Support for big integers

[LorisFriedel]

Hello!

Would it be possible to improve the current integer precision? I understand that all numbers in Jsonnet are represented as double (as defined in IEEE754), but for very big integers (larger than 2^53 IIRC) this gives a pretty poor precision.

If sticking to the language specifications is the priority, maybe it could be considered to add "big numbers" features in the standard library? Or something else?

WDYT?

[CertainLach]

There is an exp-bigint experimental feature available on the master branch.
I am not sure how it should be exposed to the end user; I think about not implementing it in Jrsonnet itself but allowing users of Jrsonnet as a library to implement such type itself as Val::Custom, so don't expect it to be there forever, it is made only for experimentation. Or should there be a proposal to upstream jsonnet to support this feature?

For now, the syntax is

// https://github.com/CertainLach/jrsonnet/issues/103
local bigintEq(a, b) = a >= b && a <= b;

std.type(std.bigint(123)) == 'bigint' &&
bigintEq(std.bigint(20) * std.bigint(30) , std.bigint(600))

For now, it only works with the basic operators and not with std.pow/other standard library functions.