Setting Solr's default search operator to AND for the Discovery index leads to
search behaviour that's closer to what our users expect -- "do what Google
does". There is a configuration option for the search operator in dspace.cfg.
However, that setting is used only by the legacy DSQuery class for the now
deprecated Lucene-based search.
Contribution from CILEA funded by the Hub project from HKU
(http://hub.hku.hk)
Faceting, filtering (autocomplete), sidebar facet for the site home page,
community and collections are all implemented.
Changes to the Discovery API/configuration:
1) changed the unique field for the SOLR document, now is used the
concatenation of ID and TYPE-ID (in future we want to index also
object that have not an handle)
2) the prune query has been changed in search.resourcetype:[2 TO 4] so to not remove eventually extra data loaded in the SOLR search core
3) added defaultRpp parameter
Main differences from the XMLUI implementation:
1) facets doesn't have a "...More" link but there are pagination to
scroll facet in the context (search, home page, community, etc.)
2) facets doesn't show the values already selected
3) autocomplete is done against user input and does not dump all the
values (this was a performance issue in XMLUI < 3.0, with 90k items
I see JSON around 2Mb). With the new Discovery improvements the
autocomplete feature in XMLUI seems to be turned off
4) to enable JSPUI discovery you need to add some extra plugins in
dspace.cfg (instructions are provided as comment in the discovery.cfg)
This contribution is supported by the University of Hong Kong that has
adopted the first version of this code since July 2010
Software development by the CILEA SURplus Team (http://www.cilea.it)
This is a full featured provider for the DSpace Browse System that read
data from the Discovery Search SOLR core.
The browse supports:
- all the actual browse system configuration options (item and metadata
browse, pagination, sorting)
- browse of withdrawn items
- browse of private items (XMLUI only)
- search for authors used in the item mapper tool
Major changes:
- add two new search methods to the SearchService that allow you to
specify if you want or not withdrawn items in your result. The old
methods just call these new methods asking for no withdrawn item (this
assure backcompatiliby)
- the SearchServiceImpl now index also withdrawn item, new SolrDoc field
"withdrawn"
- DiscoverResult.Facet now keep information also about authority so
that is possible distinct facets related to homonyms authors, etc.
(this feature can be disabled in the discovery.cfg)
- changed the schema.xml the default should be to ignore field if not
otherwise specified (see also below)
- better support for projection, metadata are stored (not indexed) with
all related informations attached (authority, language). Metadata to
store for projection can be defined in the discovery.cfg
Other notes:
- SolrBrowseDAO/SolrCreateBrowseDAO use settings defined in the main
dspace.cfg (the browse section) this mean that sorting configuration is
not related to the discovery search sorting etc.
- Browse consumer should be disabled (if not only overhead but it still
works), the actual index work is done using the Discovery "plugin"
SolrServiceIndexPlugin (configurated via spring)
- SolrCreateBrowseDAO is able to remove the old tables created by the
DBMS DAO implementations simply using the browse command line script
(dspace index -d -f)
- to build the browse index it is required to run the discovery index
client (dspace update-discovery-index -b)
* fixes typos almost exclusively in comments and strings
* fixes whitespace where combination of tabs and spaces may break indentation
* doesn't touch typos in config property names to prevent breakage
