I now have a working configuration of VuFind + Solr 3.1 – you can try it out by checking out this branch:
Some important notes:
- SolrMarc’s direct index writing feature does not seem to work with the new Solr release, so I had to reconfigure the import scripts to use HTTP posts instead of index writes. You may have to make some extra edits to import/import.properties to get everything working (i.e. if you’re not using the standard Solr port of 8080). I may change this back in the future if we find a way to get direct writing working, though HTTP has some advantages anyway, and it might not be such a bad thing if we keep this configuration.
- Many of the analyzer chains in the Solr schema are significantly changed because I have decided to try using the new built-in ICUFoldingFilterFactory for Unicode normalization instead of the custom filter we used to use. Unlike the previous custom component, ICUFoldingFilterFactory also handles lowercasing, and it also has its own companion tokenizer. The good news is that overall, this shortens and simplifies our analyzer chains; the bad news is that it’s different enough that there are likely to be some undesired side effects to the change. Please let me know if you have any known problematic search strings – it would be interesting to start throwing old problems against the new code.
- I have included the fixed version of Mark Triggs’ alphabetic browse tools, and at first glance it seems to be working (thanks, Mark!).
- There are several minor configuration adjustments to eliminate deprecated settings, and I had to change the way we process terms component responses since the format has been revised.
Please let me know if you have a chance to look at this – I’ll be very interested to hear feedback. Hopefully by the next developers call, we can discuss this in more detail and decide whether or not to push it to the trunk.