Use the Ickle query language with Infinispan caches to efficiently and quickly gain real-time insights into your data. Learn how to configure indexing and perform queries on remote and embedded caches.

1. Indexing Infinispan caches

Infinispan can create indexes of values in your caches to improve query performance, providing faster results than non-indexed queries. Indexing also lets you use full-text search capabilities in your queries.

Infinispan uses Apache Lucene technology to index values in caches.

1.1. Configuring Infinispan to index caches

Enable indexing in your cache configuration and specify which entities Infinispan should include when creating indexes.

You should always configure Infinispan to index caches when using queries. Indexing provides a significant performance boost to your queries, allowing you to get faster insights into your data.

  1. Enable indexing in your cache configuration.

        <!-- Indexing configuration goes here. -->

    Adding an indexing element to your configuration enables indexing without the need to include the enabled=true attribute.

    For remote caches adding this element also implicitly configures encoding as ProtoStream.

  2. Specify the entities to index with the indexed-entity element.


Protobuf messages

  • Specify the message declared in the schema as the value of the indexed-entity element, for example:


    This configuration indexes the Book message in a schema with the book_sample package name.

    package book_sample;
    /* @Indexed */
    message Book {
        /* @Text(projectable = true) */
        optional string title = 1;
        /* @Text(projectable = true) */
        optional string description = 2;
        // no native Date type available in Protobuf
        optional int32 publicationYear = 3;
        repeated Author authors = 4;
    message Author {
        optional string name = 1;
        optional string surname = 2;

Java objects

  • Specify the fully qualified name (FQN) of each class that includes the @Indexed annotation.

import org.infinispan.configuration.cache.*;

ConfigurationBuilder config=new ConfigurationBuilder();

1.1.1. Index configuration

Infinispan configuration controls how indexes are stored and constructed.

Index storage

You can configure how Infinispan stores indexes:

  • On the host file system, which is the default and persists indexes between restarts.

  • In JVM heap memory, which means that indexes do not survive restarts.
    You should store indexes in JVM heap memory only for small datasets.

File system
  <indexing storage="filesystem" path="${}/baseDir">
    <!-- Indexing configuration goes here. -->
JVM heap memory
  <indexing storage="local-heap">
    <!-- Additional indexing configuration goes here. -->
Index path

Specifies a filesystem path for the index when storage is 'filesystem'. The value can be a relative or absolute path. Relative paths are created relative to the configured global persistent location, or to the current working directory when global state is disabled.

By default, the cache name is used as a relative path for index path.

When setting a custom value, ensure that there are no conflicts between caches using the same indexed entities.

Index startup mode

When Infinispan starts caches it can perform operations to ensure the index is consistent with data in the cache. By default no indexing operation takes place when a cache starts but you can configure Infinispan to:

  • Clear the index when the cache starts.

    • Infinispan performs the clear (purge) operation synchronously. The cache becomes available only when the purge completes.

  • Reindex the cache when it starts.

    • Infinispan performs the reindex operation asynchronously. The reindex operation might take a longer time to complete, depending on the size of the cache.

  • Automatically clear or reindex the cache.

    • If data is volatile and the index is persistent then Infinispan clears the cache when it starts.

    • If data is persistent and the index is volatile then Infinispan reindex the cache when it starts.

Clear the index when the cache starts
  <indexing storage="filesystem" startup-mode="purge">
    <!-- Additional indexing configuration goes here. -->
Rebuild the index when the cache starts
  <indexing storage="local-heap" startup-mode="reindex">
    <!-- Additional indexing configuration goes here. -->
Indexing mode

indexing-mode controls how cache operations are propagated to the indexes.


Infinispan immediately applies any changes to the cache to the indexes. This is the default mode.


Infinispan updates indexes only when the reindex operation is explicitly invoked. Configure manual mode, for example, when you want to perform batch updates to the indexes.

Set the indexing-mode to manual:

  <indexing indexing-mode="manual">
    <!-- Additional indexing configuration goes here. -->
Index reader

The index reader is an internal component that provides access to the indexes to perform queries. As the index content changes, Infinispan needs to refresh the reader so that search results are up to date. You can configure the refresh interval for the index reader. By default Infinispan reads the index before each query if the index changed since the last refresh.

  <indexing storage="filesystem" path="${}/baseDir">
    <!-- Sets an interval of one second for the index reader. -->
    <index-reader refresh-interval="1000"/>
    <!-- Additional indexing configuration goes here. -->
Index writer

The index writer is an internal component that constructs an index composed of one or more segments (sub-indexes) that can be merged over time to improve performance. Fewer segments usually means less overhead during a query because index reader operations need to take into account all segments.

Infinispan uses Apache Lucene internally and indexes entries in two tiers: memory and storage. New entries go to the memory index first and then, when a flush happens, to the configured index storage. Periodic commit operations occur that create segments from the previously flushed data and make all the index changes permanent.

The index-writer configuration is optional. The defaults should work for most cases and custom configurations should only be used to tune performance.

  <indexing storage="filesystem" path="${}/baseDir">
    <index-writer commit-interval="2000"
      <index-merge calibrate-by-deletes="true"
    <!-- Additional indexing configuration goes here. -->
Table 1. Index writer configuration attributes
Attribute Description


Amount of time, in milliseconds, that index changes that are buffered in memory are flushed to the index storage and a commit is performed. Because operation is costly, small values should be avoided. The default is 1000 ms (1 second).


Maximum number of entries that can be buffered in-memory before they are flushed to the index storage. Large values result in faster indexing but use more memory. When used in combination with the ram-buffer-size attribute, a flush occurs for whichever event happens first.


Maximum amount of memory that can be used for buffering added entries and deletions before they are flushed to the index storage. Large values result in faster indexing but use more memory. For faster indexing performance you should set this attribute instead of max-buffered-entries. When used in combination with the max-buffered-entries attribute, a flush occurs for whichever event happens first.


Number of threads that execute write operations to the index.


Number of internal queues to use for each indexed type. Each queue holds a batch of modifications that is applied to the index and queues are processed in parallel. Increasing the number of queues will lead to an increase of indexing throughput, but only if the bottleneck is CPU. For optimum results, do not set a value for queue-count that is larger than the value for thread-pool-size.


Maximum number of elements each queue can hold. Increasing the queue-size value increases the amount of memory that is used during indexing operations. Setting a value that is too small can block indexing operations.


Enables low-level trace information for indexing operations. Enabling this attribute substantially degrades performance. You should use this low-level tracing only as a last resource for troubleshooting.

