Michael Bridge

Let's imagine a multi-tenant application where each tenant operates within its own PG schema, and each tenant PG schema is exactly the same as every other tenant schema. Say there...

After further consideration, I agree with the search path approach (which is a clean solution in other related contexts). I am well aware of the advantages and limitations (as referenced...

After further investigation, setting `search_path` on a per-request basis is relatively simple: ``` pgSettings: req => { [ ... ] return { 'search_path': jwtClaims['tenant'] + ', public', } } ```...

Agreed -- I intended to request the cleanest entry point for revising core. After further review, this would seem to be `QueryBuilder.build`, so as to leave the introspection mechanics untouched.

As mentioned by [robertjpayne](https://github.com/robertjpayne), this use case is easily solvable with `search_path` if the Postgres schema is not forcibly included. Are any of the maintainers able to suggest how we...

I've been poking around the code looking for the logic that defaults to qualifying table identifiers with schema names. @janpio any suggestions? I think a very simple patch can likely...

I think maybe this is buried in a binary; I've heard nothing from the maintainers and moved on to a more transparent lib.

Thanks for your response. While I follow a similar task for scheduled work, this flow is extremely useful for unscheduled work or minor PRs that don't have a corresponding ticket....

Agreed this would be a big lift. I experimented with wrapping the generated client. #2 is quite doable but #1 proved tricky to implement in all circumstances (especially for nested...

~~After further consideration, I think the simpler approach may be to wrap the input as follows:~~ ``` type CompanyUpdate { id: ID! display_name: String! description: String } type Mutation {...