Hello, J. Random programmer here, William K. Volkman wrote: > Once an application is designed, the tables and columns don't change > much, why do they need to be able to queried?
Because the programmer is lazy and doesn't want to repeat the list of columns (and types) both in the database schema and in his code. That's why I autoload. I believe the startup cost to be negligible. And then comes my ZeosLib (sqlalchemy for Kylix) experience, where the first time each cursor ran, it would reflect the column information to get field values by name and validate the new values as much client-side as possible. This makes the autoload cost at startup irrelevant. You did raise the point of changing schema information too often. But is that a real objection *against* the ability to autoload? Those "security restrictions" impede autoload for no gain, except data-based customer lock-in. Hardly security related, in my arrogant opinion. > Web applications exacerbate the problem, something that > starts up, does something, then shuts down. CGI? Maybe I've been spending too much time on the web application server camp. > Do we really want/need the overhead of fetching table and column > information each time? No, not really. Don't autoload in that case. -- Luís Bruno _______________________________________________ Sqlalchemy-users mailing list Sqlalchemy-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/sqlalchemy-users