To configure how Infinispan merges index segments, you use the index-merge sub-element.

Table 2. Index merge configuration attributes
Attribute Description


Maximum number of entries that an index segment can have before merging. Segments with more than this number of entries are not merged. Smaller values perform better on frequently changing indexes, larger values provide better search performance if the index does not change often.


Number of segments that are merged at once. With smaller values, merging happens more often, which uses more resources, but the total number of segments will be lower on average, increasing search performance. Larger values (greater than 10) are best for heavy writing scenarios.


Minimum target size of segments, in MB, for background merges. Segments smaller than this size are merged more aggressively. Setting a value that is too large might result in expensive merge operations, even though they are less frequent.


Maximum size of segments, in MB, for background merges. Segments larger than this size are never merged in the background. Settings this to a lower value helps reduce memory requirements and avoids some merging operations at the cost of optimal search speed. This attribute is ignored when forcefully merging an index and max-forced-size applies instead.


Maximum size of segments, in MB, for forced merges and overrides the max-size attribute. Set this to the same value as max-size or lower. However setting the value too low degrades search performance because documents are deleted.


Whether the number of deleted entries in an index should be taken into account when counting the entries in the segment. Setting false will lead to more frequent merges caused by max-entries, but will more aggressively merge segments with many deleted documents, improving query performance.

Index sharding

When you have a large amount of data, you can configure Infinispan to split index data into multiple indexes called shards. Enabling data distribution among shards improves performance. By default, sharding is disabled.

Use the shards attribute to configure the number of indexes. The number of shards must be greater then 1.

    <index-sharding shards="6" />

1.2. Infinispan native indexing annotations

When you enable indexing in caches, you configure Infinispan to create indexes. You also need to provide Infinispan with a structured representation of the entities in your caches so it can actually index them.

Overview of the Infinispan indexing annotations


Indicates entities, or Protobuf message types, that Infinispan indexes.

To indicate the fields that Infinispan indexes use the indexing annotations. You can use these annotations the same way for both embedded and remote queries.


Supports any type of field. Use the @Basic annotation for numbers and short strings that don’t require any transformation or processing.


Use this annotation for fields that represent decimal values.


Use this annotation for fields that are strings and intended for exact matching. Keyword fields are not analyzed or tokenized during indexing.


Use this annotation for fields that contain textual data and are intended for full-text search capabilities. You can use the analyzer to process the text and to generate individual tokens.


Use this annotation to mark vector fields representing embeddings, on which can be defined kNN-predicates.


Use this annotation to mark a field as an embedded object within the parent entity. The NESTED structure preserves the original object relationship structure while the FLATTENED structure makes the leaf fields multivalued of the parent entity. The default structure used by @Embedded is NESTED.

Each of the annotations supports a set of attributes that you can use to further describe how the entity is indexed.

Table 3. Infinispan annotations and supported attributes
Annotation Supported attributes


searchable, sortable, projectable, aggregable, indexNullAs


searchable, sortable, projectable, aggregable, indexNullAs, decimalScale


searchable, sortable, projectable, aggregable, indexNullAs, normalizer, norms


searchable, projectable, norms, analyzer, searchAnalyzer, termVector


searchable, projectable, dimension, similarity, beamWidth, maxConnections

Using Infinispan annotations

You can provide Infinispan with indexing annotations in two ways:

  • Annotate your Java classes or fields directly using the Infinispan annotations.
    You then generate or update your Protobuf schema, .proto files, before uploading them to Infinispan Server.

  • Annotate Protobuf schema directly with @Indexed and @Basic, @Keyword or @Text.
    You then upload your Protobuf schema to Infinispan Server.

    For example, the following schema uses the @Text annotation:

      * @Text(projectable = true)
    required string street = 1;

1.3. Rebuilding indexes

Rebuilding an index reconstructs it from the data stored in the cache. You should rebuild indexes when you change things like the definitions of indexed types or analyzers. Likewise, you can rebuild indexes after you delete them for whatever reason.

Rebuilding indexes can take a long time to complete because the process takes place for all data in the grid. While the rebuild operation is in progress, queries might also return fewer results.


Rebuild indexes in one of the following ways:

  • Call the reindexCache() method to programmatically rebuild an index from a Hot Rod Java client:


    For remote caches you can also rebuild indexes from Infinispan Console.

  • Call the method to rebuild indexes for embedded caches as follows:

    Indexer indexer = Search.getIndexer(cache);
    CompletionStage<Void> future =;
    • Check the status of reindexing operation with the reindexing attribute of the index statistics.

1.4. Updating index schema

The update index schema operation lets you add schema changes with a minimal downtime. Instead of removing previously indexed data and recreating the index schema, Infinispan adds new fields to the existing schema. Updating index schema is much faster than rebuilding the index but you can update schema only when your changes do not affect fields that were already indexed.

You can update index schema only when your changes does not affect previously indexed fields. When you change index field definitions or when you delete fields, you must rebuild the index.

  • Update index schema for a given cache:

    • Call the updateIndexSchema() method to programmatically update the index schema from a Hot Rod Java client:


      For remote caches, you can update index schema from the Infinispan Console or using the REST API.

Additional resources

1.5. Non-indexed queries

Infinispan recommends indexing caches for the best performance for queries. However you can query caches that are non-indexed.

  • For embedded caches, you can perform non-indexed queries on Plain Old Java Objects (POJOs).

  • For remote caches, you must use ProtoStream encoding with the application/x-protostream media type to perform non-indexed queries.

2. Creating Ickle queries

Infinispan provides an Ickle query language that lets you create relational and full-text queries.

2.1. Ickle queries

To use the API, call the cache .query() method and provide the query string.

For instance:

// Remote Query using protobuf
Query<Transaction> q = remoteCache.query("from sample_bank_account.Transaction where amount > 20");

// Embedded Query using Java Objects
Query<Transaction> q = cache.query("from org.infinispan.sample.Book where price > 20");

// Execute the query
QueryResult<Book> queryResult = q.execute();

A query will always target a single entity type and is evaluated over the contents of a single cache. Running a query over multiple caches or creating queries that target several entity types (joins) is not supported.

Executing the query and fetching the results is as simple as invoking the execute() method of the Query object. Once executed, calling execute() on the same instance will re-execute the query.

2.1.1. Pagination

You can limit the number of returned results by using the Query.maxResults(int maxResults). This can be used in conjunction with Query.startOffset(long startOffset) to achieve pagination of the result set.

