Are these bounds actually required?

FromSqlOwned is required for#[derive(FromSql)] which is needed to run the test. Or were you asking about the derivedDebug andPartialEq?

Edit: I also put the bounds in the two different positions to test that bounds and where clauses work as expected.

Derive logic should normally insert those bounds on the generated impls as necessary - you don't needwhere T: Clone at the struct definition when using#[derive(Clone)] for example.

Just to make sure I understand the change request, the idea is add a FromSql<'a> bound on any generic type in the FromSql derive logic?

Yep.

So I started implementing this, and ran into an ambiguity (where this differs fromClone, which is in the rust prelude). I need to create a path to the FromSql and ToSql traits, however it could be one of a few options:

1. From the postgres_derive crate (postgres_derive::FromSql<'a, T>, but I think most uses will not import this crate directly)
2. From the postgres_types crate (postgres_types::FromSql<'a, T>)
3. Some other crate that re-exports or renames those types (no way to use a path from a crate name)
4. Assume that it'sused somewhere (FromSql<'a, T>)
   Any suggestions on which assumption to use?

Here's how the path is already produced in the existing logic:https://github.com/sfackler/rust-postgres/blob/master/postgres-derive/src/fromsql.rs#L90

Thanks for pointing that out. Based on that, looks like it was option (2). Branch is updated, and tests pass locally.