Solr allows you to load custom code to perform a variety of tasks within Solr – from custom Request Handlers to process your searches, to custom Analyzers and Token Filters for your text field, even custom Field Types.
How to Load Plugins
Plugin code can be loaded into Solr by putting your classes into a JAR file, and then configuring Solr to know how to find them.
If you want to use multiple SolrCores, and have a plugin available to all of them, you can place your JAR files in a directory specified using the "sharedLib" attribute in your solr.xml file prior to starting your servlet container. In version 4.3.0 and earlier, you can set the sharedLib attribute to "lib" to load from $SOLR_HOME/lib ... but in 4.3.1 and later, if you want to use $SOLR_HOME/lib, you must NOT configure sharedLib. The lib directory in the solr home is automatically loaded in newer versions.
For loading plugins in individual SolrCores, you have two options:
- Place your JARs in a
libdirectory in the instanceDir of your SolrCore. In the example program, the location is example/solr/lib. This directory does not exist in the distribution, so you would need to do
mkdirfor the first time.
- Use the lib directive in your solrconfig.xml file to specify an arbitrary JAR path, directory of JAR files, or a directory plus regex that JAR file names must match. Do not load jars in this way that have already been loaded from $SOLR_HOME/lib or $INSTANCEDIR/lib.
Loading plugins uses a custom Class Loader. It has been tested with a variety of Servlet Containers, but given the multitudes of servlet containers available in the wild it may not always work with every servlet container.
The Old Way
Another method that works consistently on any servlet container is to:
- unpack the solr.war
- add a jar containing your custom classes to the
- repack your new, customized, solr.war and use it.
Classes that are 'Pluggable'
The following is a complete list of every API that can be treated as a plugin in Solr, with information on how to use that configure your Solr instance to use an instance of that class.
Instances of SolrRequestHandler define the logic that is executed for any request. Multiple handlers (including multiple instances of the same SolrRequestHandler class with different configurations) can be specified in your solrconfig.xml...
for more info, see: SolrRequestHandler
for more info, see: SearchComponent
First, implement a subclass of QParserPlugin. This consists primarily of implementing the parse() method to construct the appropriate Query objects.
Next register it in solrconfig.xml like this:
See SolrQuerySyntax and it's list of useful parser implementations for examples of writing a !QParserPlugin.
Having done this, you can choose to use your query parser on a one-time basis using the defType query parameter, like this:
You can also specify your query parser as part of the q parameter, like this:
For more permanent use, you will likely want to define a separate SolrRequestHandler for your parser, like this:
Use this to plugin your own custom functions see FunctionQuery.
register in solrconfig.xml directly under the <config> tag
The class must implement org.apache.solr.search.ValueSourceParser
:TODO: NEED DOCS
:TODO: NEED DOCS
:TODO: NEED DOCS
:TODO: NEED DOCS
Instances of QueryResponseWriter define the formatting used to output the results of a request. Multiple writers (including multiple instances of the same QueryResponseWriter class with different configurations) can be specified in your solrconfig.xml...
The Similarity class is a native Lucene concept that determines how much of the score calculations for the various types of queries are executed. For more information on how the methods in the Similarity class are used, consult the Lucene scoring documentation. If you wish to override the DefaultSimilarity provided by Lucene, you can specify your own subclass in your schema.xml...
The CacheRegenerator API allows people who are writing custom SolrRequestHandlers which utilize custom User Caches to specify how those caches should be populated during autowarming. A regenerator class can be specified when the cache is declared in your solrconfig.xml...
The Analyzer class is a native Lucene concept that determines how tokens are produced from a piece of text. Solr allows Analyzers to be specified for each fieldtype in your schema.xml that uses the TextField class – and even allows for different Analyzers to be specified for indexing text as documents are added, and parsing text specified in a query...
Solr also provides a SolrAnalyzer base class which can be used if you want to write your own Analyzer and configure the "positionIncrementGap" in your schema.xml...
Specifying an Analyzer class in your schema.xml makes a lot of sense if you already have an existing Analyzer you wish to use as is, but if you are planning to write Analysis code from scratch that you would like to use in Solr, you should keep reading the following sections...
Tokenizer and TokenFilter
In addition to specifing specific Analyzer classes, Solr can construct Analyzers on the fly for each field type using a Tokenizer and any number of TokenFilters. To take advantage of this functionality with any Tokenizers or TokenFilters you may have (or may want to implement) you'll need to provide a TokenizerFactory and TokenFilterFactory which takes care of any initialization and configuration, and specify these Factories in your schema.xml...
If you have very specialized data type needs, you can specify your own FieldType class for each <
fieldtype> you declare in your schema.xml, to control how the values for those fields are encoded in your index...
newSearcher events trigger the
newSearcher() method with the appropriate args,
postOptimize events will trigger the
The UpdateHandler API allows you to specify a custom algorithm for determining how sequences of adds and deletes are processed by Solr. The UpdateHandler you wish to use can be configured in your solrconfig.xml, but implementing a new UpdateHandler is considered extremely advanced and is not recommended....
To develop your own plugins, add the
apache-solr-*.jar jar files to the classpath you use to compile your code. They contains all of the Solr Interfaces and Class files you may need. If you are developing plugins that know about lower level Lucene interfaces, you may need to also include the
lucene-*.jar jar files from the
lib/ directory of your Solr distribution.
Plugins can also be initialized with init( PluginInfo info ). These Plugins can have children ( any Subnode with a "class" attribute is considered a child ). If a pluginInfoInitialized Plugin wants to initialise these children in its
inform( core ) method via
, the children must not be SolrCoreAware. see PluginInfo.
Solr3.1 As of Solr3.1 any subnode which is not a "lst","str","int","bool","arr","float" or "double" is considered a child. (Yes "long" is missing - see SOLR-2541)
Classes that need to know about the [SolrCore] can implement SolrCoreAware. Classes implementing SolrCoreAware must not have a dedicated Constructor. Valid SolrCoreAware classes include:
The initialization lifecycle is: