Hi Luke, Thank you for analyzing the code and kindly providing a detailed explanation.
I now understand that it is not possible to retrieve the partial result set or its highlighted snippets for a query interrupted due to exceeding the "timeAllowed" setting value. I truly appreciate your help once again. Regards, Zhao 2024年11月15日(金) 21:46 Luke Kot-Zaniewski (BLOOMBERG/ 919 3RD A) < lkotzanie...@bloomberg.net>: > From reading the code I don't think it is possible. If your timeout is > occurring in the core search the SolrIndexSearcher will throw a certain > type of ExitingReaderException that is caught very high up (in the > SearchHandler). This happens as the SearchHandler is looping through > "SearchComponents" one of which is the HighlightComponent. Effectively this > means the search processing is short-circuited as no downstream components > get invoked. Not sure if changing this behavior is the right thing since > highlighting itself can be expensive so the timeout would be misleading. I > suppose you could consider implementing more granular timeouts but not sure > about the return on investment of something like that. > > Sent from Bloomberg Professional for Android > > ----- Original Message ----- > From: Mingchun Zhao <users@solr.apache.org> > To: users@solr.apache.org > At: 11/13/24 20:38:49 UTC-05:00 > > > Hi Luke, > > Thank you very much for the valuable information. I’ll give the Unified > Highlighter’s Postings feature a try. > > That said, I’d still like to know more about the behavior of the > "timeAllowed" parameter, especially whether it’s possible to retrieve > partial search results (both matching documents and highlighted snippets) > when a query times out. > > If you have any information on this, I’d appreciate it. > > Regards, > Zhao > > 2024年11月13日(水) 21:50 Luke Kot-Zaniewski (BLOOMBERG/ 919 3RD A) < > lkotzanie...@bloomberg.net>: > > > Just throwing this out there, have you tried the unifiedHighlighter with > > termVectors/character offsets? The default configuration re-analyzes the > > text of the result-set which can be cpu-bound for large docs/fields. But > > there are ways to get around this by storing the character offsets and > > termVectors in the index, which can be a good tradeoff in many > situations, > > i.e. "Postings" > > > > > https://solr.apache.org/guide/solr/latest/query-guide/highlighting.html#:~:text=over%20highlighting%20performance.-,Postings,-%3A%20Supported%20by%20the > > > > I know this doesn't answer your question but just wanted to mention an > > alternative solution to relying on time-outs/partial-results, which may > be > > frustrating for other reasons. > > Sent from Bloomberg Professional for Android > > > > ----- Original Message ----- > > From: Mingchun Zhao <users@solr.apache.org> > > To: users@solr.apache.org > > At: 11/12/24 21:55:36 UTC-05:00 > > > > > > Hi, > > > > If a query is interrupted due to timeAllowed, does it still return the > > partially matched documents and highlight snippets on it? Or does the > > behavior vary depending on the Solr version? > > In my testing, when the response header includes "partialResults=true", > the > > partial search results or snippets were not included in the response > body. > > > > Regards, > > Zhao > > >
