Generating a pretty AST with LPeg

[Matthias Kluwe <mkluwe@...>]

Hi!

I've nearly finished my first parsing task using LPeg (which was parsing
SAS-code to visualize the "data flow") and it was fun.

As I'm lacking some experience my "AST" resulting from the parsing step
is not too pretty. Let me give you a simplified example of what my "AST"
looks now. Consider parsing "function calls" in some hypothetical
language:

Using

    local WS    = lpeg.S' \n\t'^0
    local NAME  = lpeg.C( lpeg.R'az'^1 )
    local PLIST = WS * ( NAME * ( WS * ',' * WS * NAME )^0 )^-1
    local FCALL = lpeg.Cc'FCALL'
                  * lpeg.Ct( WS * NAME * '(' * PLIST * WS * ')' )
    local ALL   = lpeg.Ct( FCALL^0 )

with

    lpeg.match( ALL, [[
    foo( bar, baz )
    asdf( qwer )]] )

results in the table

    {
      'FCALL', { 'foo', 'bar', 'baz' },
      'FCALL', { 'asdf', 'qwer' }
    }

which can be interpreted as the list of expression types ('FCALL'),
followed by a data structure containing the function name and the
arguments.

For better readability and "type safety" I'd rather like it to be

    {
      'FCALL', { name = 'foo', args = { 'bar', 'baz' } },
      'FCALL', { name = 'asdf', args = { 'qwer' } }
    }

Currently, I'm achieving this by a second transformation step.

I'm pretty sure this can be directly done by some clever trick in the
LPeg expressions above.

Please drop me a hint on how to do it.

Any other suggestions are welcome, even a "This is not the way to do it,
usually you do [...]".

Regards,
Matthias