I believe Mongo adapter is analogous to Druid's although it supports inputs no problem. I understand that Mongo adapter being different from Druid could be a design decision but I hope you can tell me the difference between both designs.
Is Mongo supporting inputs because it has it's own convention and follows the method of visiting input nodes to build the query, while druid follows the Bindable convention which doesn't require converting a node's inputs into the same convention ? I'm just trying to figure out the design difference, pros and cons of both. Thanks, Gelbana On Thu, Mar 15, 2018 at 9:57 PM, Julian Hyde <[email protected]> wrote: > As far as Calcite is concerned, a Druid query is a black box. A blob of > JSON sent to a REST service from which comes data. Therefore it is a leaf > node, like TableScan and Values. > > It has 0 inputs before you copy it, and should still have 0 inputs after > you copy it, hence the assert. > > > On Mar 15, 2018, at 4:25 AM, Muhammad Gelbana <[email protected]> > wrote: > > > > I see that *DruidQuery* intentionally avoids having inputs by not > > overriding *AbstractRelNode.getInputs()* and asserting that no inputs are > > providing while copying itself: > > > > @Override > >> public RelNode copy(RelTraitSet traitSet, List<RelNode> inputs) { > >> *assert inputs.isEmpty();* > >> return this; > >> } > >> > > > > Would someone please explain the purpose of doing this ? > > > > Thanks, > > Gelbana > >
