can this phpbb query be optimized? - mysql

Question

Here's some code adapted from phpBB. Near as I can tell it's trying to delete all topics wherein the only poster is the target user.
(note that for testing purposes I changed the final query from a DELETE to a SELECT)
<?php
$user_id = 66275;
mysql_connect('localhost', 'username', 'password');
mysql_select_db('db_name');
$start = microtime(true);
$total = 0;
define('POSTS_TABLE', 'phpbb_posts');
define('TOPICS_TABLE', 'phpbb_topics');
$sql = 'SELECT topic_id, COUNT(post_id) AS total_posts
FROM ' . POSTS_TABLE . "
WHERE poster_id = $user_id
GROUP BY topic_id";
$result = mysql_query($sql);
$topic_id_ary = array();
while ($row = mysql_fetch_assoc($result))
{
$topic_id_ary[$row['topic_id']] = $row['total_posts'];
}
mysql_free_result($result);
if (sizeof($topic_id_ary))
{
$sql = 'SELECT topic_id, topic_replies, topic_replies_real
FROM ' . TOPICS_TABLE . '
WHERE ' . sql_in_set('topic_id', array_keys($topic_id_ary));
$result = mysql_query($sql);
$del_topic_ary = array();
while ($row = mysql_fetch_assoc($result))
{
if (max($row['topic_replies'], $row['topic_replies_real']) + 1 == $topic_id_ary[$row['topic_id']])
{
$del_topic_ary[] = $row['topic_id'];
}
}
mysql_free_result($result);
if (sizeof($del_topic_ary))
{
$sql = 'SELECT topic_id FROM ' . TOPICS_TABLE . '
WHERE ' . sql_in_set('topic_id', $del_topic_ary);
$result = mysql_query($sql);
while ($row = mysql_fetch_assoc($result))
{
$total++;
echo $row[topic_id] . "\r\n";
}
}
}
function sql_in_set($field, $array, $negate = false, $allow_empty_set = false)
{
if (!sizeof($array))
{
if (!$allow_empty_set)
{
// Print the backtrace to help identifying the location of the problematic code
$this->sql_error('No values specified for SQL IN comparison');
}
else
{
// NOT IN () actually means everything so use a tautology
if ($negate)
{
return '1=1';
}
// IN () actually means nothing so use a contradiction
else
{
return '1=0';
}
}
}
if (!is_array($array))
{
$array = array($array);
}
if (sizeof($array) == 1)
{
#reset($array);
$var = current($array);
return $field . ($negate ? ' <> ' : ' = ') . $var;
}
else
{
return $field . ($negate ? ' NOT IN ' : ' IN ') . '(' . implode(', ', $array) . ')';
}
}
$elapsed = microtime(true) - $start;
echo "\r\ntook $elapsed seconds";
echo "\r\ngot $total rows back";
?>
This does three queries. First gets all the topics the target user has posted in and the number of times they've posted in each topic. The second gets how many replies each topic in the first query actually has. Then there's some PHP code to see which topics have had all their posts made by the target user. After that the code (prior to my changes) DELETEs all those topics.
Overall it seems to me that this could be written better by doing something like this:
SELECT t.topic_id
FROM phpbb_topics AS t
JOIN phpbb_posts AS p1
ON p1.topic_id = t.topic_id
AND p1.poster_id = $poster_id
LEFT JOIN phpbb_posts AS p2
ON p2.topic_id = t.topic_id
AND p2.poster_id <> $poster_id
WHERE p2.poster_id IS NULL;
Or maybe this:
SELECT t.topic_id
FROM phpbb_topics AS t
JOIN phpbb_posts AS p1
ON p1.topic_id = t.topic_id
AND p1.poster_id = $poster_id
AND t.topic_poster = $poster_id
AND t.topic_last_poster_id = $poster_id
LEFT JOIN phpbb_posts AS p2
ON p2.topic_id = t.topic_id
AND p2.poster_id <> $poster_id
WHERE p2.poster_id IS NULL
Testing this is actually quite difficult thanks to MySQLs caching but... from what testing I have been able to do it seems like the way phpBB is currently doing it is in fact faster. Which is surprising to me.
Any ideas?

Answer 1

It looks to me like you are on the right track. Try adding indexes to all the columns you are using in the joins as this can often drastically increase the speed of joins.