> You can, and depending on how often you intend to execute said code, it is > probably the better way. It also requires pl/pgsql while CREATE OR REPLACE > "just works" as a single SQL command. It seems easier to give you the > simple answer than to work out the details for the more complex one.
Is it the same from the DB server POV? Meaning it is also the same 1/2 hit depending on the existence? Also performance-wise querying and this method is the same, right?
If you care about performance here I'd suggest you measure it. The absence of catalog bloat is the more meaningful benefit IMO.
If the function doesn't exist CREATE FUNCTION likely wins hands-down. If the function name/signature does exist you are starting to compare apples and oranges.