Re: Why PHPLIB::DB sucks

Hi! Bertrand Mansion wrote:

I looked a bit closer at the code and found out that this 'emulation' actually does not use native sequences at all. Instead it creates a table for all the sequences and locks it exclusively when a need to get a next id arises. This is even more brain-damaged, exclusive lock in PostgreSQL is usually acquired only by DROP TABLE, ALTER TABLE, CREATE INDEX and similar commands. This creates a bottleneck that will nullify all possible speed increases with this 'light' layer.

Isn't that the way they do it in PEAR::DB too ?

No. PostgreSQL driver uses native sequences, MySQL driver uses sequence emulation, but it relies on auto_increment feature and does not require locking -> no bottlenecks.