regex issue ... context selection based on word

utahnix

Hey there,

I'm building a site search feature for our website, and I'm having a hard time. The results come from a MySQL fulltext search, which works fine. But I want PHP to select an excerpt from the body of the article so that the user knows in what context the search was made.

So, if a word is found in the body, PHP should select X words to the left and right of the word. It should account for punctuation on either side of the matched word (things like commas, exclamations, colons, semi-colons, periods and so on).

It should also be able to handle beginning and end (maybe the search word was the first or last word in the article and doesn't have X words to the left and right of it).

I've been working on a php regex pattern to do this, but it doesn't exactly do what I want. It works mostly, but misses the target in one situation.

Suppose the search word was 'love'. If the sentence read "... and he has a deep love and appreciation for the job..." it should match the word "love" (plus X words to the left and right) because love is its own word.

Another example: suppose the sentence said "... the job was true love. He knew he was at the right...". This should match the word love, even though it ends with a period.

But if the sentence read "... and he loves the job and appreciates ... " then it shouldn't match. But it does with my code.

All in all, I just want the behavior to match that of the MySQL FULLTEXT search (using MySQL 5.0.26 on SUSE Linux). I want the search excerpts to match that of what the database found.

Suggestions?

NogDog

Change the "20" in the two "{0,20}" expressions to the max number of characters you want on either side of the search word:

$regex = '/\b.{0,20}\b' . trim(preg_quote($search)) . '\b.{0,20}\b/is';
preg_match($regex, $string, $matches);
echo $matches[0];

utahnix

That line doesn't work...

This is what I have so far:

(\W\S){0,$maxoff}$word(\W\S){0,$maxoff}

Where $maxoff is the maximum word offset. I want to go left and right $maxoff words, not $maxoff characters. That's where the problem comes in.

But the line I have above, although work most of the time, misses about 10-15% of the time.

Ideas to make this work?