Thank you very much, was able to find the Jira that triggered the problematic for me code change. Just wonder how to proceed … My company is now in the middle of a process to upgrade from Solr 5.4 to Solr 8.7 . Performance tests showed that performance got worse … An investigation (jstack, debug etc.) showed, that the cause is the call to the method: https://github.com/apache/solr/blame/1cb0850b70a7583501718ed635964f2f605d1742/solr/core/src/java/org/apache/solr/search/Grouping.java#L586 3 years ago 773 if (secondPass != null) { 774 result = secondPass.getTopGroups(0); 775 populateScoresIfNecessary(); 776 } This change was done as part of: https://github.com/apache/solr/commit/55bfadbce115a825a75686fe0bfe71406bc3ee44 What seems to happen is that since my system is asking for ‘score’ in fl (required for a certain business need). This code makes the document collection be re-triggered, which takes time (probably similar to the first time documents are collected). According to the Jira, this is done to support getting the score when sorting by a field (not by “score”). I did a small patch and removed this code, and indeed when sorting by ‘score’ all works as expected, and I get the score per document when I ask for it in fl, but if I sort by some other field, the score field returns with value: NaN. In means of business of my product, I need the score to return only when the search is sorted by score. I can live with this ugly patch, but probably the real solution is not to call the populateScoresIfNecessary method if the sort is by ‘score’ and not a by certain field, since it seems not necessary. I opened a bug in Jira for this, and will update it with this conclusion, but would like to be as sure as I can that they will fix it. Do know of a way to “push” Solr developers to fix a certain issue?
Again thank you for the advise. From: Eric Pugh <ep...@opensourceconnections.com> Sent: Wednesday, 18 August 2021 14:55 To: dev@solr.apache.org Subject: Re: Tracking out when and why a certain code was added in specific Solr class CAUTION: EXTERNAL Email. I just did this to try and understand when the “indent” checkbox in the Solr Admin changed behavior! I go into Github, and find the file, and then use the Blame link: https://github.com/apache/solr/blame/720fa009118c229e0969fa21523ff3c4cf91f931/solr/webapp/web/partials/query.html<https://urldefense.com/v3/__https:/github.com/apache/solr/blame/720fa009118c229e0969fa21523ff3c4cf91f931/solr/webapp/web/partials/query.html__;!!NkSChLMW7qAT!rCpJfBKMtOaY8Oe8ectnxnqdN7aIRKeRkIDir44Xd4j2l9ODumezRN9sBAqDL9dLqI1pwu0F8A$> That will link you to the JIRA issue, in my case it was SOLR-10494 that changed things. You can also do the same thing off of various branches. On Aug 18, 2021, at 7:29 AM, yael.mushin...@exlibrisgroup.com.INVALID<mailto:yael.mushin...@exlibrisgroup.com.INVALID> wrote: Hello, Can someone assist in explaining how I can understand regarding a certain code change I see between versions of Solr, when and part of what commit it was added/changed? Thank you, _______________________ Eric Pugh | Founder & CEO | OpenSource Connections, LLC | 434.466.1467 | http://www.opensourceconnections.com<https://urldefense.com/v3/__http:/www.opensourceconnections.com/__;!!NkSChLMW7qAT!rCpJfBKMtOaY8Oe8ectnxnqdN7aIRKeRkIDir44Xd4j2l9ODumezRN9sBAqDL9dLqI2DuzbVtA$> | My Free/Busy<https://urldefense.com/v3/__http:/tinyurl.com/eric-cal__;!!NkSChLMW7qAT!rCpJfBKMtOaY8Oe8ectnxnqdN7aIRKeRkIDir44Xd4j2l9ODumezRN9sBAqDL9dLqI3B_1-8tA$> Co-Author: Apache Solr Enterprise Search Server, 3rd Ed<https://urldefense.com/v3/__https:/www.packtpub.com/big-data-and-business-intelligence/apache-solr-enterprise-search-server-third-edition-raw__;!!NkSChLMW7qAT!rCpJfBKMtOaY8Oe8ectnxnqdN7aIRKeRkIDir44Xd4j2l9ODumezRN9sBAqDL9dLqI0SZK8mtA$> This e-mail and all contents, including attachments, is considered to be Company Confidential unless explicitly stated otherwise, regardless of whether attachments are marked as such.