// sorted by year and match all books that have "clustering" in their title
// and return the third page of 10 results
Query<Book> query = cache.query("FROM org.infinispan.sample.Book WHERE title like '%clustering%' ORDER BY year").startOffset(20).maxResults(10)

If you don’t explicitly set the maxResults for a query instance, Infinispan limits the number of results returned by the query to 100. You can change the default limit by setting the query.default-max-results cache property.

2.1.2. Number of hits

The QueryResult object includes the .hitCount() method, which returns a hit count value that represents the total number of results from a query, regardless of any pagination parameter.

Additionally, QueryResult object contains a boolean value returned by the .isExact() method which indicates whether the hit count number is exact or a lower bound. The hit count is only available for indexed queries for performance reasons.

Hit count accuracy

You can limit the required accuracy of hit counts by setting hit-count-accuracy attribute. When dealing with large data sets, precise hit counts can impact performance. Setting a limit to the hit count accuracy, lets you achieve faster query responses while ensuring that the provided hit counts remain sufficiently accurate for your application’s needs.

The default accuracy of the hit-count-accuracy attribute is limited to 10000. This means that for any query, Infinispan provides exact hit count up to maximum of 10,000. If the effective hit count is higher than 10,000, Infinispan returns a lower bound estimate of the count. You can change the default limit by setting the query.hit-count-accuracy cache property. Alternatively, it can be set on each query instance.

When the actual hit count exceeds the limit set by the hit-count-accuracy, the .isExact() method or the hit_count_exact JSON field will be false, indicating that the returned hit count is an estimation. Setting this value to Integer.MAX would return accurate results for any query, but this can severely impact query performance.

For optimal performance set the property value slightly above the expected hit count. If you do not require accurate hit counts, set it to a low value.

2.1.3. Iteration

The Query object has the .iterator() method to obtain the results lazily. It returns an instance of CloseableIterator that must be closed after usage.

The iteration support for Remote Queries is currently limited, as it will first fetch all entries to the client before iterating.

2.1.4. Named query parameters

Instead of building a new Query object for every execution it is possible to include named parameters in the query which can be substituted with actual values before execution. This allows a query to be defined once and be efficiently executed many times. Parameters can only be used on the right-hand side of an operator and are defined when the query is created by supplying an object produced by the org.infinispan.query.dsl.Expression.param(String paramName) method to the operator instead of the usual constant value. Once the parameters have been defined they can be set by invoking either Query.setParameter(parameterName, value) or Query.setParameters(parameterMap) as shown in the examples below. ⁠

// Defining a query to search for various authors and publication years
Query<Book> query = cache.query("SELECT title FROM org.infinispan.sample.Book WHERE author = :authorName AND publicationYear = :publicationYear").build();

// Set actual parameter values
query.setParameter("authorName", "Doe");
query.setParameter("publicationYear", 2010);

// Execute the query
List<Book> found = query.execute().list();

Alternatively, you can supply a map of actual parameter values to set multiple parameters at once: ⁠

Setting multiple named parameters at once
Map<String, Object> parameterMap = new HashMap<>();
parameterMap.put("authorName", "Doe");
parameterMap.put("publicationYear", 2010);


A significant portion of the query parsing, validation and execution planning effort is performed during the first execution of a query with parameters. This effort is not repeated during subsequent executions leading to better performance compared to a similar query using constant values instead of query parameters.

2.1.5. Query execution

The Query API provides two methods for executing Ickle queries on a cache:

  • Query.execute() runs a SELECT statement and returns a result.

  • Query.executeStatement() runs a DELETE statement and modifies data.

You should always invoke executeStatement() to modify data and invoke execute() to get the result of a query.

2.2. Ickle query language syntax

The Ickle query language is subset of the JPQL query language, with some extensions for full-text.

The parser syntax has some notable rules:

  • Whitespace is not significant.

  • Wildcards are not supported in field names.

  • A field name or path must always be specified, as there is no default field.

  • && and || are accepted instead of AND or OR in both full-text and JPA predicates.

  • ! may be used instead of NOT.

  • A missing boolean operator is interpreted as OR.

  • String terms must be enclosed with either single or double quotes.

  • Fuzziness and boosting are not accepted in arbitrary order; fuzziness always comes first.

  • != is accepted instead of <>.

  • Boosting cannot be applied to >,>=,<,<= operators. Ranges may be used to achieve the same result.

2.2.1. Filtering operators

Ickle support many filtering operators that can be used for both indexed and non-indexed fields.

Operator Description Example


Checks that the left operand is equal to one of the elements from the Collection of values given as argument.

FROM Book WHERE isbn IN ('ZZ', 'X1234')


Checks that the left argument (which is expected to be a String) matches a wildcard pattern that follows the JPA rules.

FROM Book WHERE title LIKE '%Java%'


Checks that the left argument is an exact match of the given value.

FROM Book WHERE name = 'Programming Java'


Checks that the left argument is different from the given value.

FROM Book WHERE language != 'English'


Checks that the left argument is greater than the given value.

FROM Book WHERE price > 20


Checks that the left argument is greater than or equal to the given value.

FROM Book WHERE price >= 20


Checks that the left argument is less than the given value.

FROM Book WHERE year < 2020


Checks that the left argument is less than or equal to the given value.

FROM Book WHERE price ⇐ 50


Checks that the left argument is between the given range limits.

FROM Book WHERE price BETWEEN 50 AND 100

2.2.2. Boolean conditions

Combining multiple attribute conditions with logical conjunction (and) and disjunction (or) operators in order to create more complex conditions is demonstrated in the following example. The well known operator precedence rule for boolean operators applies here, so the order of the operators is irrelevant. Here and operator still has higher priority than or even though or was invoked first.

# match all books that have "Data Grid" in their title
# or have an author named "Manik" and their description contains "clustering"

FROM org.infinispan.sample.Book WHERE title LIKE '%Data Grid%' OR = 'Manik' AND description like '%clustering%'

Boolean negation has highest precedence among logical operators and applies only to the next simple attribute condition.

# match all books that do not have "Data Grid" in their title and are authored by "Manik"
FROM org.infinispan.sample.Book WHERE title != 'Data Grid' AND = 'Manik'

2.2.3. Nested conditions

Changing the precedence of logical operators is achieved with parenthesis:

# match all books that have an author named "Manik" and their title contains
# "Data Grid" or their description contains "clustering"
FROM org.infinispan.sample.Book WHERE = 'Manik' AND ( title like '%Data Grid%' OR description like '% clustering%')

2.2.4. Projections with SELECT statements

In some use cases returning the whole domain object is overkill if only a small subset of the attributes are actually used by the application, especially if the domain entity has embedded entities. The query language allows you to specify a subset of attributes (or attribute paths) to return - the projection. If projections are used then the QueryResult.list() will not return the whole domain entity but will return a List of Object[], each slot in the array corresponding to a projected attribute.

