parsing tree-like non-XML file

[marenz]

I'm strugling to parse this exotic, nested file that isn't XML.

here's a sample

F(
T(22)
p(
	X(401) Y(3) W(15) F(0)
	L(2)
s(
	T(n() C(1) t(1))
	Folders()
	c(

I built a character-by-character scanner that works well enough:

	$limit=strlen($file);
	echo "Begin token ingestion.  ($limit)\n";
	for ($i=0;$i<=$limit;$i++){
			$chr=substr($file,$i,1);
		if ($chr=='('){
			$lvl++;
			$numTok++;
			$tokens[$numTok]= new tok($numTok);
			$tokens[$numTok]->id=$numTok;
			$tokensId[$lvl]=$numTok;
			$tokensStart[$lvl]=$i+1;
			$prev=substr($file,0,$i);
			$st1=strrpos($prev,"\n");
			$st2=strrpos($prev,"(");
			$st3=strrpos($prev,")");
			if ($st2>$st1){
				$start=$st2+1;
			} else {
				$start=$st1+1;
			}
			if ($st3>$start){
				$start=$st3+1;
			}
			$len=$i-$start;
			$tmpText=substr($file,$start,$len);
			$tmpText=str_replace(" ","",$tmpText);
			$tmpText=str_replace("\t","",$tmpText);
			$tmpText=str_replace("\n","",$tmpText);
			$tmpLabel[$lvl]=$tmpText;
			if ($tmpLabel[$lvl]=='Group'){
				$numGrps++;
				$grp=$numGrps;
				array_push($grpList,$grp);
			}

	}
	if ($chr==')'){
		$stop=$i-$tokensStart[$lvl];
		$num=$tokensId[$lvl];
		$tokens[$num]->label=$tmpLabel[$lvl];
		$tmpStr=substr($file,$tokensStart[$lvl],$stop);
		$tokens[$num]->val=$tmpStr;
		$tokens[$num]->lvl=$lvl;
		$tokens[$num]->grp=$grp;
		$prev=$lvl-1;
		$tokens[$num]->parent=$tokensId[$prev];	
		if ($tmpLabel[$lvl]=='Group'){
			array_pop($grpList);
			$tokens[$num]->grpSub=$grp;
			$lastGrp=$grpList[(count($grpList)-1)];
			$grp=$lastGrp;
			$tokens[$num]->grp=$lastGrp;
			echo "lastgrp= $lastGrp\n";
		}
		$lvl--;
	}	
}

... but for longer files it really chokes up. Folks tell me I should be able to use classes and regular expressions to type the file into a linked list, but I can't find anything on the web about how to do that.

update: I've tried using preg_match_all to grab items delimited by '(' and ')' but it doesn't do the trick since I end up with a non-nested mess.

[rehfeld]

not sure if its just because of you posting a small snippit example of the text to parse, but are there really not always some type of end tag/token to work with?

i see this "tag" open,

p(

but i dont see a closing ) for it....

i guess you just have to infer a closing token when you find a new opening token?

that looks tough....

regardless i think regex could prob make it a bit easier/ faster if used correctly.

at a glance, since it does kinda look hierarchial / xml like, i think i would try and use preg_split() or preg_match_all() to break things into sub peices and then work on those subpeices.

btw- i doubt you will see any major speed improvement, but instead of using substr() to get a single character, you can do this

$str = 'abc';
echo $str{1}; // b
echo $str{2}; // c

[marenz]

i see this "tag" open,

p(
[/code] [/B]

I should have explained, the problem is that the whole file is essentially one big object with containers, an "E(" tag that stays open the entire time. Thanks for the advice, I think nixing the substr will help. I also split the conversion into two scripts, one to open the original file and spit out a more easily parsed one with just what I need (50% or more of the file will be ignored) and the other to do the actual work of conversion.

EDIT: I also messed about with PHP's preg_match_all and split functions, and they work fairly well for flat instances. However, they fail to recognize nested instances.

For example:
A(
B(
C(1)
)
)

shows up as
array[0] => A(\n B(\nC(1)
array[1]=> )
etc.

depending on how I set up my regexp.

[marenz]

In an effort to speed things up, I nixed the idea of stepping through every character and made a loop that checked for the next positions of '(' or ')' depending on which was closer, skipping past all the other stuff. for reasons I cannot possibly fathom, this method turned out to be far slower. WTF?