[
https://issues.apache.org/jira/browse/OAK-6304?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16042533#comment-16042533
]
Thomas Mueller commented on OAK-6304:
-------------------------------------
> oak.query.ast package... do you remember why it was exported in the first
> place?
No, sorry I don't remember. It could have been due to fulltext, which was later
moved to another package.
> JoinConditionImpl#SPECIAL_PATH_PREFIX
In my view, this could be moved to QueryConstants.
> Would it be worth to consider moving the Operator to an spi package
What about creating a new enum in the Filter interface, that just does what we
need in FilterImpl, like so:
{noformat}
enum ConditionOperator {
EQUAL, NOT_EQUAL, GREATER_THAN, GREATER_OR_EQUAL, LESS_THAN,
LESS_OR_EQUAL, LIKE
};
{noformat}
It would add an indirection (Operator would internally use ConditionOperator).
> org.apache.jackrabbit.oak.query.ast.SelectorImpl: this is the "nasty" one as
> it is a param of the FilterImpl constructor and used the public method
> getSelector that is not part of the Filter API contract. Thomas Mueller, any
> idea/hint?
FilterImpl is just used in tests in oak-lucene and oak-solr-* (right?).
Possibly we could create a base class of FilterImpl that doesn't have a
selector, just for testing. That would need some work.
> Cyclic dependency between oak.spi.query an oak.query.*
> ------------------------------------------------------
>
> Key: OAK-6304
> URL: https://issues.apache.org/jira/browse/OAK-6304
> Project: Jackrabbit Oak
> Issue Type: Task
> Components: core, indexing, query
> Reporter: angela
> Attachments: OAK-6304.patch, OAK-6304-test.patch
>
>
> while working on OAK-6069 i noticed that there exist IMO troublesome
> dependencies from _o.a.j.oak.spi.query_ back to _o.a.j.oak.query.*_. While i
> don't know the historical reasons that led to this setup it feels wrong to me
> that the interfaces and classes in _o.a.j.oak.spi.query_ would have
> dependencies to packages that from my understanding are supposed to contain
> implementations.
> while utilities like {{o.a.j.oak.spi.query.Cursors}} may simply have ended up
> in the wrong bucket and could possibly moved to e.g _o.a.j.query.index_, i
> couldn't find an easy solution for an interface like
> {{o.a.j.oak.spi.query.Filter}} that defines methods returning
> {{o.a.j.query.fulltext.FullTextExpression}} and
> {{o.a.j.query.QueryEngineSettings}}, while at the same time being tied to the
> {{o.a.j.oak.spi.query.QueryIndex}} interface.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
