[
https://issues.apache.org/jira/browse/CALCITE-3550?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16988264#comment-16988264
]
Julian Hyde commented on CALCITE-3550:
--------------------------------------
Can we adopt the naming convention that plural class names are only for
utilities? I know SqlTypeAssignmentRules doesn't quite adhere to that (though
it has some static methods), but new classes should. Also, I think the new base
class should be an interface.
I agree we don't need {{synchronized}} any more. Perhaps Guava's
{{Suppliers.memoize}} will help with race conditions.
I agree that {{ThreadLocal}} is sufficient. It will avoid changing lots of
code. Mention the ThreadLocal in the SqlValidator javadoc, though.
> Make SqlTypeAssignmentRules conversion mapping pluggable
> --------------------------------------------------------
>
> Key: CALCITE-3550
> URL: https://issues.apache.org/jira/browse/CALCITE-3550
> Project: Calcite
> Issue Type: Improvement
> Components: core
> Affects Versions: 1.21.0
> Reporter: Danny Chen
> Priority: Major
> Labels: pull-request-available
> Fix For: 1.22.0
>
> Time Spent: 20m
> Remaining Estimate: 0h
>
> In CALCITE-2302, we introduce implicit type coercion for Calcite builtin
> conversion behaviors, we also expand the conversion matrix to support more
> type conversion compared to the old matrix [1], that means, we implement a
> conversion matrix of our Calcite style.
> As a library of other sql engines, i think we should make the conversion
> matrix pluggable, so that each engine can customize if the conversion is
> valid from a type to another.
> We have made the TypeCoercion pluggable/customizable, so if this matrix is
> also pluggable, then, the whole conversion behavior can be customized.
> That would give the type conversion of Calcite a pretty good flexibility.
> [1]
> https://docs.google.com/spreadsheets/d/1GhleX5h5W8-kJKh7NMJ4vtoE78pwfaZRJl88ULX_MgU/edit?usp=sharing
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
