"Lookahead exceeds token buffer length" with simple grammar

[Leandros]

Trying to parse anything with a parser generated from this elementary grammar:

%start Foo
%title "Foo grammar"
%comment "Empty grammar generated by `parol`"
%line_comment "//"

%%

Rjs: 'Hello' 'world';

Trying to parse

Hello world

results in the following errors:

  × Failed parsing file ./test.txt
  ├─▶ Failed accessing lookahead token!
  ╰─▶ Lookahead exceeds token buffer length

The error can be fixed by adding the empty case to the production:

Rjs: 'Hello' 'world' | ;

Is this intended behavior?

[jsinger67]

@Leandros Thank you for reporting this.
I know about this behavior but it is not good documented and actually easy to fix.
The cause of this is that your first grammar is actually k0. This results in a lookahead buffer of size 0.
I think I will fix this simply by setting the lower size limit of the lookahead buffer to 1.

Edit: I think I should elaborate a bit on this weird k0 thingy.
From the perspective of the parser there is no need to choose between different productions for a non-terminal when there's only one production. So it doesn't need no lookahead at all to solve this decision problem. This is the reason for taking a zero value as lookahead size (k0) for this simple grammar.
In the real world the scanner must be able to deliver consumable tokens for the parser. So buffer size 0 won't work here.
That's the point here. An edge case but valid.

[jsinger67]

Fixed with commit c302ba2.

[Leandros]

Wonderful. Thank you!