Wojciech Przytuła
Wojciech Przytuła
Docstrings in `speculative_execution.rs` need adjustment as well: `host` -> `target`.
I'm convinced that with the new strongly typed serialization framework we really need `BoundStatement` for both compatibility and performance purposes. I believe we should work on it together with exploring...
Ref #977
Have you tested your solution? How? Could you write some unit tests?
I can see that you merged `main` into your branch. As our policy is to rebase, could you instead rebase your branch on `main`, please?
The approach with `enforce_target_node()` is very interesting and it seems to be working, at least conceptually. I'm a bit worried, however, that it may turn out to be somewhat fragile....
> I'm making the call to the non-fallback function of self.fallback inside the pick function, so that should be pretty inexpensive for each message whose shard/batching task has already been...
> No, rather I have an app that receives millions of lines per second to insert. These lines are processed, spread over multiple threads, in a manner that involves just...
> > Are the lines that you get every second merely bind values to be sent with prepared statements with fixed query string? > > Basically yes. There's a bunch...
Maybe you could write an example called, say, `shard_aware_batching.rs`, and present a complete minimal way to utilise `shard_for_statement()` and `enforce_target_node()` - similar to what you put in the docstring, but...