why Zend_Db_Adapter is much slower than mysql_query - php

Question

I use Zend Framework in my project. I need to insert multiple records and I found that Zend_Db suprisingly slower thatn my_sql query (several times), that made me think I did something wrong.
Here are two examples.
Zend_Db_Adapter:
$startTime = microtime(true);
$db = Zend_Db_Table::getDefaultAdapter();
$db->beginTransaction();
$dateAdded = date('Y-m-d H:i:s');
$lastChanged = $dateAdded;
foreach ($importDataNamespace->data as $subscriberNum => $subscriber)
{
foreach ($fieldsMap as $fieldNumber => $fieldTag) {
if (isset($subscriber[$fieldNumber])) {
$subscriberData[$fieldTag] = $subscriber[$fieldNumber];
} else {
$subscriberData[$fieldTag] = '';
}
}
$query = 'INSERT INTO subscribers (list_id, account_id, email_address, first_name, last_name, date_added, last_changed) ' .
'VALUES (' . 52 . ', ' . 29 . ', ' . $db->quote($subscriberData['EMAIL']) . ', ' . $db->quote($subscriberData['FNAME']) .
', ' . $db->quote($subscriberData['LNAME']) . ', ' . $db->quote($dateAdded) . ', ' . $db->quote($lastChanged) . ')';
$db->query($query);
}
$db->commit();
$this->view->time = microtime(true) - $startTime;
Example with mysql_query:
$startTime = microtime(true);
$user = 'root';
$password = 'password';
$db = 'database';
$connect = #mysql_connect('localhost',$user,$password) or die("Failed to connect database");
#mysql_select_db($db) or die("Failed to select database");
$dateAdded = date('Y-m-d H:i:s');
$lastChanged = $dateAdded;
$result = mysql_query('SET autocommit = 0');
foreach ($importDataNamespace->data as $subscriberNum => $subscriber)
{
foreach ($fieldsMap as $fieldNumber => $fieldTag) {
if (isset($subscriber[$fieldNumber])) {
$subscriberData[$fieldTag] = $subscriber[$fieldNumber];
} else {
$subscriberData[$fieldTag] = '';
}
}
$query = 'INSERT INTO subscribers (list_id, account_id, email_address, first_name, last_name, date_added, last_changed) ' .
'VALUES (' . 52 . ', ' . 29 . ', \'' . mysql_real_escape_string($subscriberData['EMAIL']) . '\', \'' . mysql_real_escape_string($subscriberData['FNAME']) .
'\', \'' . mysql_real_escape_string($subscriberData['LNAME']) . '\', \'' . $dateAdded . '\', \'' . $lastChanged . '\')';
mysql_query($query);
}
$result = mysql_query('SET autocommit = 1');
$result = mysql_query('COMMIT;');
$this->view->time = microtime(true) - $startTime;
In the first case it took 14.8 seconds, in the second 3.7.
Could you tell me why does it happend and what do you do wrong?
If I delete any quote for Zend_Db it took 12 seconds from 14 with quote, but it's still much more slower than with mysql_query:
$startTime = microtime(true);
$db = Zend_Db_Table::getDefaultAdapter();
$db->beginTransaction();
$dateAdded = date('Y-m-d H:i:s');
$lastChanged = $dateAdded;
foreach ($importDataNamespace->data as $subscriberNum => $subscriber)
{
foreach ($fieldsMap as $fieldNumber => $fieldTag) {
if (isset($subscriber[$fieldNumber])) {
$subscriberData[$fieldTag] = $subscriber[$fieldNumber];
} else {
$subscriberData[$fieldTag] = '';
}
}
$query = 'INSERT INTO subscribers (list_id, account_id, email_address, first_name, last_name, date_added, last_changed) ' .
'VALUES (' . 52 . ', ' . 29 . ', \'' . $subscriberData['EMAIL'] . '\', \'' . $subscriberData['FNAME'] .
'\', \'' . $subscriberData['LNAME'] . '\', \'' . $dateAdded . '\', \'' . $lastChanged . '\')';
$db->query($query);
}
$db->commit();
$this->view->time = microtime(true) - $startTime;
Thank you for any information about this issue.
This code takes about 0.065 seconds with mysql_query:
$dateAdded = date('Y-m-d H:i:s');
$lastChanged = $dateAdded;
$startTime = microtime(true);
$result = mysql_query('BEGIN');
for ($i = 0; $i < 100; $i++) {
$email = 'test_ ' . $i . '#gmail.com';
$query = 'INSERT INTO subscribers (list_id, account_id, email_address, first_name, last_name, date_added, last_changed) ' .
'VALUES (' . 52 . ', ' . 29 . ', \'' . mysql_real_escape_string($email) . '\', \'' . mysql_real_escape_string($firstName) .
'\', \'' . mysql_real_escape_string($lastName) . '\', \'' . mysql_real_escape_string($dateAdded) . '\', \'' . mysql_real_escape_string($lastChanged) . '\')';
mysql_query($query);
}
$result = mysql_query('COMMIT');
$time = microtime(true) - $startTime;
echo 'Using mysql_query: ' . $time . '<br />';
exit();
Code with benchmark of Zend_Db_Adapter (I didn't even use quote in this case):
$db = Zend_Db_Table::getDefaultAdapter();
$db->getProfiler()->setEnabled(true);
$profiler = $db->getProfiler();
$startTime = microtime(true);
$db->beginTransaction();
for ($i = 0; $i < 100; $i++)
{
$email = 'test_' . $i . '#gmail.com';
$query = 'INSERT INTO subscribers (list_id, account_id, email_address, first_name, last_name, date_added, last_changed) ' .
'VALUES (' . 52 . ', ' . 29 . ', \'' . $email . '\', \'' . $firstName .
'\', \'' . $lastName . '\', \'' . $dateAdded . '\', \'' . $lastChanged . '\')';
$db->query($query);
}
$db->commit();
$time = microtime(true) - $startTime;
echo 'Time of transaction Zend_Db_Adapter query: ' . $time . '<br />';
echo 'Total time ' . $profiler->getTotalElapsedSecs() . '<br />';
$count = 0;
$totalTime = 0;
foreach ($profiler->getQueryProfiles() as $query) {
$count++;
$elapsedTime = $query->getElapsedSecs();
$totalTime += $elapsedTime;
echo $count . ' ' . $elapsedTime . ' ' . $query->getQuery() . '<br />';
}
echo 'Sum time: ' . $totalTime . '<br />';
Here are some results:
Time of transaction Zend_Db_Adapter query: 0.23094701767
Total time 0.0949234962463
1 0.00199699401855 connect
2 0.000336885452271 begin
3 0.000540018081665 INSERT INTO subscribers (list_id, account_id, email_address, first_name, last_name, date_added, last_changed) VALUES (52, 29, 'test_0#gmail.com', 'John', 'Clinton', '2011-01-28 15:25:21', '2011-01-28 15:25:21')
4 0.000504016876221 INSERT INTO subscribers (list_id, account_id, email_address, first_name, last_name, date_added, last_changed) VALUES (52, 29, 'test_1#gmail.com', 'John', 'Clinton', '2011-01-28 15:25:21', '2011-01-28 15:25:21')
It's very strange. The time of transacation to insert 100 records is 2.5 times more than executing of all queries.
If I try to meach the time of forming strings in the loop it (if we delete query) doesn't take so much time.

