Identifiers must have an alpha-character in them.

[asoffer]

We use _ as a digit separator in numbers, and allow them to be placed anywhere, so 0_ would be considered a valid number, but _0 would be an identifier. This seems to be needlessly subtle.

We should change the rules so that an identifier must

• Have only alphanumeric and underscore characters
• Not start with a numeric character
• Have at least one alpha character.

These rules seem relatively straightforward and have the added benefit of making _ not a valid identifier (reserving it for future use like indicating ignored values or pattern matching).

[perimosocordiae]

Why not ban trailing underscores in numeric literals? Effectively this means underscores may only appear internally in numbers, as implied by the name "separator".

Then state: numbers must start with a digit, and identifiers must not start with a digit. Seems pretty clear to me.

[asoffer]

This still means that 12_3 and 1_23 are numbers but _123 is an identifier. I might be okay banning this, but I think it's best to demand that a sequence of characters containing only digits and underscores is not an identifier. It might be a number, and I'm okay banning some outright, but it shouldn't be an identifier.

[perimosocordiae]

Python supports _1 as an identifier, but I'm fine banning it.

[asoffer]

Interesting. That's pretty compelling. I didn't realize Python allowed underscore as a digit separator. With that in mind, I'm just going to close this and stick with what we have. The fact that Python represents prior art makes me think this is probably fine. The only differences I can discern is that Python doesn't allow trailing underscores or multiple underscores. But neither of those feel like the concerning bits here.