Grammar sends Earley implementation to an infinite loop

[erezsh]

Earley implementation never terminates when running with the input "a", and the following grammar:

start: a
a: a | "a"

Of course the grammar is badly formed, but since we aspire to parse "any grammar", an error is a more appropriate behavior.

[lucaswiman]

@erezsh If I understand correctly, this grammar is not malformed, but also enters into an infinite loop:

from lark import Lark

l = Lark('''
  start: bar+
  bar: /a|b|c*/ "foo"
''')

l.parse('afoobfooccfoo')

[lucaswiman]

It seems the issue is that the regex can match zero characters. This one works fine:

Lark('''
  start: bar+
  bar: /a|b|c+/ "foo"
''')

and this one fails in the same way as the example above:

Lark('''
  start: bar+
  bar: /a|b|/ "foo"
''')

[erezsh]

Thanks, good catch!

I think the correct response is to forbid empty regexps.

[erezsh]

Both issues resolved.

[erezsh]

Fixed!