[
https://issues.apache.org/jira/browse/SOLR-15437?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Tobias Kässmann updated SOLR-15437:
-----------------------------------
Description:
[~TomGilke], [~tboeghk] and I are currently working a lot on the Solr LTR
feature. In our setup we're using a custom sort function for a solid base
scoring of the documents.
We found out that a plain SolrCloud will return random sorted docs if you
define a custom sort field or function. This problem is also mentioned
[here|https://lucene.472066.n3.nabble.com/Ranking-issue-when-combining-sorting-and-re-ranking-on-SolrCloud-multiple-shards-td4457723.html]
and is not really fixed.
You can do the following steps to reproduce this issue:
[here|https://gist.github.com/tboeghk/0844d310641b994dac924acaf94fd78f].
In the meantime we're also investing on this issue and try to debug the
QueryComponent. But if there is any hint or thought available, we're very happy
to hear from you. Maybe [~cpoerschke] can help here a little bit?
What we've found out so far:
* It only happens in SolrCloud mode (shards > 1)
* Maybe is has something to do with the
[MergeStrategy|https://github.com/apache/solr/blob/main/solr/core/src/java/org/apache/solr/handler/component/QueryComponent.java#L188]
in the QueryComponent.
The following test will help to reproduce this issue (place it into the
TestLTROnSolrCloud file):
{code:java}
@Test
public void testSimpleQueryCustomSort() throws Exception {
SolrQuery query = new SolrQuery("*:*");
query.setRequestHandler("/query");
query.setFields("*,score,[shard]");
query.setParam("rows", "8");
query.setParam("sort", "id asc");
query.add("rq", "{!ltr model=powpularityS-model reRankDocs=8}");
QueryResponse queryResponse =
solrCluster.getSolrClient().query(COLLECTION, query);
assertEquals(8, queryResponse.getResults().getNumFound());
assertEquals("8", queryResponse.getResults().get(0).get("id").toString());
assertEquals("7", queryResponse.getResults().get(1).get("id").toString());
assertEquals("6", queryResponse.getResults().get(2).get("id").toString());
assertEquals("5", queryResponse.getResults().get(3).get("id").toString());
assertEquals("4", queryResponse.getResults().get(4).get("id").toString());
assertEquals("3", queryResponse.getResults().get(5).get("id").toString());
assertEquals("2", queryResponse.getResults().get(6).get("id").toString());
assertEquals("1", queryResponse.getResults().get(7).get("id").toString());
}
{code}
was:
[~TomGilke], [~tboeghk] and I are currently working a lot on the Solr LTR
feature. In our setup we're using a custom sort function for a solid base
scoring of the documents.
We found out that a plain SolrCloud will return random sorted docs if you
define a custom sort field or function. This problem is also mentioned
[here|https://lucene.472066.n3.nabble.com/Ranking-issue-when-combining-sorting-and-re-ranking-on-SolrCloud-multiple-shards-td4457723.html]
and is not really fixed.
You can do the following steps to reproduce this issue:
[here|https://gist.github.com/tboeghk/0844d310641b994dac924acaf94fd78f].
In the meantime we're also investing on this issue and try to debug the
QueryComponent. But if there is any hint or thought available, we're very happy
to hear from you. Maybe [~cpoerschke] can help here a little bit?
What we've found out so far:
* It only happens in SolrCloud mode (shards > 1)
* Maybe is has something to do with the
[MergeStrategy|https://github.com/apache/solr/blob/main/solr/core/src/java/org/apache/solr/handler/component/QueryComponent.java#L188]
in the QueryComponent.
The following test will help to reproduce this issue:
{code:java}
@Test
public void testSimpleQueryCustomSort() throws Exception {
SolrQuery query = new SolrQuery("*:*");
query.setRequestHandler("/query");
query.setFields("*,score,[shard]");
query.setParam("rows", "8");
query.setParam("sort", "id asc");
query.add("rq", "{!ltr model=powpularityS-model reRankDocs=8}");
QueryResponse queryResponse =
solrCluster.getSolrClient().query(COLLECTION, query);
assertEquals(8, queryResponse.getResults().getNumFound());
assertEquals("8", queryResponse.getResults().get(0).get("id").toString());
assertEquals("7", queryResponse.getResults().get(1).get("id").toString());
assertEquals("6", queryResponse.getResults().get(2).get("id").toString());
assertEquals("5", queryResponse.getResults().get(3).get("id").toString());
assertEquals("4", queryResponse.getResults().get(4).get("id").toString());
assertEquals("3", queryResponse.getResults().get(5).get("id").toString());
assertEquals("2", queryResponse.getResults().get(6).get("id").toString());
assertEquals("1", queryResponse.getResults().get(7).get("id").toString());
}
{code}
> ReRanking does work in combination with custom sort and SolrCloud
> -----------------------------------------------------------------
>
> Key: SOLR-15437
> URL: https://issues.apache.org/jira/browse/SOLR-15437
> Project: Solr
> Issue Type: Bug
> Security Level: Public(Default Security Level. Issues are Public)
> Components: contrib - LTR, query parsers, SolrCloud
> Affects Versions: 8.8.2
> Reporter: Tobias Kässmann
> Priority: Major
>
> [~TomGilke], [~tboeghk] and I are currently working a lot on the Solr LTR
> feature. In our setup we're using a custom sort function for a solid base
> scoring of the documents.
> We found out that a plain SolrCloud will return random sorted docs if you
> define a custom sort field or function. This problem is also mentioned
> [here|https://lucene.472066.n3.nabble.com/Ranking-issue-when-combining-sorting-and-re-ranking-on-SolrCloud-multiple-shards-td4457723.html]
> and is not really fixed.
> You can do the following steps to reproduce this issue:
> [here|https://gist.github.com/tboeghk/0844d310641b994dac924acaf94fd78f].
> In the meantime we're also investing on this issue and try to debug the
> QueryComponent. But if there is any hint or thought available, we're very
> happy to hear from you. Maybe [~cpoerschke] can help here a little bit?
> What we've found out so far:
> * It only happens in SolrCloud mode (shards > 1)
> * Maybe is has something to do with the
> [MergeStrategy|https://github.com/apache/solr/blob/main/solr/core/src/java/org/apache/solr/handler/component/QueryComponent.java#L188]
> in the QueryComponent.
> The following test will help to reproduce this issue (place it into the
> TestLTROnSolrCloud file):
> {code:java}
> @Test
> public void testSimpleQueryCustomSort() throws Exception {
> SolrQuery query = new SolrQuery("*:*");
> query.setRequestHandler("/query");
> query.setFields("*,score,[shard]");
> query.setParam("rows", "8");
> query.setParam("sort", "id asc");
> query.add("rq", "{!ltr model=powpularityS-model reRankDocs=8}");
> QueryResponse queryResponse =
> solrCluster.getSolrClient().query(COLLECTION, query);
> assertEquals(8, queryResponse.getResults().getNumFound());
> assertEquals("8", queryResponse.getResults().get(0).get("id").toString());
> assertEquals("7", queryResponse.getResults().get(1).get("id").toString());
> assertEquals("6", queryResponse.getResults().get(2).get("id").toString());
> assertEquals("5", queryResponse.getResults().get(3).get("id").toString());
> assertEquals("4", queryResponse.getResults().get(4).get("id").toString());
> assertEquals("3", queryResponse.getResults().get(5).get("id").toString());
> assertEquals("2", queryResponse.getResults().get(6).get("id").toString());
> assertEquals("1", queryResponse.getResults().get(7).get("id").toString());
> }
> {code}
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@solr.apache.org
For additional commands, e-mail: issues-h...@solr.apache.org
