The full array of Lucene analysis tools can now be applied to Lux’s XML text fields simply by configuring the fields in the Solr schema (or in the Java API by supplying a Lucene Analyzer to a Lux FieldDefinition).
Lux’s API now provides control over element visibility as a tradeoff between index size and power.
SolrCloud support! Lux now handles distributed, sharded indexes by using Solr’s distributed query and update functionality. The short story is that everything works when you have your documents spread across multiple machines, and this paves the way to scale out to very large indexes. All the xquery functions are now cloud-aware as well: lux:insert() will distribute documents using the default Solr hashing strategy based on the document URI as the unique key. Please be aware the implementation is complete, but has seen only limited testing, so there will be bugs. However, the bulk of the complexity is handled in Solr, not in Lux.
Solr schema types are now supported for XPath fields. Previously, if you declared an XPath field (in Lux’s configuration in solrconfig.xml), and its name matched a type in the Solr schema (in schema.xml), that fact was ignored, and its value was always treated as a string. Now we recognize the association and use Solr’s conversion rules to index the appropriately-typed value. Thanks to Mark Lawson for pointing this out.
Analyzers declared in solr schema for type of field named lux_text (or whatever the xml text field is called) are used when indexing and querying xml text field, and element and attribute text fields.
We’ve simplified and reorganized the index configuration representation of fields and field naming: created field roles, eliminated static singleton field definitions, tracked field renaming in the field object rather than in a map in the configuration object.
lux:field-values() was deprecated (in favor of lux:key) and is now eliminated
Element transparency: the contents of transparent elements are also indexed as part of enclosing elements, while the contents of opaque elements are not. Further, elements may be declared (by name) to be container elements, which “see into” descendant opaque elements, indexing all descendant content, with the exception of hidden elements, whose content is not indexed at all. Elements are opaque by default - in previous releases they were effectively transparent - all text was tagged with all enclosing element names.
Fixed a bug (introduced w/0.11.0) that caused exceptions in the URI resolver when calling doc().
Upgraded to Saxon 9.5. This change should be largely invisible to users, but people who roll their own installations will also need to upgrade since we use API calls in Saxon 9.5 that did not exist in 9.4.
Upgraded to Solr 4.4.0. Again, the changes are largely invisible, but installations must upgrade due to backwards incompatible changes. The index format did not change, though.
Experimental SolrCloud support.