CodesInChaos

> With Queryable you're going to change the way it's decoded by That's one way to resolve is ambiguity. > we probably not going to implement QueryArg for `u8` This...

I don't like `scalars`, since that's a technical detail, users won't care about (and some might not even know what a scalar is). I like `model` and `data` best, but...

Conceptually it should contain edgedb types a statically typed application would use in its data model (as opposed to edgedb-cli which is mostly dynamically typed via `Value`). Currently they're all...

edgedb-derive causes some complications with that. It has to choose the path from which to import types used by the derived code. * Currently it hardcodes `edgedb-protocol`. This means any...

Something like `&*(mmap.as_ptr() as *const TWeights)` should work. I think this will produce a reference with an unconstrained lifetime, which acts similar to an `&'static T`. To check the alignment,...

A relatively simple improvement would be to add a `separator` parameter to `cases`, which is inserted after an alignment point, similar to how `gap` controls vertical separation. The caller can...

Unfortunately this isn't the only open bug in scalar constraint validation. I don't think #7648 has been fixed yet.

I believe EdgeDB relies on `serializable` to ensure essential invariants (like link cardinality/optionality). So supporting `snapshot` isolation for writing transactions (or writing queries outside transactions) is probably not realistic. However...

It's not a bug, because a tuple will become null as a whole if you attempt to construct it with a null field. However I don't understand why this work-around...

Option 1 to 3 have two problems that option 4 can fix: * non canonical empty ranges, like an exclusive range where `start == end`, or an inclusive range where...