Laurence Isla

I submitted the issue upstream here: https://github.com/yesodweb/wai/issues/976

> But somehow, I was not able to reproduce this in the spec tests here, yet. The fix should be simple, by returning fldNames instead of ["*"]. But we'd need...

Treating the `null` value in the query string as a text string is a current limitation for `isdistinct`. It also happens for `eq`, `like` and similar comparison operators. Only `is`...

> Note that quotes already dropped inside an `or` (or similar construct): `or=(data.eq."foo")` is the same as `or=(data.eq.foo)` That's interesting, then maybe we aren't being completely consistent in parsing operators,...

> How do I assign the list to the config to test it? Using a string doesn't work? ```python env = { **defaultenv, "PGRST_DB_HOISTED_TX_SETTINGS": "work_mem,statement_timeout", } ```

I tried to reproduce this issue to no avail at first. But the `created_at` name made me realize that it could be used for other tables too, so I overloaded...

> Postgrest is unable to determine relationship between user-defined type and table using computed relationship function. I think that, by design, relationships should only be created between tables/views. Since `search_source`...

Ah, I see, it's an interesting use case. My opinion is that a `TYPE` shouldn't be candidate for resource embedding, but I'm not sure if the rest of the team...

To test this using our fixtures: ```bash curl "localhost:3000/entities?select=name,child_entities(grandchild_entities())&child_entities.grandchild_entities=not.is.null&child_entities=not.is.null" ``` The response is: ```json [{"name":"entity 1","child_entities":[{}, {}]}] ``` The expected one is: ```json [{"name":"entity 1"}] ```

@steve-chavez Managed to make it reload the pool configurations, but I'm not very familiarized with the process: I'm not sure if it needs to destroy the old pool or where....