OpenSearch Notes

Implementation Notes

• added:  GET:site/search?format=opensearch to return an OpenSearch-compliant atom feed.
  • The PHP search engine (/extensions/LuceneSearch.php) uses the atom feed to display results.  The link dialog will continue to use format=search until we decided to update that code.
• added:  GET:site/search/description to return the OpenSearch description document (used to discover the Uri format of the search engine)
  
• added a preview field to the lucene index
  • this field holds the first 1024 characters of a page (without markup) or the first 1024 characters of an attachment if search filter was available to convert the attachment to plain text
• OpenSearch has a <totalResults> element.  Since we apply permissions to search results (and those permissions aren't stored in the lucene index) I had to hard-code the totalResults value to some arbitrarily high number (100,000)
  • A possible workaround for this is to actually store page permissions in the lucene index.  However, this would require us to re-index a document/attachment each time page permissions change.
    
• Paging:  The paging is done via offset/limit.  If no atom:entry elements appear in the feed it's assumed we've reached the end of the results.  We might also want to consider implementing
  
• Output format: OpenSearch results can be returned in RSS 2.0, Atom 1.0, and html format.  I've only implemented and atom feed at this time.
• Extension elements: The atom feed returned by Deki Wiki's OpenSearch adds some additonal information to search results (which the client (PHP) uses to provide additional context. 
  
  • The following namespaces is used to denote Deki Wiki extension elements:  xmlns:dekilucene="http://services.mindtouch.com/deki/d...06/luceneindex"
  • <dekilucene:path>Deki_Wiki/Installation_and_Upgrade/Installation_FAQ/Deki_Wiki_Dependency_Lists</dekilucene:path>
    • For pages, this is the path within the heirarchy
    • for attachments, this is the path to the page where it's attached
  • <dekilucene:size>4443</dekilucene:size> - the number of characters in the page or the size (in bytes) of an attachment
  • <dekilucene:wordcount>367</dekilucene:wordcount> - The number of words in a page.  This value is always 0 for attachments
  • <dekilucene:id.file>1831</dekilucene:id.file> - the id of the file (this element doesn't exist for page entries)
  • <dekilucene:page.parent id="4037" path="Deki_Wiki/Installation_and_Upgrade/Installation_FAQ" title="Installation FAQ" href="http://dekidev/@api/deki/pages/4037"/> - This is provided so the client (PHP) can display a link to the parent page without having to hit the API again
    • For files, this is the page it's attached to