Fixes #2981

[xixixao]

To fix #2981, an example of going the way of checking for EOF, which probably should be done, to avoid problems such as #2984 (this PR includes that fix).

One thing to watch out for is that someone might be relying on generated tokens and this obviously adds the EOF token to the list, changing the compiler's output.

(parser compiled with Jison 0.4.4)

[vendethiel]

I'd really love if we could remove parts of this file, instead of adding more...

@michaelficarra @jashkenas Are you fine with EOF being last of each Root items? sigh

[xixixao]

I was pleasantly surprised by the conciseness of this grammar, given the freedom of the syntax the language provides. But of course, keeping grammar to minimum is always favorable.

[michaelficarra]

I agree with @Nami-Doc that it's not exactly a good solution. Can we not come up with anything better?

[vendethiel]

This is what a more concise grammar would look like. Even doing a lot more things ;). Kudos to satyr, as usual.

[jashkenas]

@satyr's use of ditto in that file is really silver-tongued.

[michaelficarra]

Ugh, the stateful implementation is awful. I would prefer something more declarative. Maybe a block where rules within it have the given action.

ruleBlock = (action, block) -> block action

ruleBlock ((a, b, c) -> some action involving a, b and c), (action) ->
  o 'other rules', action

[vendethiel]

Congratulations, now you can't read a single thing anymore :/. I don't think that'd improve anything.
Your function is just a do

[michaelficarra]

Yeah, I was trying to implement a let, and I forgot we have do which is the almost equivalent.

do (sharedAction = (a, b) -> some action involving a and b) ->
  o 'other rules', sharedAction
  o 'more rules', sharedAction

edit: The point is being explicit rather than implicit. It might hurt the eyes, but it still improves readability.

[vendethiel]

And we need concats because there are arrays, and not every single combination in the block share the same action

[vendethiel]

@jashkenas what's your stance on this ?

[jashkenas]

On this PR? Iffy. Wouldn't it be simpler to strip the EOF before parsing, instead of forcing it?

On a stateful implementation of ditto for the parser? Sounds fine. The semantics of the word "ditto" are "stateful" in English as well.

[vendethiel]

strip the EOF before parsing,

I don't understand what you mean

[xixixao]

The EOF is added before parsing, to make sure the parser goes through all the tokens. Is there a simpler way to enforce that?

[vendethiel]

Alternative got merged long time ago.