# match all books that have "Data Grid" in their title or description
# and return only their title and publication year
SELECT title, publicationYear FROM org.infinispan.sample.Book WHERE title like '%Data Grid%' OR description like '%Data Grid%'
Project cache entry version

It is possible to project the cache entry version, using the version projection function.

# return the title, publication year and the cache entry version
SELECT b.title, b.publicationYear, version(b) FROM org.infinispan.sample.Book b WHERE b.title like '%Data Grid%'
Project cache entry value

It is possible to project the cache entry value together with other projections. It can be used for instance to project the cache entry value together with the cache entry version in the same Object[] returned hit.

# return the cache entry value and the cache entry version
SELECT b, version(b) FROM org.infinispan.sample.Book b WHERE b.title like '%Data Grid%'
Project the score

If the query is indexed, it is possible to project the score obtained by each matching together with other projections. It can be used for instance to project the cache entry value together with the score in the same Object[] returned hit.

# return the cache entry value and the the score of the matching
SELECT b, score(b) FROM org.infinispan.sample.Book b WHERE b.title like '%Data Grid%'


Ordering the results based on one or more attributes or attribute paths is done with the ORDER BY clause. If multiple sorting criteria are specified, then the order will dictate their precedence.

# match all books that have "Data Grid" in their title or description
# and return them sorted by the publication year and title
FROM org.infinispan.sample.Book WHERE title like '%Data Grid%' ORDER BY publicationYear DESC, title ASC

2.2.5. Grouping and aggregation

Infinispan has the ability to group query results according to a set of grouping fields and construct aggregations of the results from each group by applying an aggregation function to the set of values that fall into each group. Grouping and aggregation can only be applied to projection queries (queries with one or more field in the SELECT clause).

The supported aggregations are: avg, sum, count, max, and min.

The set of grouping fields is specified with the GROUP BY clause and the order used for defining grouping fields is not relevant. All fields selected in the projection must either be grouping fields or else they must be aggregated using one of the grouping functions described below. A projection field can be aggregated and used for grouping at the same time. A query that selects only grouping fields but no aggregation fields is legal. ⁠ Example: Grouping Books by author and counting them.

SELECT author, COUNT(title) FROM org.infinispan.sample.Book WHERE title LIKE '%engine%' GROUP BY author

A projection query in which all selected fields have an aggregation function applied and no fields are used for grouping is allowed. In this case the aggregations will be computed globally as if there was a single global group.


You can apply the following aggregation functions to a field:

Table 4. Index merge attributes
Aggregation function Description


Computes the average of a set of numbers. Accepted values are primitive numbers and instances of java.lang.Number. The result is represented as java.lang.Double. If there are no non-null values the result is null instead.


Counts the number of non-null rows and returns a java.lang.Long. If there are no non-null values the result is 0 instead.


Returns the greatest value found. Accepted values must be instances of java.lang.Comparable. If there are no non-null values the result is null instead.


Returns the smallest value found. Accepted values must be instances of java.lang.Comparable. If there are no non-null values the result is null instead.


Computes the sum of a set of Numbers. If there are no non-null values the result is null instead. The following table indicates the return type based on the specified field.

Table 5. Table sum return type
Field Type Return Type

Integral (other than BigInteger)


Float or Double






Evaluation of queries with grouping and aggregation

Aggregation queries can include filtering conditions, like usual queries. Filtering can be performed in two stages: before and after the grouping operation. All filter conditions defined before invoking the groupBy() method will be applied before the grouping operation is performed, directly to the cache entries (not to the final projection). These filter conditions can reference any fields of the queried entity type, and are meant to restrict the data set that is going to be the input for the grouping stage. All filter conditions defined after invoking the groupBy() method will be applied to the projection that results from the projection and grouping operation. These filter conditions can either reference any of the groupBy() fields or aggregated fields. Referencing aggregated fields that are not specified in the select clause is allowed; however, referencing non-aggregated and non-grouping fields is forbidden. Filtering in this phase will reduce the amount of groups based on their properties. Sorting can also be specified similar to usual queries. The ordering operation is performed after the grouping operation and can reference any of the groupBy() fields or aggregated fields.

2.2.6. DELETE statements

You can delete entities from Infinispan caches with the following syntax:

DELETE FROM <entityName> [WHERE condition]
  • Reference only single entities with <entityName>. DELETE queries cannot use joins.

  • WHERE conditions are optional.

DELETE queries cannot use any of the following:

  • Projections with SELECT statements

  • Grouping and aggregation

  • ORDER BY clauses

Invoke the Query.executeStatement() method to execute DELETE statements.

2.3. Full-text queries

You can perform full-text searches with the Ickle query language.

2.3.1. Fuzzy queries

To execute a fuzzy query add ~ along with an integer, representing the distance from the term used, after the term. For instance

FROM sample_bank_account.Transaction WHERE description : 'cofee'~2

2.3.2. Range queries

To execute a range query define the given boundaries within a pair of braces, as seen in the following example:

FROM sample_bank_account.Transaction WHERE amount : [20 to 50]

2.3.3. Phrase queries

A group of words can be searched by surrounding them in quotation marks, as seen in the following example:

FROM sample_bank_account.Transaction WHERE description : 'bus fare'

2.3.4. Proximity queries

To execute a proximity query, finding two terms within a specific distance, add a ~ along with the distance after the phrase. For instance, the following example will find the words canceling and fee provided they are not more than 3 words apart:

FROM sample_bank_account.Transaction WHERE description : 'canceling fee'~3

2.3.5. Wildcard queries

To search for "text" or "test", use the ? single-character wildcard search:

FROM sample_bank_account.Transaction where description : 'te?t'

To search for "test", "tests", or "tester", use the * multi-character wildcard search:

FROM sample_bank_account.Transaction where description : 'test*'

2.3.6. Regular expression queries

Regular expression queries can be performed by specifying a pattern between /. Ickle uses Lucene’s regular expression syntax, so to search for the words moat or boat the following could be used:

FROM sample_library.Book  where title : /[mb]oat/

2.3.7. Boosting queries

Terms can be boosted by adding a ^ after the term to increase their relevance in a given query, the higher the boost factor the more relevant the term will be. For instance to search for titles containing beer and wine with a higher relevance on beer, by a factor of 3, the following could be used:

FROM sample_library.Book WHERE title : beer^3 OR wine

2.4. Vector search queries

You can perform vector kNN searches with the Ickle query language using the special operator <-> to define predicates.

