Michael J. Sullivan

Elvis and I discussed this offline and the feeling is that `Foo.link@target` *should* be deduplicated--that all paths to objects should. Making `@target` get deduplicated properly was a simple matter of...

Right, but do we want to create a compatibility hazard by adding new reserved keywords?

What is the semantics of this query supposed to be? Currently it doesn't even parse.

And json slicing behaves incorrectly on NULL subjects: ``` edgedb> select ($0)[0:]; Parameter $0 (Ctrl+D for empty set `{}`): edgedb error: InvalidValueError: cannot slice JSON UNKNOWN Hint: Slicing is only...

We disallow volatile keys there now, so we can't miscompile anything, but we don't *support* it, which we should.

Yeah, running commit hooks on every `git commit` definitely does not match my workflow. I always use `hub pull-request` to create PRs, and wouldn't mind pre-commit hooks *there*. Not sure...

This is related to #3592, which suggests that doing this ought to work. I'll leave this open as an easy "fix the error message" bug, and #3592 can be the...

Also, as a note, we'll need to make sure that we correctly find any constraints we need to generate conflict checks for

Huh, this seems to have been partially fixed at some point. It is broken when we do typename injection, but works without it. This is actually a *problem*, though, because...

Probably materialization related