[
https://issues.apache.org/jira/browse/KYLIN-1525?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15206965#comment-15206965
]
Richard Calaba commented on KYLIN-1525:
---------------------------------------
[~luke_han]
Hi Luke,
I mean no disrespect but I didn't get your argumentation for not having Kylin
XML/A enabled. Olaper is just a XML/A wrapper on top of your existing JDBC
driver so having it incorporated just enhances the portfolio of client UIs
which can be connected to Kylin - increasing potential number of users - dream
of every product owner.
Yes, Kylin query interface is based on SQL and the SQL-schema defines the
star-schema for the Cube. Also as you pointed out -> the end user runs the
query against this table based star schema and Kylin picks up the relevant Cube
(cuboid) to use it for answer of the query. That I got and it is fine - but
still the queries which you can answer are tightly related to the existing
models you define. Where is the difference to XML/A querying style ??? Does it
make difference whether the model is star-schema of relational tables or
whether MDX is used to describe the Cube layout and allow running query against
the cube of your choice ??? Practically if you are building application on top
of your BI Cubes - then anyway you have to rely on knowing particular Cube
design - so I do not see any advantage/disadvantage between both approaches ...
To contradict you 2nd statement - not enough requirements - I think its pretty
simple - allow XML/A queries to be answered by Kylin query interafce. There is
dozen's of front-end UIs which can leverage XML/A enabled sources to do
interesting things with the data (from ETL, through visualizations to whatever
advanced analytics on top of cube extracts).
When I saw Olaper I realized that it should not be that hard to conform to it's
cube metadata model in order to enable Kylin for querying through XML/A
interface - powerful product feature in my opinion.
I do not mind raising the discussion in mailing list - I just do not know how
it works. I tried twice to send emails to mailing list and get no responses.
Seems the JIRA is perfect place to have the discussion of new product features
...
> XML/A for Kylin
> ---------------
>
> Key: KYLIN-1525
> URL: https://issues.apache.org/jira/browse/KYLIN-1525
> Project: Kylin
> Issue Type: New Feature
> Components: Query Engine
> Reporter: Richard Calaba
> Assignee: liyang
>
> It would be great if Kylin can be accessed also by standard OLAP tools
> supporting XML/A so native OLAP queries can be addressed by Kylin.
> I have seen somewhere on the Internet the option to use Pentaho Mondrian with
> some Kylin-patch to achieve this. However I found much simpler solution
> open-sourced here - https://github.com/Wondersoft/olaper .
> The idea would be to include the Olaper code-line and use it to generate
> Cubes Query interface for XML/A directly in Kylin. Kylin knows both - the
> logical structure of the Cube and the storage structure - that's excatly what
> is needed for Olaper to make it work - you need JDBC Driver and then a 2
> metadata files (cube.json and table.json) to describe to the Olaper the
> logical and physical cube structure ...
> I believe that would be very powerful extension opening Kylin to other OLAP
> UI tools (like Saiku i.e.). Today we just rely on Tableau UI consuming Kylin
> through ODBC driver.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
