make production ending semicolon optional about bnf

[shnewto]

Yeah I've been thinking about this one too. Thanks for raising it.

[shnewto]

One note, I'd like it if we didn't require a newline either so breaking up long productions or expressions across lines wouldn't prohibit parsing.

[CrockAgile]

Not requiring the newline seems trickier to me, but sounds like you have an idea. How would that work?

[shnewto]

@CrockAgile I've got a lot of faith in what we can get out of nom, I could be misguided but here's my general thought. Productions can end on a ; on eof or if with peek or some similar custom mechanism we are able to see that a <...> ::= is next in line. What do you think?

[CrockAgile]

@Snewt that seems reasonable to me. I still need to brush up on nom to know how it handles forward peeking, but you are probably right it is doable. I would guess it would improve usability with negligible impact on performance or code readability

[CrockAgile]

@Snewt Would it be too slick/hacky to sidestep this issue by honoring BNF comments? If BNF comments are from ; to the end of line, then essentially they become optional. And would support comments in the input string.

[shnewto]

@CrockAgile I'm not sure that would really sidestep the issue as it stands. It could be that we add support for ; comments but as it is they are expected to only show up in specific instances.

[CrockAgile]

@Snewt Sounds good! I think I got confused anyway by glancing at the algol example in the history section on the wiki.

[shnewto]

closed by 21338d3