issue made me curious: do you need something like Context.SetRat? I've omitted Context versions of most of the SetX methods, but I'm not opposed to adding them.