This is an example of kNN query:

from play.Item i where i.myVector <-> [7,7,7]~3

This query will find the items that have the myVector fields that are within 3 nearest neighbourhood from the vector [7,7,7].

Notice that in order to use this kind of search the entity, in our example play.Item, has to be @Indexed and the field, in our example myVector, should be annotated with @Vector.

We support two kinds of vector field types:

  • byte / Byte (to work with byte vectors)

  • float / Float (to work with float vectors)

You can have different vector fields on the same entity, but in any case you can have only one vector predicate on your queries.

2.4.1. Vector search parameters

Both the k-value and the vector can be passed as query parameter. The k-value scalar can be expressed with the usual placeholder :k in the Ickle text. For the vector we can use either a placeholder for each term of the vector:

Query<Item> query = cache.query("from play.Item i where i.floatVector <-> [:a,:b,:c]~:k");
query.setParameter("a", 1);
query.setParameter("b", 4.3);
query.setParameter("c", 3.3);
query.setParameter("k", 4);

Or a placeholder can be used for the entire vector:

Query<Item> query = cache.query("from play.Item i where i.floatVector <-> [:a]~:k");
query.setParameter("a", new float[]{7.1f, 7.0f, 3.1f});
query.setParameter("k", 3);

Is very common also to return the score of the computation, using the score projection. In the case of vector search the query will be like the following:

Query<Object[]> query = cache.query("select i, score(i) from play.Item i where i.floatVector <-> [:a]~:k");
query.setParameter("a", new float[]{7.1f, 7.0f, 3.1f});
query.setParameter("k", 3);

List<Object[]> resultList = query.list();

In this case the first element of each array will contain the entity, and the second element will contain the score of the matching.

2.4.3. Filtering entities

Instead of applying the kNN search to the entire population of entities of a given type, you can limiting the searching set applying classic predicates (match, full-text-search, range, …​) to the kNN queries, defining what we call a filtering clause.

A filtering clause can contain any kind of predicates with the only exception of kNN predicates that cannot be included.

For instance the following query:

Query<Object[]> query = remoteCache.query(
   "select score(i), i from Item i where i.floatVector <-> [:a]~:k filtering (i.buggy : 'cat' or i.text : 'code')");
query.setParameter("a", new float[]{7, 7, 7});
query.setParameter("k", 3);

Will return the nearest 3 items from the point [7,7,7] selecting only the items that have a text containing the term cat or code.

The filtering queries is a way to apply the classic indexed search to the new vector search.

2.4.4. Vector field attributes

It is always required to specify the dimension of the vector field.

The other mapping attributes are optional, since Infinispan will have a default for each of them. You can configure them, for instance, to tune the desired accuracy / performance.


Different VectorSimilarity algorithms are supported

Value Distance Score Note


\(d(x,y) = \sqrt{\sum_{i=1}^{n} (x_i - y_i)^2 } \)

\(s = \frac{1}{1+d^2}\)

This is the Infinispan default


\(d(x,y) = \sum_{i=1}^{n} x_i \cdot y_i \)

\(s = \frac{1}{1+d}\)

To use this similarity efficiently, both index and search vectors must be normalized


\(d(x,y) = \sum_{i=1}^{n} x_i \cdot y_i \)

s = \begin{cases} \frac{1}{1-d} & \text{if d < 0}\\ d+1 & \text{otherwise} \end{cases}

This similarity does not require vector normalization


\(d(x,y) = \frac{1 - \sum_{i=1} ^{n} x_i \cdot y_i }{ \sqrt{ \sum_{i=1} ^{n} x_i^2 } \sqrt{ \sum_{i=1} ^{n} y_i^2 }} \)

\(s = \frac{1}{1+d}\)

This similarity cannot be of zero magnitude, e.g. when a vector is all zeroes: [0,0,0,…​ 0,0], the cosine is not defined and will result in an error.


Changing the beamWidth you can modify the the size of the dynamic list used during k-NN graph creation. It affects how vectors are stored. Higher values lead to a more accurate graph but slower indexing speed. The Infinispan default is 512.

Max Connections

The number of neighbors each node will be connected to in the HNSW graph. Modifying this value will have an impact on memory consumption. It is recommended to keep this value between 2 and 100. The Infinispan default is 16.

3. Querying remote caches

You can index and query remote caches on Infinispan Server.

3.1. Querying caches from Hot Rod Java clients

Infinispan lets you programmatically query remote caches from Java clients through the Hot Rod endpoint. This procedure explains how to index query a remote cache that stores Book instances.

  • Add the ProtoStream processor to your pom.xml.

Infinispan provides this processor for the @ProtoField annotations so you can generate Protobuf schemas and perform queries.

  1. Add indexing annotations to your class, as in the following example:
    import org.infinispan.api.annotations.indexing.Basic;
    import org.infinispan.api.annotations.indexing.Indexed;
    import org.infinispan.api.annotations.indexing.Text;
    import org.infinispan.protostream.annotations.ProtoFactory;
    import org.infinispan.protostream.annotations.ProtoField;
    public class Book {
       @ProtoField(number = 1)
       final String title;
       @ProtoField(number = 2)
       final String description;
       @ProtoField(number = 3, defaultValue = "0")
       final int publicationYear;
       Book(String title, String description, int publicationYear) {
          this.title = title;
          this.description = description;
          this.publicationYear = publicationYear;
       // public Getter methods omitted for brevity
  2. Implement the SerializationContextInitializer interface in a new class and then add the @ProtoSchema annotation.

    1. Reference the class that includes the @ProtoField annotations with the includeClasses parameter.

    2. Define a name for the Protobuf schema that you generate and filesystem path with the schemaFileName and schemaFilePath parameters.

    3. Specify the package name for the Protobuf schema with the schemaPackageName parameter.
      import org.infinispan.protostream.SerializationContextInitializer;
      import org.infinispan.protostream.annotations.ProtoSchema;
            includeClasses = {
            schemaFileName = "book.proto",
            schemaFilePath = "proto/",
            schemaPackageName = "book_sample")
      public interface RemoteQueryInitializer extends SerializationContextInitializer {
  3. Compile your project.

