pgtap
pgtap copied to clipboard
theory
Problem with function is_empty
The problem is, that another extension, pgsql-tweaks, is installing the function is_empty, but with different returning results, pgtap is returning text, pgsql-tweaks is returning a boolean, please see is_empty source. In addition the installation of either one might fail, if the other one is already installed.
I can and will solve the problem by changing the name from is_empty to is_empty_b (the b for the boolean result), if pgtap is already installed. But it would not solve the problem to install pgtap, if pgsql-tweaks is installed.
And overwriting each others function might cause problems, as they differ in their returning data type.
Just in case, the pg-tweaks repository on GitHub is only a mirror of the one on GitLab.
Just install each extension in their own schema? I always put pgtap into pgtap - and then put that into the search path, when required.
Extensions must really be able to assume their own namespace, they can't expect that no other extension clashes with any of their name.
That might be true to experienced users, like you and me, but not to most people out there using PostgreSQL and extensions. That is at least my experience from consulting work and support on Telegram channels.
Overwriting each other's functions is a limitation of Postgres extensions. Adding them to different schemas is the only supported solution I'm aware of.
Someday I'd like to see extensions get their own namespaces separate from schemas, but I don't know how do-able it'll be to get it designed, agreed to, and implemented.
@sjstoelting can you provide an example of the error you get here? I suspect it wouldn't be terribly difficult to get the community to improve the error report to make it clearer what's happening here.
