Victor Chenal created SOLR-14215:
------------------------------------
Summary: Improve performance of SolrFeature when there are no
filter queries
Key: SOLR-14215
URL: https://issues.apache.org/jira/browse/SOLR-14215
Project: Solr
Issue Type: Improvement
Security Level: Public (Default Security Level. Issues are Public)
Reporter: Victor Chenal
In cases where a SolrFeature only contains a main query {{q}} and no filter
queries {{fq}}, it is possible to traverse documents using the doc iterator
from {{q}} instead of constructing one from a processed filter that combines
all of the queries.
In our test datasets this made retrieving the feature *10 times faster*.
Specifically, the call to
[{{getProcessedFilter}}|https://github.com/apache/lucene-solr/blob/16e537db7efe37eb7519b07dc96c5026c0a1d193/solr/contrib/ltr/src/java/org/apache/solr/ltr/feature/SolrFeature.java#L242-L243]
inside {{getDocIdSetIteratorFromQueries}} is expensive. That method aims at
returning a doc iterator that traverses over the documents that match all of
the criteria of {{q}} and {{fq}}.
When there are no filter queries (i.e. {{fq}} is null or empty), this just
returns a doc iterator for {{q}}. Using the iterator from that query directly
is a lot faster.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
