[sean0407]

Fixes #5469

The lexer currently stops before the decimal point in 1.e1 and 1.e-1, producing 1..e1 and 1..e - 1. Recognize a trailing decimal point when it is followed by a valid exponent so these inputs compile as JavaScript numeric literals.

Changes:

• Add the exponent form to the source and generated lexer.
• Add regression coverage for positive and negative exponents.

Verification:

• Generated lib/coffeescript/lexer.js from src/lexer.coffee in an isolated, network-disabled container.
• Ran 5 isolated Node assertions: both expressions evaluate to the expected values, both emitted forms retain their numeric literal, and existing 4.toString() behavior remains intact.
• git diff --check passed; patch policy accepted 3 text files and 8 changed lines.

[Inve1951]

I suspect this was intentionally not supported.
See the decimal regex on the next line: it also doesn't support a trailing dot.

As you pointed out this also changes the meaning of 42.e3 from property access to number literal, which is a breaking change.

[sean0407]

You're right: 42.e3 is currently parsed as access to the e3 property, so treating it as a number would change existing behavior. The same ambiguity applies to a signed exponent such as 1.e-1, which can currently be parsed as property access followed by subtraction.

I agree this should not be merged as a bug fix without a compatibility decision. Would you prefer to leave this syntax unsupported and close #5469, or consider the change only for a future breaking release? I'll hold off on further changes pending your preference.