in reply to Re^4: DB persistence framework for Perl Classes
in thread DB persistence framework for Perl Classes

And about the portability of Class::DBI, I can be wrong, but you are saying to me that DBI is portable over databases. Well, the SQL is not.

As somebody who fell into the "gotta write my own abstraction layer" trap, I have to disagree here.

Class::DBI exists to allow you to not write SQL; problems with database portability I've run into have been limitations in the DBD layer of DBI and not related to Class::DBI directly. A good example of this is trying to access MS SQL Server; DBD::Sybase doesn't support bind parameters using the FreeTDS library.

Code like this:

my $item = db::Inventory::Item->retrieve( 15 ); if( $item ) { $item->quantity( 15 ); }
works with every database I've tried so far where a Class::DBI driver exists.

If you're worried about database portability, you have to take into account variants of "sequence" methodologies across various databases. For example, both Postgres and Oracle have an object called a sequence, but they access them differently:

-- Oracle: INSERT INTO my_table VALUES( my_sequence.nextval, 'Some Value' ); -- Postgres: INSERT INTO my_table VALUES( nextval( 'my_sequence' ), 'Some Value' );
Heaven forbid you want to use a database where sequences don't exist but you want autoincrementing fields, say MySQL or SQL Server/Sybase:
--MySQL INSERT INTO my_table( seq_field, val_field ) VALUES (NULL, 'Some Value'); SELECT LAST_INSERT_ID(); -- SQL Server INSERT INTO my_table( val_field ) VALUES ( 'Some Value' ); SELECT SCOPE_IDENTITY();
Not to mention that MySQL will insert a requested value into the sequence field (if it doesn't violate a primary key constraint), where SQL Server will barf if you attempt to place any value into an autoincrementing field.

The code to do this with Class::DBI is exactly the same, no matter what database layer you deal with:

my $newRecord = db::Table->create( { val_field => 'Some Value' } );
And you don't have to deal with rewriting your SQL to handle different funkiness associated with different back ends. It Just Works.

Granted, there are things I don't like about Class::DBI:

What I tend to do is use Class::DBI for my user-type interactions where editing and small selection is involved (update my user preferences, edit an item, etc.) and use raw SQL for reporting in the web environment.

For back-end batch processing, I tend to do everything with raw DBI/Perl code unless I have to iterate over every single record and the queries driving my main batch loop aren't JOINed tables.