[
https://issues.apache.org/jira/browse/LUCENE-9142?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17018103#comment-17018103
]
Michael McCandless commented on LUCENE-9142:
--------------------------------------------
This is indeed dangerously sneaky code – {{SortedIntSet.equals}} has special
logic to compare only to a {{FrozenIntSet}} ... it's kinda weird that it cannot
compare against another {{SortedIntSet}}, while {{FrozenIntSet.equals}} is
symmetric (can compare against either class).
Maybe we could at least fix both of these {{equals}} methods to invoke the same
(static) {{equals}}?
Hmm, and {{FrozenIntSet.equals}} looks buggy when it's comparing to a
{{SortedIntSet}} – it's checking the length of the {{values}} array in the
{{SortedIntSet}} when I think it should instead check against {{upto}}? If
that's really it bug it may indeed be causing our determinize/minimize to not
collapse as many states as it should?
> Add documentation to Operations.determinize, SortedIntSet, and FrozenSet
> ------------------------------------------------------------------------
>
> Key: LUCENE-9142
> URL: https://issues.apache.org/jira/browse/LUCENE-9142
> Project: Lucene - Core
> Issue Type: Bug
> Components: core/search
> Reporter: Mike Drob
> Priority: Major
>
> Was tracing through the fuzzy query code, and IntelliJ helpfully pointed out
> that we have mismatched types when trying to reuse states, and so we may be
> creating more states than we need to.
> Relevant snippets:
> {code:title=Operations.java}
> Map<SortedIntSet.FrozenIntSet,Integer> newstate = new HashMap<>();
> final SortedIntSet statesSet = new SortedIntSet(5);
> Integer q = newstate.get(statesSet);
> {code}
> {{q}} is always going to be null in this path because there are no
> SortedIntSet keys in the map.
> There are also very little javadoc on SortedIntSet, so I'm having trouble
> following the precise relationship between all the pieces here.
> cc: [~mikemccand] [~romseygeek] - I would appreciate any pointers if you have
> them
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org
For additional commands, e-mail: issues-h...@lucene.apache.org