    The code examples in this procedure generate a proto/book.proto schema and an implementation of the annotated Book class.

Next steps

Create a remote cache that configures Infinispan to index your entities. For example, the following remote cache indexes the Book entity in the book.proto schema that you generated in the previous step:

<replicated-cache name="books">

The following RemoteQuery class does the following:

  • Registers the RemoteQueryInitializerImpl serialization context with a Hot Rod Java client.

  • Registers the Protobuf schema, book.proto, with Infinispan Server.

  • Adds two Book instances to the remote cache.

  • Performs a full-text query that matches books by keywords in the title.
package org.infinispan;

import java.nio.file.Files;
import java.nio.file.Path;
import java.nio.file.Paths;
import java.util.List;

import org.infinispan.client.hotrod.RemoteCache;
import org.infinispan.client.hotrod.RemoteCacheManager;
import org.infinispan.client.hotrod.Search;
import org.infinispan.client.hotrod.configuration.ConfigurationBuilder;
import org.infinispan.query.dsl.Query;
import org.infinispan.query.dsl.QueryFactory;
import org.infinispan.query.remote.client.ProtobufMetadataManagerConstants;

public class RemoteQuery {

   public static void main(String[] args) throws Exception {
      ConfigurationBuilder clientBuilder = new ConfigurationBuilder();
      // RemoteQueryInitializerImpl is generated
            .addContextInitializers(new RemoteQueryInitializerImpl());

      RemoteCacheManager remoteCacheManager = new RemoteCacheManager(;

      // Grab the generated protobuf schema and registers in the server.
      Path proto = Paths.get(RemoteQuery.class.getClassLoader()
      String protoBufCacheName = ProtobufMetadataManagerConstants.PROTOBUF_METADATA_CACHE_NAME;
      remoteCacheManager.getCache(protoBufCacheName).put("book.proto", Files.readString(proto));

      // Obtain the 'books' remote cache
      RemoteCache<Object, Object> remoteCache = remoteCacheManager.getCache("books");

      // Add some Books
      Book book1 = new Book("Infinispan in Action", "Learn Infinispan with using it", 2015);
      Book book2 = new Book("Cloud-Native Applications with Java and Quarkus", "Build robust and reliable cloud applications", 2019);

      remoteCache.put(1, book1);
      remoteCache.put(2, book2);

      // Execute a full-text query
      Query<Book> query = remoteCache.query("FROM book_sample.Book WHERE title:'java'");

      List<Book> list = query.execute().list(); // Voila! We have our book back from the cache!

3.2. Querying ProtoStream common types

Perform Ickle queries on caches that store data as ProtoStream common types such as BigInteger and BigDecimal.

  1. Add indexing annotations to your class, as in the following example:

    public class CalculusIndexed {
        @ProtoField(value = 1)
        public BigInteger getPurchases() {
          return purchases;
        @Decimal // the scale is 2 by default
        @ProtoField(value = 2)
        public BigDecimal getProspect() {
          return prospect;
  2. Set the dependsOn attribute to CommonTypes.class to indicate that the generated Protobuf schema can reference and use CommonTypes types such as BigInteger and BigDecimal:

    @ProtoSchema(includeClasses = CalculusIndexed.class, dependsOn = CommonTypes.class,
         schemaFilePath = "/protostream", schemaFileName = "calculus-indexed.proto",
         schemaPackageName = "lab.indexed")
    public interface CalculusIndexedSchema extends GeneratedSchema {
  3. Perform queries:

    Query<Product> query = cache.query("from lab.indexed.CalculusIndexed c where c.purchases > 9");
    QueryResult<Product> result = query.execute();
    // play with the result
    query = cache.query("from lab.indexed.CalculusIndexed c where c.prospect = 2.2");
    result = query.execute();
    // play with the result
Additional resources

3.3. Querying caches from Infinispan Console and CLI

Infinispan Console and the Infinispan Command Line Interface (CLI) let you query indexed and non-indexed remote caches. You can also use any HTTP client to index and query caches via the REST API.

This procedure explains how to index and query a remote cache that stores Person instances.

  • Have at least one running Infinispan Server instance.

  • Have Infinispan credentials with create permissions.

  1. Add indexing annotations to your Protobuf schema, as in the following example:

    package org.infinispan.example;
    /* @Indexed */
    message Person {
        /* @Basic */
        optional int32 id = 1;
        /* @Keyword(projectable = true) */
        required string name = 2;
        /* @Keyword(projectable = true) */
        required string surname = 3;
        /* @Basic(projectable = true, sortable = true) */
        optional int32 age = 6;

    From the Infinispan CLI, use the schema command with the --upload= argument as follows:

    schema --upload=person.proto person.proto
  2. Create a cache named people that uses ProtoStream encoding and configures Infinispan to index entities declared in your Protobuf schema.

    The following cache indexes the Person entity from the previous step:

    <distributed-cache name="people">
      <encoding media-type="application/x-protostream"/>

    From the CLI, use the create cache command with the --file= argument as follows:

    create cache --file=people.xml people
  3. Add entries to the cache.

    To query a remote cache, it needs to contain some data. For this example procedure, create entries that use the following JSON values:


    From the CLI, use the put command with the --file= argument to add each entry, as follows:

    put --encoding=application/json --file=personone.json personone

    From Infinispan Console, you must select Custom Type for the Value content type field when you add values in JSON format with custom types .

  4. Query your remote cache.

    From the CLI, use the query command from the context of the remote cache.

    query "from org.infinispan.example.Person p WHERE'Person' ORDER BY p.age ASC"

    The query returns all entries with a name that matches Person by age in ascending order.

Additional resources

3.4. Using analyzers with remote caches

Analyzers convert input data into terms that you can index and query. You specify analyzer definitions with the @Text annotation in your Java classes or directly in Protobuf schema.

  1. Annotate the property with the @Text annotation to indicate that its value is analyzed.

  2. Use the analyzer attribute to specify the desired analyzer that you want to use for indexing and searching.

Protobuf schema
/* @Indexed */
message TestEntity {

    /* @Keyword(projectable = true) */
    optional string id = 1;

    /* @Text(projectable = true, analyzer = "simple") */
    optional string name = 2;
Java classes
@Text(projectable = true, analyzer = "whitespace")
@ProtoField(value = 1)
private String id;

@Text(projectable = true, analyzer = "simple")
@ProtoField(value = 2)
private String description;

3.4.1. Default analyzer definitions

Infinispan provides a set of default analyzer definitions.

