Unify Enterprise Search Remote indexing template

From VYRE

The remote indexing template dictates what information from a content store is stored and indexed, and how.

Creating

To create a remote indexing template (RIT) for a content store go to the store (see Datastore: Content and Metadata) and click 'Edit...' and 'Remote Indexing Template'. Remember to make sure the option 'Use remote index for this store' is checked.

To submit all items for that store to the remote index click 'Edit...' and 'Send all items to remote index'.

Template Code

The template is created to represent the content store item as XML using FreeMarker.

Here is a sample template that stores two attributes from a content store:

  <SearchableItem>
    <uuid>${item.id}</uuid>
    <documentBoost>1.0</documentBoost>
    <properties>
      <entry>
        <string>title</string>
        <SearchableProperty>
          <name>title</name>
          <value><![CDATA[${item.name}]]></value>
          <boost>1.0</boost>
          <stored>true</stored>
          <indexed>true</indexed>
          <tokenized>true</tokenized>
        </SearchableProperty>
      </entry>
      <entry>
        <string>description</string>
        <SearchableProperty>
          <name>description</name>
          <value><![CDATA[${item.description}]]></value>
          <boost>1.0</boost>
          <stored>true</stored>
          <indexed>true</indexed>
          <tokenized>true</tokenized>
        </SearchableProperty>
      </entry>
    </properties>
  </SearchableItem>

As you can see the template represents the overall structure of the item as XML with content store attributes inserted or calculated using FreeMarker expressions or interpolation.

Configuration

SearchableItem

This is the item indexed by the search engine and it mandates three child nodes: uuid, documentBoost and properties which is an iterable of SearchableProperties describing each attribute.

uuid

A unique identifier for this document.

documentBoost

A float value representing the overall score multiplier given to this document at index time. As an example would a "documentBoost" value of 1.5 mean that this document would be given an inflated score that's 50% higher than its natural rank. This allows you to specify that documents from one datastore are to be considered more relevant than others.

properties

A map of searchable properties that represent each stored value.

SearchableProperty

A searchable property encapsulated how the value of a single attribute from a content store item and mandates how the search server should treat it at index time. Note that this is an XML serialization of the Java object which is why (at present) there is a redundant 'string' property present in the XML representation. TODO: investigate whether name could be used to represent a 'display value' rather than act purely as an identifier.

name

The identifier or name of the indexed attribute, e.g. 'title'

value

The value of the indexed attribute.

boost

A float value representing the overall score multiplier given to keyword hits in this attribute at query time. This means that a hit in a given attribute can be given higher or less weighting than others. As an example could a hit in a 'title' field be considered more valuable to natural rank than a hit in a 'description' or 'body' field.

stored

'true' or 'false' indicating whether the value should be stored verbatim in the index for displaying in search results.

indexed

'true' or 'false' indicating whether the document should be indexed i.e. whether it is searchable or not (i.e. not just for display purposes).

tokenized

'true' or 'false' indicating whether or not to tokenize the value of the property. The tokenization process splits the value according to predefined rules meaning that individual tokens or words will be searchable. In effect this indicates whether the individual constituent parts of the value are searchable or whether the attribute as a whole represents a single value.

Example:

If tokenized would the sentence "The quick brown fox jumps over the lazy dog" yield a result on the keyword 'fox'. However if non-tokenized would it only be considered a valid hit if a query was done for the entire string.

Advanced

TODO: Solr schema xml, Solr config files and rabida configuration files which override settings applied in the remote indexing template.

Helper Object

To help with retrieving values in the RIT, a helper object is added. Methods can be called on this object to retrieve different information.

For example ${helper.getAttributeValue( '18' )} will return the String value for attribute 18 of the current item (or the default value, see below).

Default fields indexed

The fields below are all stored automatically when sending an item to the remote index.

See Also

• Unify Enterprise Search form portlet
• Rabida search results portlet
• Set up a basic example of a rabida search page