Clarify the spec by making "empty" more explicit in the examples

[ndw]

This is an editorial issue, not a substantive one. In the "Hints for Implementors" section, and perhaps elsewhere but this is where I noticed it, there are some rewritings that are optional (in the sense that one of the alternatives is empty). For example: the right hand side of the f-star rule:

f* ⇒ f-star
-f-star: f, f-star; .

I find it very easy to miss the significance of ; . at the end of that rule. I think it would be better if we made it less, uh, "invisible", if you'll forgive the pun. I think three alternatives have floated past in email:

A comment:

f* ⇒ f-star
-f-star: f, f-star; {empty} .

The string ():

f* ⇒ f-star
-f-star: f, f-star; () .

Or an explicit nonterminal:

f* ⇒ f-star
-f-star: f, f-star; empty .
-empty: .

I don't have a strong favorite among those, but I probably lean marginally towards () for its simplicity.

Any of those would draw my attention to a detail that I've read blithely past more than once.

For completeness, I believe []? was also discussed in email, and I think it would do, but it seems just a little more confusing to explain.

[spemberton]

This is an editorial issue, not a substantive one. In the "Hints for Implementors" section, and perhaps elsewhere but this is where I noticed it, there are some rewritings that are optional (in the sense that one of the alternatives is empty). For example: the right hand side of the f-star rule: f* ⇒ f-star -f-star: f, f-star; . I find it very easy to miss the significance of ; . at the end of that rule. I think it would be better if we made it less, uh, "invisible", if you'll forgive the pun. I think three alternatives have floated past in email: In this particular case (but not the one before it in the spec), we could replace it with f* ⇒ f-star -f-star: (f, f-star)?. If we did this for the last example as well, then this section would have the nice property that each rewriting rule would be based on earlier ones. Steven

[ndw]

That sounds fine to me. It still leaves the question of what to do about "?" itself, but it's more obvious what's going on there because the name "option" is a clue, at least for English speakers, and the rule is so simple, there's less to distract one from seeing the significance of ; ..

[spemberton]

In the one place where empty occurred, it was replaced by (). Closed.