Surpressing Mulitple Matches

[Jonathan Stott]

Hi all

I'm writing a parser for some irc logs (to convert them to wiki markup and/or html) and the machine is fairly simple. 

eol = '\n';
text = any -- eol;

# action line (generated from /me)

action_line = timestamp action_nick text %print_action eol;


# text line, normal irc chat

text_line = timestamp nick text %print_text eol;


# system line. Joins, parts etc.  Included so I can ignore it.
# system_leader is a unique bit of text the irc client puts just after the timestamp for all system info. 

system_line = timestamp system_leader nick text eol;


# hr line. A convention which has developed. 
#A line of just '-'s should be converted to a <hr />

hr_line = timestamp nick '-'{3,} %print_hr eol;

line = action_line | text_line | system_line | hr_line

main := line*

print_action, print_text both print a nicely formatted line, with two newlines, suitable for the wiki. print_hr prints '----', the wiki markup for a <hr />.  Except that any hr_line, is also a valid text line, so two lines get printed.  And I can't really exclude '-' from a text_line, they do get used in typing.  Is there any solution to avoiding printing both lines? (Beyond the obvious one of some test like 'if this is all -s, don't print' in the print_text action)

Regards
Jon