Answer 1

I think that one reason is that you execute $db->quote() too many times, which is unnecessary. Do you know that $db->quote() can take an array as its argument and you can basically reduce calls to $db->quote() to only one. In addition in your mysql_query version you do not escape $dateAdded and $lastChanged, while in zend_db version you do.
EDIT: Added an example below
$db = Zend_Db_Table::getDefaultAdapter();
$input = array(
'a' => "asd'lfksd",
'b' => "asdlfk'sdfasdf",
'c' => "asd fds f saf'sdfsd",
'd' => "asd fds f saf'sdfsd"
);
// separate calls to quote
$startTime = microtime(true);
$db->quote($input['a']);
$db->quote($input['b']);
$db->quote($input['c']);
$db->quote($input['d']);
$totalTime1 = microtime(true) - $startTime;
// one call to quote
$startTime = microtime(true);
$db->quote($input);
$totalTime2 = microtime(true) - $startTime;
// show results
var_dump("Sperate calls are ". $totalTime1/$totalTime2 . " times slower");
//output: string 'Sperate calls are 3.0875831485588 times slower' (length=46)

Answer 2

mysql_* functions are natives to PHP , so they are very fast.
Zend_Db_Adapter is working with PDO so you have a first absraction layer with PDO and a second with Zend_DB_Adapter.
More layer abstraction you 've got , more the code is slowing down.
That why the MVC framework in general are more slower than procedural code.
Try again your benchmark with prepared statement and a cache mechanism and you should be close of the mysql_* functions

Answer 3

This looks like you don't have a metadata cache in Zend_Db.