`ToBigInt` and `ToBigUint` should not return an error about num HOT 5 CLOSED

Can a core team member comment on this? It's an easy change to make, and it can have huge benefits. I was thinking of changing all the std::ops traits to take a type that implements ToBigInt, so that you can do useful things like adding regular integer constants to BigInts, which would greatly reduce the verbosity of the current API. However, I can't do this because these conversion methods can fail.

from num.

@huonw, @alexcrichton, do you have thoughts on this? I'd like to implement it.

from num.

@hauleth and @cuviper, you are the new maintainers, correct? What do you think about this issue?

from num.

I'd rather use From for infallible conversions (#117) and leave these alone. You already pointed out that ToBigUint could fail on signed values, and even ToBigInt could fail if it were implemented for something like BigRational. I think it's useful to allow this possibility, and we don't need to break the current API when From and Into can do the job you want.

Expanding generic std::ops for From/Into types sounds fine, but do note the issues that came up in #125 doing something like this for Complex. You can be generic on the RHS, but the compiler won't let you implement for T on the LHS.

from num.

FYI, num-bigint has its own repo now: https://github.com/rust-num/num-bigint

But I think my prior comment still stands, and we do have From primitive integers now.

from num.