[
https://issues.apache.org/jira/browse/OPTIQ-333?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14055508#comment-14055508
]
Jacques Nadeau commented on OPTIQ-333:
--------------------------------------
Iteration seems like a poor solution since we already have what I would
describe as a better solution in the attached patch. If you're concerned about
the surface of the SPI, let's try to figure out how to abstract out what is
really necessary functionality for upstream operations versus what isn't. I
would suggest that the primary difference between my proposed SimpleOptiqSchema
and CachingOptiqSchema is that the latter is a set of optimizations based on a
particular assumed use which is inconsistent with how we want to use the
library.
> Abstract OptiqSchema as a base class and move the existing implementation to
> a separate class
> ---------------------------------------------------------------------------------------------
>
> Key: OPTIQ-333
> URL: https://issues.apache.org/jira/browse/OPTIQ-333
> Project: Optiq
> Issue Type: Improvement
> Reporter: Jacques Nadeau
> Assignee: Jacques Nadeau
>
> The implementation of OptiqSchema can make users jump through many hoops
> because it holds a complex hold internal state. A couple of times now,
> efforts to move Apache Drill to newer versions of Optiq has required all
> schema management code to be rewritten as the nature of the internally
> maintained state has changed. As such, this JIRA is about abstracting out
> the core interface and providing an alternative, near stateless
> implementation. Existing tools will continue to leverage the traditional
> existing implementation.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
