identify number literals

[getify]

it would appear number literals are complex enough to need heuristic lexing (can't just be done with regexes).

[getify]

Examples of complications:

• {x:4}-4 this is either a { .. } statement block followed by a -4 number literal expression statement, or it's an object value followed by a - subtraction operator followed by 4 number literal.
• 42.e3 is valid just like 42e3 is.

[getify]

Eh, this might not need to happen after all. -42 or +42 is what makes it contextually sensitive, but JS doesn't see a -42, it always sees a - operator followed by a 42, which means JS doesn't treat it context-sensitive. We probably shouldn't either, if for no other reason than to avoid unnecessary work. All those devs (like me) who want -42 to be treated as a literal are just out of luck.

Outstanding question: does that also go for -Infinity and -0? @mathiasbynens

[mathiasbynens]

Outstanding question: does that also go for -Infinity and -0?

Yes. Esprima confirms this: http://esprima.org/demo/parse.html?code=%5B%0A%20%20-Infinity%2C%0A%20%20-0%0A%5D

[ljharb]

At the very least, I think it would be a useful start to identify number literals even if the unary + and - weren't included - especially with the "e" syntax. I know that I'd love (in my own projects) to enforce that numbers ending in zeroes use scientific notation, for example.

[getify]

Interesting. Since the regex for identifying all the variations of number formats is non-trivial, it might make sense to have in this lib, although this is definitely not a full lexer, it's really only for the "tough parts" of lexing. Number lexing seems like it almost meets that standard. I'll think about it.