Hm, thinking again, I guess Tom Lane is right
>> Surely the initialization code would have to be run anyway ... and if
>> the function does import a pile of modules, do you really want to cache
>> all that in its pg_proc entry? What happens if some of the modules get
>> updated later?
Besides, what happens if you store compiled bytecode in a table, then
upgrade the python interpreter to a new version... would it be compatible
? I suppose so, but I don't really know...Persistent connections should be used anyway, this makes the RAM caching
good...