calcite-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Li, Yang" <>
Subject RE: Kylin
Date Mon, 13 Oct 2014 03:42:00 GMT
Hey, Kylin developer here!  Say hi to everyone.  :D

In addition to query preparation and optimization, we also leverage Calcite to execute query,
run the generated source code and produce results. Currently we are on optiq 0.8 and moving
to a stable calcite release will be a big next step.

Going forward, we like to contribute back to calcite too. There were a few chances, small
bugs fixes etc, but tight internal schedule didn't give those a priority. Now that Kylin has
gone open source, we will have more time to collaborate with big world, especially the great
calcite project.

Looking forward to work with you guys!


-----Original Message-----
From: Julian Hyde [] 
Sent: 2014年10月11日 5:37
Subject: Re: Kylin

Thanks for bringing this up, Dan. I’ve been talking to Kylin’s architects Luke Han and
Xu Jiang for a few months, and have seen demos of it in action. It is impressive.

Kylin is using Calcite/Optiq not just for parsing, but also for optimization. They have written
quite a few RelNodes and rules targeted at their execution environment, which is cubes stored
in HBase. Obviously Calcite does not have data storage, but in terms of query preparation
and optimization, there is considerable overlap with the lattice feature I’ve added to Calcite
over the last few months. (I will be demoing lattices at Strata in NY next week… more about
those soon.)

The Kylin developers are aware of the overlaps and they want to work more closely with Calcite.

I believe that they are considering the project for Apache incubation. If so, that would be
great news. Kylin is an impressive project that fills an important need.


On Oct 10, 2014, at 10:57 AM, Dan Di Spaltro <> wrote:

> Not sure if everyone knows this, but it looks like [1] which 
> was opensourced by Ebay is using the optiq parser.  They have some 
> slides explaining how they built it [2].
> [1]
> [2]
> Pretty neat stuff.
> -Dan
> --
> Dan Di Spaltro

View raw message