Adrian Mróź

`Cluster` should be just an common interface. Current `Cluster` class should be renamed to `DruidCluster` as we support only Druid database. Additionally we should `NativeCluster`.

Hey! I see your request but sadly I'm not well versed with express to help.

I'm thinking if we could create function there: https://github.com/allegro/turnilo/blob/master/src/client/views/cube-view/cube-view.tsx#L726 ? Then we would create options ad hoc to proof ourselves even more. We pass that code as anonymous lambda anyway:...

Granularity is decided there: https://github.com/allegro/turnilo/blob/08d1a160b14838e30ac69f7173ec3e20c04e5d63/src/common/models/splits/splits.ts#L123-L153 This method is called whenever splits change or filters change. We should limit that.

Hmm, looks like we have a bug there - we search clause by type, we should search by split reference and type. https://github.com/allegro/turnilo/blob/08d1a160b14838e30ac69f7173ec3e20c04e5d63/src/common/models/splits/splits.ts#L136

Observations: - We don't use data values when finding best bucketing. That probably implies that bucketing should be changed only when changes filter on the same dimension. - We don't...

Probably linked with #598

Let's merge it as temporary solution for #536

Check scenario: full reload, _viz registers downloadable dataset_, click on share menu, check if everything is wired correctly e.g. if file downloads

turnilo uses implydata/plywood to send queries to Druid and it does not support joins