Query

Component description

The MKSearch query component is responsible for handling HTTP-based search queries by general users. It's implementation is currently in the form of a single HttpQuery servlet class that forwards queries to custom JSP result pages that query and display the results with a custom ResultsRenderer type.

Completed task information has been moved to the beta 1 query plans archive.

Outstanding issues

Extended graph queries

Presently, the query builder does not handle extended RDF graphs, where a Dublin Core qualified term is used to refine a primary search field. For example, where an indexed document has a DC.Date property with a DCTERMS.W3CDTF encoding scheme, simple date queries are not matched. If no W3C date node were present, the match is successful.

Phrase search result navigation

Where quoted phrases are used in queries, page navigation is corrupted. Symptoms as follows:

• The quotes are not carried through to the navigation link.
• The number of results pages is (or may be) increased, compared with the original results page.
• Every result page in the series has the same set of results.
• The results numbering on the results pages starts with a negative value, increments through zero to positive numbers.
• The result items with negative numbering have no title or summary and link to the standard 404 error page.

Beta 2 development plans

Results cacheing

Standing RSS feeds and results paging will also require a result cache to optimise performance on the server side. However, the cache mechanism also needs to be integrated with the repository management interface, so that stale results are purged.

AND and AND NOT queries

The query builder should interpret the + and - symbols immediately before a word to mean AND and AND NOT respectively.