MYSQL ORDER BY with an array of values in WHERE clause - mysql

Question

I am working on a part of a search script where the submitted search string is parsed and each significant term placed into an array. The array is then looped through in the WHERE clause to search several columns in the MYSQL database. Here's a sample SQL code:
$sql = "SELECT title, question, tag1, tag2, tag3, tag4, tag5 FROM question WHERE ";
while(list($key,$val)=each($split_stemmed)){
if($val<>" " and strlen($val) > 0){
$sql .=
"(title LIKE '$val%'
OR question LIKE '$val%'
OR tag1 LIKE '$val%'
OR tag2 LIKE '$val%'
OR tag3 LIKE '$val%'
OR tag4 LIKE '$val%'
OR tag5 LIKE '$val%') OR";
}
}
$sql=substr($sql,0,(strLen($sql)-3));
$sql .= "GROUP BY q_id ORDER BY
((title LIKE '$val%') +
(question LIKE '$val%') +
(tag1 LIKE '$val%') +
(tag2 LIKE '$val') +
(tag3 LIKE '$val%') +
(tag4 LIKE '$val%') +
(tag5 LIKE '$val%')) desc, title asc";
The problem I am having is getting the ORDER BY to work correctly. The ORDER BY's purpose is to arrange the results of the query from the order of most hits on the search string to the least. My assumption is that I somehow need to loop through the array again in the ORDER BY clause, but am not sure of how to do this or if I am even correct in that assumption. Any helpers out there?
I know I can probably do this simpler using mysql full text search but the tables being used are InnoDB, so I dont think that is an option as of yet.

Answer 1

A couple of issues to consider...
1) a NULL value in any of the columns tag1, tag2, tag3... in the ORDER BY expression is going to return a NULL for the entire expression. Consider:
SELECT 0 + NULL + 1 + 0
If all of those columns are defined as NOT NULL, then this isn't an issue. But more generally, you'd want to insulate from NULL values...
ORDER BY ( IFNULL(title LIKE '$val%'),0) +
IFNULL(question LIKE '$val%'),0) +
IFNULL(tag1 LIKE '$val%'),0) +
IFNULL(tag2 LIKE '$val' ),0) +
-or alternatively-
ORDER BY ( IFNULL(title ,'') LIKE '$val%') +
IFNULL(question,'') LIKE '$val%') +
IFNULL(tag1 ,'') LIKE '$val%') +
IFNULL(tag2 ,'') LIKE '$val' ) +
)
(Seems like you're missing a % on tag2 based on the pattern, but it's entirely possible that omission is intentional
2) It's not at all clear why you need a GROUP BY q_id in your query.
I'd recommend you test by taking the expression in the ORDER BY clause, and copy it to your SELECT list, and run the query to see the values that it's returning.
In MySQL, if you add that expression to the SELECT list and give it an alias, you can reference the alias on the ORDER BY.
SELECT expr AS match_count, ...
FROM
ORDER BY match_count DESC

Answer 2

The functionality you're looking for -- searching multiple columns for search terms and ordering by relevance -- is exactly what full-text search was designed for. If at all possible, you should use this as it will reduce your headaches in the long run.
If you absolutely must implement this way, then it's necessary to move the LIKE statements into the SELECT clause so that you can sum them. Like this:
SUM (CASE WHEN title LIKE '$val%' THEN 1 ELSE 0 END
+ CASE WHEN question LIKE '$val%' THEN 1 ELSE 0 END
+ CASE WHEN tag1 LIKE '$val%' THEN 1 ELSE 0 END
... etc
) AS relevance
Then to select only the matches, you could use HAVING relevance > 0. And to order, you would simply use ORDER BY relevance DESC (since the relevance column as above will give you a count of the number of matching columns).