Definition Description


Splits text fields into tokens, treating whitespace and punctuation as delimiters.


Tokenizes input streams by delimiting at non-letters and then converting all letters to lowercase characters. Whitespace and non-letters are discarded.


Splits text streams on whitespace and returns sequences of non-whitespace characters as tokens.


Treats entire text fields as single tokens.


Stems English words using the Snowball Porter filter.


Generates n-gram tokens that are 3 grams in size by default.


Splits text fields into larger size tokens than the standard analyzer, treating whitespace as a delimiter and converts all letters to lowercase characters.


Converts all the letters of the text to lowercase characters, the text is not tokenized (normalizer).

These analyzer definitions are based on Apache Lucene. For more information about tokenizers, filters, and CharFilters, see the Apache Lucene documentation.

Additional resources

3.4.2. Creating custom analyzer definitions

Create custom analyzer definitions and add them to your Infinispan Server installations.

  • Stop Infinispan Server if it is running.

    Infinispan Server loads classes at startup only.

  1. Implement the ProgrammaticSearchMappingProvider API.

  2. Package your implementation in a JAR with the fully qualified class (FQN) in the following file:

  3. Copy your JAR file to the server/lib directory of your Infinispan Server installation.

  4. Start Infinispan Server.

ProgrammaticSearchMappingProvider example
import org.apache.lucene.analysis.core.LowerCaseFilterFactory;
import org.apache.lucene.analysis.core.StopFilterFactory;
import org.apache.lucene.analysis.standard.StandardFilterFactory;
import org.apache.lucene.analysis.standard.StandardTokenizerFactory;
import org.infinispan.Cache;
import org.infinispan.query.spi.ProgrammaticSearchMappingProvider;

public final class MyAnalyzerProvider implements ProgrammaticSearchMappingProvider {

   public void defineMappings(Cache cache, SearchMapping searchMapping) {
            .analyzerDef("standard-with-stop", StandardTokenizerFactory.class)

3.5. Queries by keys

If a key of a cache entry has an indexed type, it is possible to index the key fields as well the the value fields, so that also the formers could be used in the Ickle queries.

This can be done specifying the fully qualified name of the ProtocolBuffer message type to use as key type in the keyEntity attribute of the @Indexed annotation.

This feature is at the moment available only with indexed remote queries.

Specify the keyEntity of an indexed entity
import org.infinispan.api.annotations.indexing.Basic;
import org.infinispan.api.annotations.indexing.Indexed;
import org.infinispan.api.annotations.indexing.Text;
import org.infinispan.protostream.GeneratedSchema;
import org.infinispan.protostream.annotations.ProtoFactory;
import org.infinispan.protostream.annotations.ProtoField;
import org.infinispan.protostream.annotations.ProtoSchema;

@Indexed(keyEntity = "model.StructureKey")
public class Structure {

   private final String code;
   private final String description;
   private final Integer value;

   public Structure(String code, String description, Integer value) {
      this.code = code;
      this.description = description;
      this.value = value;

   public String getCode() {
      return code;

   public String getDescription() {
      return description;

   public Integer getValue() {
      return value;

   @ProtoSchema(includeClasses = { Structure.class, StructureKey.class }, schemaPackageName = "model")
   public interface StructureSchema extends GeneratedSchema {
      StructureSchema INSTANCE = new StructureSchemaImpl();
Define the key entity and its indexed fields
import org.infinispan.api.annotations.indexing.Basic;
import org.infinispan.api.annotations.indexing.Indexed;
import org.infinispan.api.annotations.indexing.Keyword;
import org.infinispan.protostream.annotations.ProtoFactory;
import org.infinispan.protostream.annotations.ProtoField;

public class StructureKey {

   private String zone;
   private Integer row;
   private Integer column;

   public StructureKey(String zone, Integer row, Integer column) { = zone;
      this.row = row;
      this.column = column;

   @Keyword(projectable = true, sortable = true)
   public String getZone() {
      return zone;

   @Basic(projectable = true, sortable = true)
   public Integer getRow() {
      return row;

   @Basic(projectable = true, sortable = true)
   public Integer getColumn() {
      return column;

3.5.1. Key property name

By default, the key fields will be targeted using the property named key.

Use key properties in the Ickle queries
select s.key.column from model.Structure s where = 'z7'

If the value already has a property named key, the definition of the key entity could create a naming conflict with the properties. For this reason but also in general, it is possible to change the name to assign as a prefix for the property keys changing the attribute keyPropertyName of the @Indexed annotation.

3.5.2. Key include depth

Since the entity key can have embedded entities we can limit the depth for the embedded entity fields that will be indexed changing the attribute keyIncludeDepth.

The default for this value is 3.

4. Querying embedded caches

Use embedded queries when you add Infinispan as a library to custom applications.

Protobuf mapping is not required with embedded queries. Indexing and querying are both done on top of Java objects.

4.1. Querying embedded caches

This section explains how to query an embedded cache using an example cache named "books" that stores indexed Book instances.

In this example, each Book instance defines which properties are indexed and specifies some advanced indexing options with Hibernate Search annotations as follows:
package org.infinispan.sample;

import java.time.LocalDate;
import java.util.HashSet;
import java.util.Set;

import org.infinispan.api.annotations.indexing.*;

// Annotate values with @Indexed to add them to indexes
// Annotate each field according to how you want to index it
public class Book {
   String title;

   String description;

   String isbn;

   LocalDate publicationDate;

   Set<Author> authors = new HashSet<Author>();
package org.infinispan.sample;

import org.infinispan.api.annotations.indexing.Text;

public class Author {
   String name;

   String surname;
  1. Configure Infinispan to index the "books" cache and specify org.infinispan.sample.Book as the entity to index.

    <distributed-cache name="books">
      <indexing path="${user.home}/index">
  2. Obtain the cache.

    import org.infinispan.Cache;
    import org.infinispan.manager.DefaultCacheManager;
    import org.infinispan.manager.EmbeddedCacheManager;
    EmbeddedCacheManager manager = new DefaultCacheManager("infinispan.xml");
    Cache<String, Book> cache = manager.getCache("books");
  3. Perform queries for fields in the Book instances that are stored in the Infinispan cache, as in the following example:

    // Create an Ickle query that performs a full-text search using the ':' operator on the 'title' and '' fields
    // You can perform full-text search only on indexed caches
    Query<Book> fullTextQuery = cache.query("FROM org.infinispan.sample.Book b WHERE b.title:'infinispan' AND'sanne'");
    // Use the '=' operator to query fields in caches that are indexed or not
    // Non full-text operators apply only to fields that are not analyzed
    Query<Book> exactMatchQuery= cache.query("FROM org.infinispan.sample.Book b WHERE b.isbn = '12345678' AND : 'sanne'");
    // You can use full-text and non-full text operators in the same query
    Query<Book> query= cache.query("FROM org.infinispan.sample.Book b where : 'Stephen' and b.description : (+'dark' -'tower')");
    // Get the results
    List<Book> found=query.execute().list();

4.2. Entity mapping annotations

Add annotations to your Java classes to map your entities to indexes.

