ajwerner

TFTR! bors r+

> I understand the motivation for this request, but the choice to represent messages as pointers was made long ago and is not something that can be easily undone. I...

> The use of mutexes in the messages is principally for record keeping used in reflection, and in fast-path encoding implementations. Extracting all of that functionality seems infeasible at this...

> I think approaching this situation as “some programs, and design principles are going to have highly nested protobufs, what can we do to not make that suck?” There's something...

Okay. I acknowledge that my approach in this issue has been myopically focused on the issue of value types in my more general attempt to move cockroach, and ideally other...

I feel like you don't need to tell sqlc about the exact schema because it doesn't ultimately manage the schema. All of the features you just mentioned are schema features...

> As long as the syntax you're using is valid PostgreSQL, sqlc should be able to handle it. We don't support any CockroachDB extensions at this time. At the end...

Sorry about the primary key ordering comment. Yeah, these are all extensions.

Much work has been done in recents months to make the cockroach SQL parser depend on less code. It's newly plausible that we'll have a `go get`-able parser one day...

@nickjackson the parser library now exists! See https://github.com/cockroachdb/cockroachdb-parser.