Regular Expressions - Expert needed!

LonelyPixel

Hi folks,

I'm trying to build some BBCode parser in PHP for a whole while now. Now I have two simple ways of how to do this: (Spaces in Tags are against this board...)

Replace any occurence of [ b ]...[ /b ] by ...
It means you can enter simple text formats but they may not be nested at all or it becomes ugly.
Replace any [ b ] by and [ /b ] by separately
Now I have no problems with nested Tags - only when I need to get the contents of my tag for further manipulation (but that's rarely nested). But it won't check for opening+closing tags and can only be translated 1:1 into HTML.

I now thought of something like
[ b ] ... [ b ] ... [ /b ] ... [ /b ]
to get nested tags, too. But if I have two serial occurences of my tag on the 2nd layer i.e., it's bad again.

Here's what I thought of by now:

<style>
b.r {color: darkred;}
b b.r {color: red;}  /* to highlight the nested "boldbold" */
b.b {color: darkblue;}
b b.b {color: blue;}  /* to highlight the non-nested "boldbold" */
</style>
<?
# Here is the source data: multiple non-nested tags and (multiple) nested tags
$source = "begin [ b ]bold[ /b ] normal [ b ]bold2[ /b ] normal2".
" [ b ]bold3[ /b ] normal3 [ b ]bold4[ /b ] end";
#$source = "begin [ b ]bold [ b ]bold2[ /b ] bold3[ /b ] end";
#$source = "begin [ b ]bold1 [ b ]boldbold1[ /b ] bold1[ /b ] ".
#"middle [ b ]bold2 [ b ]boldbold2[ /b ] bold2[ /b ] end";

# This is the translation of nested tags. It comes first
$pass1 = preg_replace(
	"/
		\[ b \]          # 1st opening tag
		(.*?)            # Data 1st layer (part 1) --> In this part, 
#there must not be any [ \/b ]
		                 #     But this part must stay limited 
#by the following [ b ]
		                 #     So I had to check the contents of this 
#parenthesis afterwards anyhow.
		\[ b \]            # 2nd opening tag
		(.*?)              # Data 2nd layer
		\[ \/b \]          # 2nd closing tag
		(.*?)            # Data 1st layer (part 2)
		\[ \/b \]        # 1st closing tag
#		((?(?!
#			.*?
#			\[ \/b \]
#			.*?
#			\[ b \]
#		).*?))
	/isx",
	"<b class=r>$1<b class=r>$2</b>$3</b>", $source);

# The commented lines in the RegExp are the try of a belated check of \1...
# But I can't just compare \1 with anything but the current data from $source
# It would be nice to have something like
# (if \1 matches [ /b ] then fail) inside my RegExp :-)

# This is the translation of non-nested tags. It comes next
$pass2 = preg_replace("/\[ b \](.*?)\[ \/b \]/is", "<b class=b>$1</b>", $pass1);

echo "<tt>$source<br>$pass1<br>$pass2</tt>";
?>

(Sorry, you'll have to remove the spaces here too since the Board translates them even INSIDE A PHP block!!)

If you're interested in my first "release" of the BBCode translator, check out http://software.unclassified.de/abbc. (I'm going to translate this into English as it's getting international now :-)

I'd really be interested in any suggestions on how to get this stuff done with open/close-tag check, with correct nested-tag handling and, if possible, with regular expressions. Otherwise I'll have to go down by scanning every single character of the user posting and check for tags for my own...

TIA

Weedpacket

The crucial point that has to be made about regular expressions is that they can't count; so a single regexp can't cope with tags that are nested arbitrarily deep. If there is a fixed limit, it can be done, but really only in principle; a solution that would also work for arbitrary nesting would be easier.

So, what sort of solution? I'm just going to sketch something out here, rather than write all the code - I'd be bound to make some silly mistake somewhere, 'cos right now my brain isn't parsing PHP to well. It uses an idea I've posted a couple of times already, in threads 10219330 and 10221160.

You find each occurrence of {b}...{/b} in the string (using {} instead of [] that does't contain any intervening {b} or {/b} tags.

Capture the bit in between, and replace it with some identifier. Store the captured bit along with the identifier in an array.

Keep doing this until there are no {b}...{/b} pairs found (this is the loop that regular expressions can't do) - in other words, the string doesn't change after the preg_replace(), so it means hanging on to the previous value of the string for a "before-and-after" comparison. The only {b} and {/b} tags left are unmatched ones.

Now go through the array you've built up. Do this in reverse order, 'cos some of the later captured bits might contain the identifiers from earlier replacements (cos of nesting). Replace identifiers in the string with corresponding captured bit.

By adding a third field to the array - "tagtype" to go alongside "captured text" and "identifier", you could proceed to match {i}...{/i} and other pairs of tags at the same time as you match and replace {b}...{/b} pairs.

csaba

You may wish to review the excellent Regular Expression primer at http://www.php.net/manual/en/pcre.pattern.syntax.php
It covers recursion within regular expressions and may be useful for your situation.

Csaba

LonelyPixel

Thanks for you help so far,
but I haven't seen a chance to get it done this way with reasonable effort. So I've decided to do the nesting scan with a loop through the whole content and only let the RegExps translate one level of opening/closing tags. That was a bit computing and logical work but it looks quite good by now.
I even have a sort of syntax highlighting now, that marks tags that are not closed, closed but never opened or nested wrongly!
At the moment I'm bothering CODE tags whose contents must not be processed and general masking of special characters. Guess I'll release it in a few days on the mentioned website 🙂