From: Demian K. <dem...@vi...> - 2009-09-18 18:09:11
|
I just plugged in your patch, deleted my index and indexed some new records with import-marc.sh, and spell check immediately started working. I didn't have to do the manual call to spellcheck.build before suggestions started appearing. Does this mean there is no problem in the current trunk, or am I just missing something -- i.e. will it work without the indexing but work better with it? Will subsequent imports cause problems? If it is indeed necessary to poll that URL periodically, it's more reason to improve our documentation on nightly update processes and provide some sample scripts. As far as doing this under Windows goes, there are Windows versions of both wget and curl, so it could be achieved through the installation of one of those packages. I'm not aware of a good way without the help of a third-party tool, though that doesn't mean there isn't one. Anyway, I'm happy to commit these changes to the trunk if nobody objects -- please let me know. I'll make the commit if I haven't heard otherwise by early next week. thanks, Demian > -----Original Message----- > From: Till Kinstler [mailto:kin...@gm...] > Sent: Thursday, September 17, 2009 4:10 PM > To: vuf...@li... > Subject: [VuFind-Tech] Spellchecker > > I just played around a little to get the spellchecker working again in > the trunk. The trivial patch attached (for r1489) made it work after > reindexing data and manually creating the spellchecking index by > calling the URL http://[Solr-server]:[Solr- > Port]/solr/biblio/select/?q=allfields:buntebank&spellcheck=true&spellch > eck.build=true > (though doing an index optimization should (re-)create that index as > well). > Please test (you need to reindex data) and put into trunk if working > reliably. > Crucial point is the creation of the spellchecking index in Solr. It > took me some trial and error to get it created. Solr isn't telling much > about success or failure when calling the URL above. Finally it worked > with the settings in schema.xml and solrconfig.xml included in the > patch. > We should get creation of the spellchecking index somehow into the > usual indexing process. One way might be setting the buildOnCommit > option in solrconfig.xml, but that slows down commits (and I am not > sure, if it really works as expected...). Or simply add a call of the > URL above to import-marc.sh and .bat (how to call an URL in a .bat > file?)? Or is that a thing to put into solrmarc? There is already an > option to call a commit after indexing in solrmarc, so maybe add one to > create a spellchecking index as well? If that's the wanted solution, > I'll take that over to the solrmarc list and implement it there (but > not before October then). > > Till |