Re: Elegant design for creating error messages in LPEG parser

[Sean Conner <sean@...>]

It was thus said that the Great joy mondal once stated:
> Just a side question.
> 
> Under what design consideration do we use 'Cmt(patt,f) ' vs 'patt/f' ?
> 
> I noticed you used Cmt , why not 'patt/f' ?

  The docs say this of lpeg.Cmt():

	Creates a match-time capture. Unlike all other captures, this one is
	evaluated immediately when a match occurs (***even if it is part of
	a larger pattern that fails later***). It forces the immediate
	evaluation of all its nested captures and then calls function.

	(emphasis added)

  If I didn't, the error checking wouldn't work as expected.  Changing the
WS production to:

	local WS = lpeg.P" "
	         + lpeg.P"\t"
	         + lpeg.P"\n" * lpeg.Carg(1)
	         / function(state)
	             print(">>>",state.line)
	             state.line = state.line + 1
	           end

causes my code to produce the following output as an error:

	Missing ']' at line 1 (start line 1)

even though I see the diagnostic print() statements printed three times:

	>>>     1
	>>>     2
	>>>     3

  I think it has something to do with lazy evaluation in LPEG in this case.

  I've also used it to fail a pattern that would otherwise match.  It is
possible to come up with a pattern that matches the numbers between 0 and
255 but it's quite involved and part of it looks like:

	dec_octet = P"25" * R"05"
	          + P"2" * P"04" * P"09"
	          + P"1" * ...

I found it easier to use Cmt() instead:

	local dec_octet = Cmt(DIGIT^1,function(_,position,capture)
	  local n = tonumber(capture)
	  if n < 256 then
	    return position
	  end
	end)

When the string of digits is greater than 255, this returns nil, which
causes the match to fail.  Doing this:

	dec_octet = DIGIT^1 
                  / function(c)
                      local n = tonumber(c)
                      if n < 256 then
                        return c
                      end
                    end

won't cause the match to fail---instead it will return nil as the captured
data.

  -spc