[
https://issues.apache.org/jira/browse/CALCITE-3093?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16849324#comment-16849324
]
Hongze Zhang commented on CALCITE-3093:
---------------------------------------
Thanks for the thoughtful ideas, [~julianhyde]. (and sorry for the typo :))
To address your suggestion I'll continue to separate the issue to this one and
another - for adding system properties and fixing PlannerImpl respectively. I
see you suggested to have two commits, I suppose we need independent JIRA
tickets for each, so correct me if I am wrong.
Also I agree that the JDBC connections to Calcite are lightweight. To be honest
I can't tell immediately how we can gain from such changes to PlannerImpl. But
from my perspective it makes the call chain to Calcite internal clearer than
before. And let's consider a bit further - do we have a chance to create a
individual maven module for Calcite JDBC driver (maybe named as
calcite-jdbc-driver or something, to distinguish with the Calcite JDBC
adapter)? Say, if users want to simply use Calcite to query their own data
source, they can import both calcite-jdbc-driver and calcite-core in his code.
If only the Planner API is needed, then the importing of calcite-jdbc-driver
can be eliminated. Just a thought.
> Remove JDBC connection calls from PlannerImpl
> ---------------------------------------------
>
> Key: CALCITE-3093
> URL: https://issues.apache.org/jira/browse/CALCITE-3093
> Project: Calcite
> Issue Type: Improvement
> Components: core
> Reporter: Hongze Zhang
> Assignee: Hongze Zhang
> Priority: Major
> Labels: pull-request-available
> Time Spent: 10m
> Remaining Estimate: 0h
>
> Currently the class {{PlannerImpl}} creates JDBC connections[1] internally to
> instantiate {{RelOptPlanner}}. For improving the design I suggest to reverse
> the call chain, for example, we can make JDBC API call PlannerImpl to create
> {{RelOptPlanner}}, {{SqlValidator}}, {{SqlToRelConverter}} instances, etc.
> This JIRA topic is the first step of the improvement - we should decouple the
> use of JDBC API from PlannerImpl. Planner API is designed to be a reusable
> toolkit and we should make it work individually without the help of JDBC API.
> [1]
> https://github.com/apache/calcite/blob/4e1b68e3c0b2a90832cc790c6166cebb14ae2970/core/src/main/java/org/apache/calcite/prepare/PlannerImpl.java#L143-L151
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
