I'd like to hear your personal opinion about PHP DB Connectors

perrohunter

Hey !

I'm in a middle of a fight with myself, this because I'd like to use the fastest DB Connector for PHP and after googling for hours I've seen that these four are the prefered DB Connectors for PHP

1.- MySQL
2.- MySQLi
3.- PDO
4.- AdoDB

I've seen that on some benchmarks PDO is the winner, but I don't understand why, and I really don't know if there are any other connectros.

What do you suggest ? In your experience which connector is the faster one ?
thanks !🙂

NogDog

I would first just drop the basic MySQL extension from the mix, as it does not support as many features as the newer extensions plus it is at best only in maintenance mode as far as development goes and is not enabled by default in PHP5+.

I don't know where AdoDB will fall into this, but my understanding is that the underlying driver is the same for MySQLi and PDO (or will be as of PHP 5.3.0). This article seems to indicate that initially MySQLi was faster than PDO, but in more recent releases the difference has disappeared -- I'm guessing due to the unification of using the same underlying driver now.

This article includes information on the functional differences between the MySQLi, PDO, and MySQL extensions.

Personally, I go with MySQLi if I only need to support MySQL on PHP5+, PDO if I need to be able to support MySQL plus other DBMS's, and MySQL if I need to support PHP4.

Weedpacket

NogDog wrote:
PDO if I need to be able to support MySQL plus other DBMS's,

PDO doesn't provide database abstraction; you're still writing SQL in whatever dialect is used by the DBMS you're connecting to. It also emulates transactions in environments that don't support it natively (and uses the native support when it is there), and a few other things. The only abstraction it does is that it avoids having different interfaces for different DBMSs.

NogDog

Weedpacket;10897794 wrote:
PDO doesn't provide database abstraction; you're still writing SQL in whatever dialect is used by the DBMS you're connecting to. It also emulates transactions in environments that don't support it natively (and uses the native support when it is there), and a few other things. The only abstraction it does is that it avoids having different interfaces for different DBMSs.

This is true -- I was mainly thinking of connectivity rather than abstraction (which PDO does not really do at all).

Currently for abstraction I'm playing around with the Code Igniter framework and the "Active Record Class" it uses for DB abstraction.

laserlight

NogDog wrote:
I was mainly thinking of connectivity rather than abstraction (which PDO does not really do at all).

I may be a little biased because I do not like the prepared statement binding interface provided by MySQLi, but why not choose PDO even if you currently only need to support MySQL on PHP5+? It makes sense to me to not have to switch to another programming interface, which is one of the touted advantages of PDO in the first place.

NogDog

laserlight;10897817 wrote:
I may be a little biased because I do not like the prepared statement binding interface provided by MySQLi, but why not choose PDO even if you currently only need to support MySQL on PHP5+? It makes sense to me to not have to switch to another programming interface, which is one of the touted advantages of PDO in the first place.

I guess my rationalization is that MySQLi "Supports all of MySQL 4.1+" while PDO supports "most'. However, since I don't actually know what part is not included by "most", there's a good chance that most or even all of the time so far it has not mattered. Probably it's mostly inertia: I've learned to use MySQLi and am comfortable with it and have not worked yet with PDO.

Weedpacket

I just reread my post and realised that something dropped out during the editing; it's prepared statement support that PDO emulates, rather than transactions.

The original question is what was the "fastest" connection. At this level, though, the connection speed is not going to be the bottleneck; it's going to be the design of the database and the queries written on it. Even on the "fastest" connection, a 1000-row sequential scan is going to take longer to get results from than a 1000-row indexed scan.

Without knowing anything about what the benchmark tests were it's impossible to say why any might be faster than any other, but being able to compile a query into an execution plan once and run it multiple times with different arguments is going to be preferable to compiling it afresh for every use.