Hi,
I'm considering changing my database schema upgrade routines to recreate
functions and all objects depending on them (e.g. triggers, views) just after
the database server is started. It would make the routines much simpler since
they wouldn't have to depend on the history of the schema.
Does anyone has any experience with recreating all functions and triggers to
upgrade a database schema assuming that during the upgrade there are no client
connections to the database?
Does anyone see any problems with such approach in terms of consistency and
performance? I'm not familiar with PostgreSQL internals and I'm not sure how it
would affect the planner when using various function volatile categories.
Konrad
