Re: lookahead ?

On Thu, 26 Apr 2001, Damian Conway wrote:

> You really do want to use a regex here (to get the lookahead/backtracking
> that RecDescent doesn't do).
>
>    line: m/^(\s*        		# leading whitespace
>    	      (.*?)        		# the title
>               (?:\s+(?=\d+\.\d+\s*$)) 	# the space preceeding the numbers
> 	    )
>             (\d+)        		# the major version number
>             \.
>             (\d+)        		# the minor version number
> 	 /x
> 	 <reject: length $1 != 47>
> 	 { @{$return}{title major minor)} = ($2,$3,$4) }

Ok, but do I need to insert a <skip: ''>, so that the leading whitespace
gets captured (and not skipped)?

I'm still wrapping my mind around all this: when you say that RD doesn't
backtrack .. isn't that what trying alternative rules does, effectively
(as long as there hasn't been a <commit>)?  In other words, how about
something like this:

line: title '' <reject: $thiscolumn != 47> version

title: titleword(s)
titleword: ...!version word

version: minor '.' major
minor: digits
major: digits

word: /\S+/
digits: /\d+/

Hmm, I guess that would fail in an instance where a substring that looked
like "version" occured prior to column 47, even if a valid "version" also
occurred at column 47, something like this:

01        11        21        31        41    47
   Here is a title with 4.3 in it             5.6

where the title here should be "Here is a title with 4.3 in it"

Hmm, how about a rule that grabs words until $thiscolumn >= 47 ?

title: titleword(s)
titleword: word <reject: do { $thiscolumn >= 47 }>

And then I start thinking about <leftop> and <rightop>, because it seems
like what I really want is:

title: word '' <reject: $thiscolumn != 47> |
       word word '' <reject: $thiscolumn != 47> |
       word word word '' <reject: $thiscolumn != 47> |
       ... etc

since I could easily insert '' between each word there, perhaps I could do
this:

title: <leftop: word '' word> '' <reject: $thiscolumn != 47>

I guess that's still a "greedy" match that would pass column 47 to include
the version, and then be rejected in it's entirety.

So how about:

title: <leftop: word '' <reject: $thiscolumn == 47> word>

Sorry for the perhaps random musings, and thank you for your assistance in
this most academic matter.

-Aaron