MDB structure question

|Hi, a while back I introduced a structural change to MDB that I borrowed from Manuels latest Metabase beta: All the management related methods were moved to a separate class. If the method is now called MDB checks if an instance of that class has already been created and if not creates it and then the call is "proxied" to the class. So for the user nothing really changed. The structure sort of looks like this now MDB_common | | |-(proxy)- MDB_manager_database_class | | | | | ---- MDB_manager_mysql_class | --- MDB_mysql So MDB_mysql extends MDB_common and MDB_manager_mysql_class extends MDB_manager_database_class and MDB_common proxies calls to MDB_manager_database_class For example in MDB_common alter table now looks like: function alterTable($name, &$changes, $check) { $result = $this->loadManager("Alter table"); if (MDB::isError($result)) { return ($result); } return ($this->manager->alterTable($this, $name, $changes, $check)); } So the number of methods really stay the same. The Mysql driver is smaller now though in terms of code (down from 80k its now at 50k). Obviously calling a management method now has a bit more overhead. The management class however is only instantiated once. So after the first call the overhead is much less for the next calls. I did some benchmarks (using John Lims benchmark suite) with and without this structure with the latest MDB version and the results are as follows: MDB with proxy: Queried 400 times in a couple runs evened out to about 9.83 seconds MDB without proxy: Queried 400 times in a couple runs evened out to about 9.88 seconds The benchmark suite is not all that exact and I am not working my butt of to ensure that it can produce totally accurate results (nor am I experienced enough in running benchmarks to begin with). But it looks like using the proxy (as it can be found in the current MDB release) seems to give a very little performance boost if you don’t use the management methods. It does add a bit of structural complexity though. So the question is: Should I remove the proxy stuff or not? As an aside: Stripping down both the common.php and mysql.php to only the needed methods (5k and 11k respectively) resulted in about 9.78 seconds. All in all I expected a much greater influence from parsing and compiling. With a script cacher this should be close to no difference? Best regards, Lukas Smith smith@dybnet.de _______________________________ DybNet Internet Solutions GbR Reuchlinstr. 10-11 Gebäude 4 1.OG Raum 6 (4.1.6) 10553 Berlin Germany Tel. : +49 30 83 22 50 00 Fax : +49 30 83 22 50 07 www.dybnet.de info@dybnet.de