This documentation relates to an old version of DSpace, version 5.x. Looking for another version? See all documentation.
Please note, that as of DSpace 4.0, the Solr-based Discovery search is on by the default in both JSPUI and XMLUI. This page describes the older Lucene-based search and DBMS browse indices. Neither the DBMS browse tables nor the Lucene search indices are used anymore (unless you explicitly disable SolrBrowseDAO and enable search artifacts). This page was previously called ReIndexing Content with the old legacy providers (DBMS for Browse or Lucene for Search)
DSpace offers two options to index content for Browsing & Searching:
- Faceted/Filtered Search & Browse (via Solr & DSpace Discovery) - enabled by default since DSpace 4.0
- Traditional Browse & Search (via Lucene & Database tables) - this is disabled by default, it was used in older versions of dspace and is being phased out
This particular page only describes the "Traditional Browse & Search" indexing processes. For more information on Faceted/Filtered Browse & Search, please see DSpace Discovery, in particular .
Re-Enabling the legacy Lucene Search and/or DBMS Browse providers
The old Lucene/DB search and browse will still work, but will need to be configured to work. The three options are SOLR (default), or Postgres/Lucene, or OracleDB/Lucene. When you set the value to these configs, ensure that at most only one provider is chosen.
If a DAOs configuration is not provided the system will default to using the SOLR Browse Engine
Configure search and browse to use PostgreSQL
This option enables the browse engine to store its indexes in PostgreSQL database tables. All browsing is then performed via queries to those database tables. This is the traditional browsing option for users of PostgreSQL.
Alter dspace.cfg to set the browseDAO to postgres.
Alter dspace.cfg to have ItemCount use Postgres
Configure search and browse to use Oracle
This option enables the browse engine to store its indexes in Oracle database tables. All browsing is then performed via queries to those database tables. This is the traditional browsing option for users of Oracle.
Alter dspace.cfg to set the browseDAO to Oracle
Alter dspace.cfg to have ItemCount use Oracle
Creating the Browse & Search Indexes
It is possible that the database tables to hold the search and browse data do not exist, so they must be created:
To create (or recreate) all the various browse/search indexes that you define as described in this page there are a variety of options available to you. You can see these options below in the command table.
Arguments short and long forms):
Should we rebuild all the indexes, which removes old tables and creates new ones. For use with
Execute all the remove and create SQL against the database. For use with
Actually do the indexing. Mutually exclusive with
Write the remove and create SQL to the stdout. For use with
Create the tables only, do no attempt to index. Mutually exclusive with
Make the tables, and do the indexing. This forces
Print extra information to the stdout. If used in conjunction with
Delete all the indexes, but do not create new ones. For use with
Show this help documentation. Overrides all other arguments.
If you are using the Solr Browse DAOs, that is the default since DSpace 4.0, it is not required to run this script as the data are stored in the Solr search core that need to be recreated using the Discovery maintenance script
Running the Indexing Programs
Complete Index Regeneration
Requires that you stop Tomcat first
Because this command actually deletes existing Browse Index tables, you must stop Tomcat (or your Servlet Container of choice) before executing
index-lucene-init. After the indexing command completes, you can restart Tomcat.
Known Oracle Issues
In many Oracle based DSpace installations, index-lucene-init often malfunctions because of Oracle specific permissions. It is therefore advised to stick to index-lucene-update instead
[dspace]/bin/dspace index-lucene-init you will completely regenerate your indexes, tearing down all existing tables and reconstructing with the new configuration.
Updating the Indexes
[dspace]/bin/dspace index-lucene-update you will reindex your full browse & search indexes without modifying the DSpace table structure. (This should be your default approach if indexing, for example, via a cron job periodically). Because it does not "tear down" the existing tables, this command can be run while DSpace (and Tomcat or similar) is still running.
If you are using the Solr Browse DAOs, that is the default since DSpace 4.0, you don't need to run this script as the data are stored in the Solr search core. You need to recreate the indexes using the Discovery maintenance script
Destroy and Rebuild Browse Tables
This is really not recommended unless you know what you are doing.
You can destroy and rebuild the database, but do not do the indexing. Output the SQL to do this to the screen and a file, as well as executing it against the database, while being verbose.
At the CLI screen:
Browse Index Customization
DSpace provides robust browse indexing. It is possible to expand upon the default indexes delivered at the time of the installation. The System Administrator should review Browse Index Configuration to become familiar with the property keys and the definitions used therein before attempting heavy customizations.
Through customization is is possible to:
- Add new browse indexes besides the four that are delivered upon installation. Examples:
- Specific subject fields (Library of Congress Subject Headings). (It is possible to create a browse index based on a controlled vocabulary or thesaurus.)
- Other metadata schema fields
- Combine metadata fields into one browse
- Combine different metadata schemas in one browse
Examples of new browse indexes that are possible. (The system administrator is reminded to read the section on Browse Index Configuration )
- Add a Series Browse. You want to add a new browse using a previously unused metadata element.
webui.browse.index.6 = series:metadata:dc.relation.ispartofseries:text:single
- Note: the index # need to be adjusted to your browse stanza in the _dspace.cfg_ file. Also, you will need to update your Messages.properties file.
- Combine more than one metadata field into a browse.You may have other title fields used in your repository. You may only want one or two of them added, not all title fields. And/or you may want your series to file in there.
webui.browse.index.3 = title:metadata:dc.title,dc:title.uniform,dc:relation.ispartofseries:title:full
- Separate subject browse.You may want to have a separate subject browse limited to only one type of subject.
webui.browse.index.7 = lcsubject.metdata:dc.subject.lcsh.text:single
As one can see, the choices are limited only by your metadata schema, the metadata, and your imagination.
