I did a quick spike thinking this would be easy 🙄
The sticking point is that pg's 11 and 12 by default give ownership of pg_replication_origin and sql_languages to test_user. pgbedrock itself should be able to work against postgres 11 and 12 as-is, but updating the tests to have the correct initial state is a bit of a chore.
One idea is to add "except" support to ownership as well as schemas/tables - this would make the changes less verbose than trying to enumerate all the stuff in information_schema. For schemas/tables, we depend on the fact that we can scope tables under "read" and "write" which gives us a way to add "except" to the mix. We don't have this convenient "in" for ownership. We bounced around the idea of extending the DSL to allow a dict in addition to a string name / glob.
Unfortunately, I gotta drop this, but I wanted to push what I had to give anyone interested in this a starting point to jump from.
Hi, is there any chance to see this PR merged ? We are waiting for pg 11 support, and to be honest not confindent on a PR owned by a ghost ;-)
Thanks for all your work on this project
I did a quick spike thinking this would be easy 🙄
The sticking point is that pg's 11 and 12 by default give ownership of
pg_replication_origin
andsql_languages
totest_user
. pgbedrock itself should be able to work against postgres 11 and 12 as-is, but updating the tests to have the correct initial state is a bit of a chore.One idea is to add "except" support to ownership as well as schemas/tables - this would make the changes less verbose than trying to enumerate all the stuff in
information_schema
. For schemas/tables, we depend on the fact that we can scope tables under "read" and "write" which gives us a way to add "except" to the mix. We don't have this convenient "in" for ownership. We bounced around the idea of extending the DSL to allow a dict in addition to a string name / glob.Unfortunately, I gotta drop this, but I wanted to push what I had to give anyone interested in this a starting point to jump from.