note 71153 deleted from function.rand by felipe

Note Submitter: dsoussan at gmail dot com 

----

The problem with these random row functions below is that they require you to fetch all the query results first. Now this may be OK for a small resultset - but in that case why not just use the SQL  'ORDER BY RAND() LIMIT 10' to do the job.

The real problem arises when you want a small random selection from a large query resultset; in that case the SQL RAND() function causes the query to run very slowly. It is then that we need to do the processing in php, and fetching every row from a large query resultset first rather defeats the purpose.

The following is a far simpler and more efficient solution to this problem. It will also cope with queries that do not return distinct rows.

<?php
function randomselection($c, $sql) {
	// run the query
	$result=mysql_query($sql) or die($sql);	
	// get the upper limit for rand()
	$up = mysql_num_rows($result)-1;
	
	// check that there is more data than rows required
	if ($up <= $c) {
		while ($row=mysql_fetch_assoc($result)) {
			$selection[] = $row;
		}
		return $selection;
	}
	
	// set up array to hold primary keys and elliminate duplicates
	// since random DOES NOT mean unique
	// or the query may not return DISTINCT rows
	$keys = array();
	// get random selection
	while (count($keys) < $c) {
		// get random number for index 
   		$i = rand(0, $up);
   		// move pointer to that row number and fetch the data
   		mysql_data_seek($result, $i);
   		$row = mysql_fetch_assoc($result);
  		//  check if the primary key has already been fetched
  		if (!in_array($row['id'], $keys)) {
     		// store the row
     		$selection[] = $row;
    		// store the id to prevent duplication
    		$keys[] = $row['id'];
    	}
    }
    return $selection;
}
?>

Fully tested and certified bug-free ;)