Kuba Podgórski

@skokenes - you can compose `Row` queries with boolean operators, so in your case: ``` Union(Row(stargazer=14), Row(stargazer=22), Row(stargazer=23), ...) ``` `Union`, `Intersect`, `Difference`, `Xor`, `Not` - all of them return...

Million is not high ;) But if we talk about query length, you an take a look into `Range` queries. It really depends on use case.

How about one index per DB and field per Table (because you can merge bitmaps across fields from the same index). In other words, imagine that you have one big...

Couple questions: - do we want to expose _pure_ container's metrics or mainly _driver's_ metrics? - do drivers have/collect any prometheus metrics? - if we want to propagate metrics from...

So, why don't we expose prometheus on each driver. When you install a new driver it will get a random available port on bblfshd which will be mapped to container's...

I think adding a new drivers does not happen very often. Although, registering new node a new node in _prom._ should be a big deal. Moreover if _prom._ tries to...

How about following design: `bblfshd` runs a local prometheus instance and the global prometheus pulls metrics from all `bblfshd` instances. Next to the prometheus server on `bblfshd` (it can be...

Partly yes (as I understood correctly @dennwc concerns). On the other hand, we create new and kill old drivers. Drivers run in containers and have to expose some random port...

In my opinion, we can try implement 2 alternative solutions: - [ ] A simple config file or any other static list of addresses or (if provided) - [ ]...

IMO, it would be totally fine to update/fix `go.mod` file and release `v2.17.0`