Evan Huus

Without digging too much into the implications my default preference would be: - Keep validation and coercion together since in a lot of more complex cases coercing requires validating of...

An interface should only be hidden if it has no references, not if it has no implementations? I'd think ```graphql type QueryRoot { foo: Foo! } interface Foo { bar:...

🤔 without thinking about the implementation too much I'd think `{ foo { bar } }` should be an executable query on the schema I provided. But I guess `{...

I am thinking e.g. of mysql, where you have a connection pool; the individual connections are synchronous but the pool itself is concurrent. I was assuming the driver as a...

OK, adapter mutexing takes care of the first issue (`retrieve_or_register`) but I believe the circuit-breaker point is still valid since `request_allowed?` and other such methods are expected to be publicly...

Whether a server does the coercion at execution time or build time is largely an implementation detail, but I do think we should specify the more natural behaviour that comes...

Another alternative syntax (probably not great, but I haven't seen it yet so worth thinking about): `field` for fields, `input < field1: Int, field2: Int >` for types, i.e. using...

I gave this another read and while I haven't gone over it in painstaking detail yet, I'm happy with the shape of what's here. @rmosolgo this is one to pay...

Note: did a little more digging and wanted to clarify, this is _not_ the same issue as solved by `g:cpsm_match_empty_query` (or at least, that is not sufficient). Unsetting that will...

Not perfect, but at most `g:ctrlp_mruf_max` entries will be MRUs, and the open buffers are easy enough to determine yourself by asking vim, so everything else is safe to re-order....