From: Demian K. <dem...@vi...> - 2013-04-10 17:46:49
|
It sounds like the problem here is that text at indexing time is getting normalized differently from text at search time... e.g. if your searchspecs.yaml uppercases user search terms, then your indexing code should do the same thing to ensure that everything lines up. This sort of work is normally handled internally to Solr using the analyzer chain mechanism, but because of the somewhat unusual nature of the call number search (with its wildcarding, left anchored nature, etc.), it's not currently using the standard Solr features. Perhaps this can be revisited in the not-too-distant future when we upgrade to a more robust Solr version... but for now you might want to make sure your indexing and searching are more closely aligned, and then things may start working better. - Demian From: Nathan Tallman [mailto:nta...@gm...] Sent: Wednesday, April 10, 2013 11:28 AM To: Demian Katz Cc: vufind-tech Subject: Re: [VuFind-Tech] Call Number Searching I think it's a combination of two things. A call number search searches callnumber_exact and callnumber_fuzzy (neither of which I can find in the Solr schema...). The values in callnumber_exact is "NEARPRINTGEOGRAPHY". So, a search for "NEARPRINTGEOGRAPH*<http://americanjewisharchives.org/catalog/Search/Results?lookfor=NEARPRINTGEOGRAPH*&type=CallNumber&submit=Find>" achieves my goal. (It doesn't work in lowercase.) I did find callnumber-a in the schema, whose value is "Nearprint Geography". I can also find my goal by searching the callnumber-a field specifically, while escaping the space, "callnumber-a:Nearprint\ Geograph*<http://americanjewisharchives.org/catalog/Search/Results?lookfor=callnumber-a%3ANearprint%5C+Geograph*&type=CallNumber&submit=Find>". So, it's both a field value and space issue. I was trying to add callnumber-a to the callnumber configuration in /web/conf/searchspecs.yaml to no avail, however I did I think of something else. From looking at the custom munge, I determined that it shouldn't have to use an asterisk, as one is already inserted, after the fact. So... a call number search of "Nearprint Geograph<http://americanjewisharchives.org/catalog/Search/Results?lookfor=Nearprint+Geograph&type=CallNumber&submit=Find>" achieves my goal. Guess, I'll just have to leave off the asterisks when searching call numbers, which we frequently do. Thanks, Nathan On Wed, Apr 10, 2013 at 8:43 AM, Demian Katz <dem...@vi...<mailto:dem...@vi...>> wrote: Have you turned on debug, looked at the generated Solr queries, and tried adding debugQuery=true to figure out what is going on at the Solr side? I'm suspicious that the space in "Nearprint Geograph" might be part of the problem, but it will take a little more analysis to figure out exactly what is going on. Let me know if you need more details or turn up any new clues! - Demian From: Nathan Tallman [mailto:nta...@gm...<mailto:nta...@gm...>] Sent: Tuesday, April 09, 2013 4:49 PM To: vufind-tech Subject: [VuFind-Tech] Call Number Searching I'm having some issues using wildcards to search call numbers. Sometimes they work, sometimes they don't. For example, searching "MS-*<http://americanjewisharchives.org/catalog/Search/Results?lookfor=MS-*&type=CallNumber&submit=Find>" returns all manuscript collections. But, searching "Nearprint Geograph*<http://10.1.2.118/vufind/Search/Results?lookfor=Nearprint+Geograph*&type=CallNumber&submit=Find>" returns no results, even though we have 1800+ records with "Nearprint Geography" as a call number. I'm not sure why the wild card is working sometimes and not others. Thanks, Nathan |