Regex problem

avataru

Hello, I am having trouble building a regex to match the last word from a string.

Normally it would be /\s+?(\S+)?$/ so it matches anything from the last space to the end of the string. The catch is that when I have html or bb tags it would break the code, so I need it to catch those too. Note: BB code is delimited by [ ].

There are several cases:
foo bar -> " bar"

foo <tag attrib="bar">foo</tag> -> " <tag attrib="bar">foo</tag>"

foo <tag attrib="bar">fo</tag>o -> " <tag attrib="bar">fo</tag>o"

foo f<tag attrib="bar">oo</tag> -> " f<tag attrib="bar">oo</tag>"

foo f<tag attrib="bar">o</tag> -> " f<tag attrib="bar">o</tag>o"

foo <tag attrib="bar">foo b</tag>ar foo-> " foo"

So far I made /(<[a-zA-Z]+?)??\s+?(\S+)?$/ but it won't catch the characters before the "<". This won't break the code but will break words.

My next step was /\s+?(\S<.+? >)?(\S+)?$/ but this is bad (slow) for long strings and in some cases, results in catastrophic backtracking.

Thanks in advance.

xblue

My first simple idea would be to get rid of the tags before matching. However if you need the tags to be included in the match this does not make any sense.

Second idea - not at all sure how efficient this is compared with your approach, but as it avoids ungreedy matching, maybe it helps to increase performance:
(<[^>]+>|\S)+$

avataru

If you are familiar with Smarty, this is actualy part of a modified version of the official truncate modifier plugin.

Anyway, if I don't match the tags <> or [], the string will break in bad places since it matches from the last space, to the end of the string in the original version. That means that if I have for example "foo b<a href="#">ar</a>" it would match "href="#">ar</a>" so the string will end up being "foo b<a" which is a bad thing in my case. The correct result in that case should be "foo" since the "b" is part of the next word, even if it has tags (by word I mean any substring of characters without a space between them).

I hope it's a bit clearer now.

xblue

Well, it explains things a bit better than the original posts (though I do not work with smarty myself).

What I still do not completely understand: Applied to your example
"foo b<a href="#">ar</a>"
I think the pattern I posted should exactly match
"b<a href="#">ar</a>"
and the slightly modified
\s+(<[^>]+>|\S)+$
should match
" b<a href="#">ar</a>",
thus leaving you with "foo" as desired when using it with preg_replace.

Isn't that what you are after? Or doesn't it work? Or is it not efficient enough?

avataru

Your last idea is really close, thank you. It still has a problem but I should be able to solve it.

A string:

L<span style="background-color:#FFC73D">o</span>rem ipsum d<span style="background-color:#FFC73D">o</span>l<span style="background-color:#FFC73D">o</span>r sit amet, c<span style="background-color:#FFC73D">o</span>k

Matches fine but it won't work if the last tag is unfinished, like this:

L<span style="background-color:#FFC73D">o</span>rem ipsum d<span style="background-color:#FFC73D">o</span>l<span style="background-color:#FFC73D">o</span>r sit amet, c<span style="background-color:#FFC73D"

This will match

 style="background-color:#FFC73D"

Also, I'll have to make it match [] as well as <> tags but this should be easy too, I think, using the |.

Anyway, thanks for your help. When I'll get it to work I'll post here as it might be useful to someone else in the future 🙂

Aurasia

I did something like this quite recently ... I tried for a day to get a work around together ... and what I settled with was this.

Basically, if from the end of the string there's a '<' ... but no '>' then you need to delete the entire last part of the string. It's all inside of a tag (this assumes that all the < will have >'s in your string segments.

Then simply take the last word when that's done. I don't have the code with me since it was part of another project, but it can't be difficult to build in one of those RegExpression editors ... can it?

Aurasia

function finalBracket($len,$txt){ 
	if(strlen($txt)>$len) { 
		$check="1"; 
		while($check!=">") { 
			$check=substr($txt,$len-1,1); 
			$len--; 
		} 
	$txt=substr($txt,0,$len); 
	}

return $txt; 
}

function nearestWord($len,$txt){ 
	if(strlen($txt)>$len) { 
		$check="1"; 
		while($check!=" ") { 
			$check=substr($txt,$len-1,1); 
			$len--; 
		} 
	$txt=substr($txt,0,$len); 
	}

return $txt; 
}

I'm sure there's a more elegant solution. But the function above will find the last angled bracket. The second will find the last word. If you use the location of the '<' and the final '>' you could figure out whether the last tag was closed or not. If it's not close, take the entire last part of the string including the '<' and delete it. Then run the nearestWord function. 🙂 I hope that helps!

xblue

I'd probably add that unclosed tag the end the pattern as optional part.
\s+(<[^{>]+>|\S)+(<[^>]+)?$}

To deal with the angular brackets, I thought you could use backreferences first, but I doubt that you can use them in negated character classes, so probably with alternation rather (starts looking ugly 🙂 and may contain typos).
\s+(<[^{>]+>|[[^{]]]|\S)+(<[^{>]+|[[^]]+)?$}}}