Wrong positions with GLRParser and EMPTY

[Hyldrean]

• parglare version: 0.12.1.dev13+gc0df483
• Python version: 3.5.3
• Operating System: Debian GNU/Linux 9 (stretch) 64 bits

Description

The start/end positions are sometimes wrong when we use the GLRParser and grammar with EMPTY. NB: I couldn't reproduce the bug with the LR Parser.

What I Did

Here is a minimal working example

from parglare import GLRParser, Grammar

rules = """
S: A A B;
A: letter;
B: letter | EMPTY;

terminals
letter: /\w/;
"""
g = Grammar.from_string(rules)
parser = GLRParser(g, build_tree=True, debug=False, debug_colors=True)
result = parser.parse("ab")
print(result[0].tree_str())

It gives the tree

S[0->1]
A[0->1]
  letter[0->1, "a"]
A[1->2]
  letter[1->2, "b"]
B[1->1]

One can easily see that it's rather

S[0->2]
A[0->1]
  letter[0->1, "a"]
A[1->2]
  letter[1->2, "b"]
B[2->2]

[igordejanovic]

Thanks for the report. Could you make a PR with failing regression test?

[Hyldrean]

It is the very first time I use py.test and I do a pull request. I hope all is fine.

[igordejanovic]

@Hyldrean It's perfect. Thanks!

[igordejanovic]

Fixed on master

[Hyldrean]

Thanks ! It works for me ;)