BigInt: Stage 2

[littledan]

New Issue: babel/proposals#2

---

Champion: @littledan
Spec Repo: https://github.com/tc39/proposal-bigint
Spec Text: https://tc39.github.io/proposal-bigint
Slides (to Stage 3 at May TC39 Meeting): https://docs.google.com/presentation/d/1lrcjQzIFgdUXczeeAzs4GkTXJsRQU21UhtmXef70Qic/edit#slide=id.p

Parser

• Similar to Numeric Separator: Stage 1 #538, we'll need to modify the tokenizer to accept 10n
  as a valid number. (Check that PR)
• From a syntactic perspective, the only thing added is numeric literals of the form 987235987032n. These permit hexadecimal, octal and binary prefixes, but no exponential form, and no decimal place. See the grammar in this section of the proposed spec.

Transform

• All arithmetic operations and comparisons should be replaced by calls to functions which provide the appropriate behavior depending whether arguments are BigInts or Numbers.
  • The transform is slow, so it will really be more of a toy rather than something that should be used in production. For this reason, I'd argue it should be left out of Babel presets.
• Polyfill needed--ideally, this would be a thin wrapper around another BigNum library from npm. There are just a few methods to support. Some functions will have to be a bit different to get close to the right behavior, for example, valueOf() should throw a TypeError, so that implicitly casting to Number throws a TypeError, even if it would really return the BigInt. Additionally, the polyfill needs BigInt64Array and DataView operations.

[wdhorton]

I'd like to take on this issue!

[amilajack]

All arithmetic operations and comparisons should be replaced by calls to functions which provide the appropriate behavior depending whether arguments are BigInts or Numbers.

I've done some work on this in a babel plugin that prevents coercion in JS. Here's the related code for that

[littledan]

@wdhorton Let me know if you want any help with reviews, interpreting the spec, etc!