Re: LPEG and indentation based syntax

["Francisco Sant'anna" <francisco.santanna@...>]

What about this use for nested html tags?

tag = {
'<'* Bs(1-'>') *'>'            -- matches any <tag1>
* (1                           -- followed by any char
   - (#('<'*(1-'/')) * 
m.V(1)) -- unless opening another <tag2>
   - ('</'*Br(1)*'>'   )       --   or closing it with </tag1>
   ) ^0
}

On 4/16/07, Roberto Ierusalimschy <roberto@inf.puc-rio.br> wrote:

Conventional backreferences (indexed from the beginning of the pattern)
is not very meaningful in LPeg, but I guess that a backreference indexed
from the beginning of the rule could solve your problem.

Assume that lpeg.Bs(patt) stores a backreference to the substring that
matched 'patt', and that lpeg.Br(n) matches the 'n'-th stored substring
in the current rule. Then I think the next pattern would do your trick:

-- Not tested (as lpeg does not support Bs-Br...)

S = lpeg.P" "
line =  -S * (1 - lpeg.P"\n")^0 * "\n"

block = { -- defines an idented block
    lpeg.Bs(S^0) * line   -- capture identation of first line
  * ( lpeg.Br(1) * line   -- block may contain lines with the same identation
    + #(lpeg.Br(1) * S^1) * lpeg.V(1)  -- or idented blocks
    )^0
}

I am not much in favor of adding backreferences to LPeg. But, if that
code is correct, it would be the second example of a "useful" application
for them (the first being long strings in Lua).

-- Roberto