@muzarski Please investigate the failing CI, especially cassandra tests.

Is this still needed after yesterdays changes to serialization in cpp-rust-driver?

Is this still needed after yesterdays changes to serialization in cpp-rust-driver?

As I see it, one of scylla_cql's roles is to be a low-level API for interacting with the DB, also to be used by interested parties independently from scylla crate. For that to be possible, all its functionality should be public.

Is this still needed after yesterdays changes to serialization in cpp-rust-driver?

As I see it, one of scylla_cql's roles is to be a low-level API for interacting with the DB, also to be used by interested parties independently from scylla crate. For that to be possible, all its functionality should be public.

We can export a function after 1.0 if needed, but we can't un-export it, and exporting everything makes it much harder to make any changes.
If those functions are not needed by either scylla or cpp-rust-driver then I see no reason to export them until there is a use case for that.

Is this still needed after yesterdays changes to serialization in cpp-rust-driver?

As I see it, one of scylla_cql's roles is to be a low-level API for interacting with the DB, also to be used by interested parties independently from scylla crate. For that to be possible, all its functionality should be public.

We can export a function after 1.0 if needed, but we can't un-export it, and exporting everything makes it much harder to make any changes. If those functions are not needed by either scylla or cpp-rust-driver then I see no reason to export them until there is a use case for that.

scylla_cql's public API is not public API of scylla, so I don't see any hazard in exporting everything from scylla_cql. It's on scylla crate to control re-exports from scylla_cql properly.

Is this still needed after yesterdays changes to serialization in cpp-rust-driver?

As I see it, one of scylla_cql's roles is to be a low-level API for interacting with the DB, also to be used by interested parties independently from scylla crate. For that to be possible, all its functionality should be public.

We can export a function after 1.0 if needed, but we can't un-export it, and exporting everything makes it much harder to make any changes. If those functions are not needed by either scylla or cpp-rust-driver then I see no reason to export them until there is a use case for that.

scylla_cql's public API is not public API of scylla, so I don't see any hazard in exporting everything from scylla_cql. It's on scylla crate to control re-exports from scylla_cql properly.

https://rust-lang.github.io/api-guidelines/necessities.html#public-dependencies-of-a-stable-crate-are-stable-c-stable

When we release scylla 1.0 scylla-cql will need to already be 1.0 and so have its API stable. Good luck making any changes after that if you export everything.

scylla-cql, like any other library in existance, has its public API. In scylla-cql public API for vints is SerializeValue implementation and implementations for deserialization traits - and this is what cpp-rust-driver is using now IIUC.
vint_[en/de]code are internal implementation detail and unless there is a need for them to be public they should not be.

I'll close it. I agree that for now, there is no need to make vint_ util functions public.