[
https://issues.apache.org/jira/browse/SOLR-17190?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17822641#comment-17822641
]
Michael Gibney commented on SOLR-17190:
---------------------------------------
Another improvement: the size arg to {{org.apache.solr.util.LongSet}} ctor,
like that of {{HashMap}}, configures the size of the backing element array; but
as invoked in code it is used with the intention of sizing to accommodate a
specific number of expected elements. This is fixed by replacing with hppc
LongHashSet, whose ctor takes load factor into account and sizes the backing
array appropriately.
> Replace org.apache.solr.util.LongSet with hppc LongHashSet
> ----------------------------------------------------------
>
> Key: SOLR-17190
> URL: https://issues.apache.org/jira/browse/SOLR-17190
> Project: Solr
> Issue Type: Improvement
> Security Level: Public(Default Security Level. Issues are Public)
> Affects Versions: main (10.0), 9.5.0
> Reporter: Michael Gibney
> Assignee: Michael Gibney
> Priority: Minor
>
> The functionality provided by {{org.apache.solr.util.LongSet}} is redundant
> to that provided by {{com.carrotsearch.hppc.LongHashSet}}. Since hppc is
> already a dependency of solr-core, we should replace usages of
> {{org.apache.solr.util.LongSet}} and remove/deprecate it.
> The motivation for this is practical as well: the [hash function used by
> org.apache.solr.util.LongSet|https://github.com/apache/solr/blob/148abec1a37b89f9bbd1835c239746773b374125/solr/core/src/java/org/apache/solr/util/LongSet.java#L78]
> is quite suboptimal for certain common usage patterns (specifically, if the
> values stored vary by powers of two (the larger the worse), values end up
> disproportionately hashing to the same bucket and you basically have an array
> scan for get and put).
> This issue proposes to replace usages on {{main}} and {{branch_9x}}, remove
> {{org.apache.solr.util.LongSet}} on main, and either remove it or deprecate
> it on {{branch_9x}}.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@solr.apache.org
For additional commands, e-mail: issues-h...@solr.apache.org