Because Browse Indexes are stored in database tables, remember to run
index-lucene-init after adding any new definitions in the
dspace.cfg to have the indexes created and the data indexed.
Since DSpace 4.0 the Solr DAOs implementation of the browse engine is used by default you don't need to run the script described in this page at least if you have re-enabled the legacy DBMS provider. Instead use the Discovery maintenance script. Browse indexing in Solr is done within the Search Indexing process.
Search Index Customization
Configuring Lucene Search Indexes
Search indexes can be configured and customized easily in the dspace.cfg file. This allows institutions to choose which DSpace metadata fields are indexed by Lucene.
Where to put the search index files
By setting higher values of search.max-clauses will enable prefix searches to work on larger repositories.
It is possible to create a 'delayed index flusher'. If a web application pushes multiple search requests (i.e. a barrage or sword deposits, or multiple quick edits in the user interface), then this will combine them into a single index update. You set the property key to the number of milliseconds to wait for an update. The example value will hold a Lucene update in a queue for up to 5 seconds. After 5 seconds all waiting updates will be written to the Lucene index.
Which Lucene Analyzer implementation to use. If this is omitted or commented out, the standard DSpace analyzer (designed for English) is used by default. This standard DSpace analyzer removes common stopwords, lowercases all words and performs stemming (removing common word endings, like "ing", "s", etc).
Instead of the standard DSpace Analyzer (DSAnalyzer), use an analyzer which doesn't "stem" words/terms. When using this analyzer, a search for "wellness" will always return items matching "wellness" and not "well". However, similarly a search for "experiments" will only return objects matching "experiments" and not "experiment" or "experimenting". When using this analyzer, you may still use WildCard searches like "experiment*" to match the beginning of words.
Instead of the standard English analyzer, the Chinese analyzer is used.
Boolean search operator to use. The currently supported values are OR and AND. If this configuration item is missing or commented out, OR is used. AND requires all the search terms to be present. OR requires one or more search terms to be present.
This is the maximum number of terms indexed for a single field in Lucene. The default is 10,000 words‚ often not enough for full-text indexing. If you change this, you will need to re-index for the change to take effect on previously added items. -1 = unlimited (Integer.MAG_VALUE)
This property determines which of the metadata fields are being indexed for search. As an example, if you do not include the title field here, searching for a word in the title will not be matched with the titles of your items..
For example, the following entries appear in the default DSpace installation:
search.index.1 = author:dc.contributor.*
search.index.2 = author:dc.creator.*
search.index.3 = title:dc.title.*
search.index.4 = keyword:dc.subject.*
search.index.5 = abstract:dc.description.abstract
search.index.6 = author:dc.description.statementofresponsibility
search.index.7 = series:dc.relation.ispartofseries
search.index.8 = abstract:dc.description.tableofcontents
search.index.9 = mime:dc.format.mimetype
search.index.10 = sponsor:dc.description.sponsorship
search.index.11 = id:dc.identifier.*
search.index.12 = language:dc.language.iso
The format of each entry is
search.index.<id> = <search index name> : <schema> . <metadata field>[:index type] where:
is an incremental number to distinguish each search index entry
is the identifier for the search field this index will correspond to
is the schema used. Dublin Core (DC) is the default. Others are possible.
is the DSpace metadata field to be indexed.
can be used to specify how manipulate the values before indexing.
Example: search.index.12 = language:dc.language.iso:inputform
Possible values are:
text - default, no special treatment. Metadata value are passed to lucene as text
timestamp - the values are interpreted as date with second granularity. An additional index postfixed with .year is created with year granularity
date - the values are interpreted as date with day granularity. An additional index postfixed with .year is created with year granularity
inputform - in addition to the values stored in the metadata the displayed form of this value as derivable from the input-form (in any of the available languages) are stored
In the example above,
search.index.3 are configured as the
author search field. The
author index is created by Lucene indexing all
description.statementofresponsibility metadata fields.
After changing the configuration run
/[dspace]/bin/dspace index-lucene-init to regenerate the indexes.
While the indexes are created, this only affects the search results and has no effect on the search components of the user interface.
In the above examples, notice the asterisk (
*). The metadata field (at least for Dublin Core) is made up of the "element" and the "qualifier". The asterisk is used as the "wildcard". So, for example,
keyword.dc.subject.* will index all subjects regardless if the term resides in a qualified field. (subject versus subject.lcsh). One could customize the search and only index LCSH (Library of Congress Subject Headings) with the following entry
keyword:dc.subject.lcsh instead of
Authority Control Note:
Although DSIndexer automatically builds a separate index for the authority keys of any index that contains authority-controlled metadata fields, the "Advanced Search" UIs do not allow direct access to it. Perhaps it will be added in the future. Fortunately, the OpenSearch API lets you submit a query directly to the Lucene search engine, and this may include the authority-controlled indexes.
Customize the advanced search form
As the previous configuration apply only to the indexing and querying phase one will need to customize the user interface to reflect the changes, for example, to add the a new search category to the Advanced Search.
XML UI requires manual coding of the involved templates instead the JSP UI provides specific configuration to set the index to show in the advanced search dropdown. Below are listed the configuration parameters
jspui.search.index.display.1 = ANY
|Set the N-value of the index dropdown in the advanced search form. The value must match one of the defined index|