Hibernate Search API

Infinispan uses the Hibernate Search API to define fine grained configuration for indexing at entity level. This configuration includes which fields are annotated, which analyzers should be used, how to map nested objects, and so on.

The following sections provide information that applies to entity mapping annotations for use with Infinispan.

For complete detail about these annotations, you should refer to the Hibernate Search manual.


Unlike Hibernate Search, using @DocumentId to mark a field as identifier does not apply to Infinispan values; in Infinispan the identifier for all @Indexed objects is the key used to store the value. You can still customize how the key is indexed using a combination of @Transformable , custom types and custom FieldBridge implementations.

@Transformable keys

The key for each value needs to be indexed as well, and the key instance must be transformed in a String. Infinispan includes some default transformation routines to encode common primitives, but to use a custom key you must provide an implementation of org.infinispan.query.Transformer .

Registering a key Transformer via annotations

You can annotate your key class with org.infinispan.query.Transformable and your custom transformer implementation will be picked up automatically:

@Transformable(transformer = CustomTransformer.class)
public class CustomKey {

public class CustomTransformer implements Transformer {
   public Object fromString(String s) {
      return new CustomKey(...);

   public String toString(Object customType) {
      CustomKey ck = (CustomKey) customType;
      return ...
Registering a key Transformer via the cache indexing configuration

Use the key-transformers xml element in both embedded and server config:

<replicated-cache name="test">
      <key-transformer key="com.mycompany.CustomKey"

Alternatively, use the Java configuration API (embedded mode):

   ConfigurationBuilder builder = ...
         .addKeyTransformer(CustomKey.class, CustomTransformer.class);

5. Creating continuous queries

Applications can register listeners to receive continual updates about cache entries that match query filters.

5.1. Continuous queries

Continuous queries provide applications with real-time notifications about data in Infinispan caches that are filtered by queries. When entries match the query Infinispan sends the updated data to any listeners, which provides a stream of events instead of applications having to execute the query.

Continuous queries can notify applications about incoming matches, for values that have joined the set; updated matches, for matching values that were modified and continue to match; and outgoing matches, for values that have left the set.

For example, continuous queries can notify applications about all:

  • Persons with an age between 18 and 25, assuming the Person entity has an age property and is updated by the user application.

  • Transactions for dollar amounts larger than $2000.

  • Times where the lap speed of F1 racers were less than 1:45.00 seconds, assuming the cache contains Lap entries and that laps are entered during the race.

Continuous queries can use all query capabilities except for grouping, aggregation, and sorting operations.

How continuous queries work

Continuous queries notify client listeners with the following events:


A cache entry matches the query.


A cache entry that matches the query is updated and still matches the query.


A cache entry no longer matches the query.

When a client registers a continuous query listener it immediately receives Join events for any entries that match the query. Client listeners receive subsequent events each time a cache operation modifies entries that match the query.

Infinispan determines when to send Join, Update, or Leave events to client listeners as follows:

  • If the query on both the old and new value does not match, Infinispan does not sent an event.

  • If the query on the old value does not match but the new value does, Infinispan sends a Join event.

  • If the query on both the old and new values match, Infinispan sends an Update event.

  • If the query on the old value matches but the new value does not, Infinispan sends a Leave event.

  • If the query on the old value matches and the entry is then deleted or it expires, Infinispan sends a Leave event.

5.1.1. Continuous queries and Infinispan performance

Continuous queries provide a constant stream of updates to applications, which can generate a significant number of events. Infinispan temporarily allocates memory for each event it generates, which can result in memory pressure and potentially lead to OutOfMemoryError exceptions, especially for remote caches. For this reason, you should carefully design your continuous queries to avoid any performance impact.

Infinispan strongly recommends that you limit the scope of your continuous queries to the smallest amount of information that you need. To achieve this, you can use projections and predicates. For example, the following statement provides results about only a subset of fields that match the criteria rather than the entire entry:

SELECT field1, field2 FROM Entity WHERE x AND y

It is also important to ensure that each ContinuousQueryListener you create can quickly process all received events without blocking threads. To achieve this, you should avoid any cache operations that generate events unnecessarily.

5.2. Creating continuous queries

You can create continuous queries for remote and embedded caches.

  1. Create a Query object.

  2. Obtain the ContinuousQuery object of your cache by calling the appropriate method:

    • Remote caches: org.infinispan.client.hotrod.Search.getContinuousQuery(RemoteCache<K, V> cache)

    • Embedded caches: org.infinispan.query.Search.getContinuousQuery(Cache<K, V> cache)

  3. Register the query and a ContinuousQueryListener object as follows:

    continuousQuery.addContinuousQueryListener(query, listener);
  4. When you no longer need the continuous query, remove the listener as follows:


Continuous query example

The following code example demonstrates a simple continuous query with an embedded cache.

In this example, the listener receives notifications when any Person instances under the age of 21 are added to the cache. Those Person instances are also added to the "matches" map. When the entries are removed from the cache or their age becomes greater than or equal to 21, they are removed from "matches" map. ⁠

Registering a Continuous Query
import org.infinispan.query.api.continuous.ContinuousQuery;
import org.infinispan.query.api.continuous.ContinuousQueryListener;
import org.infinispan.query.Search;
import org.infinispan.query.dsl.QueryFactory;
import org.infinispan.query.dsl.Query;

import java.util.Map;
import java.util.concurrent.ConcurrentHashMap;


// We have a cache of Person objects.
Cache<Integer, Person> cache = ...

// Create a ContinuousQuery instance on the cache.
ContinuousQuery<Integer, Person> continuousQuery = Search.getContinuousQuery(cache);

// Define a query.
// In this example, we search for Person instances under 21 years of age.
Query query = cache.query("FROM Person p WHERE p.age < 21");

final Map<Integer, Person> matches = new ConcurrentHashMap<Integer, Person>();

// Define the ContinuousQueryListener.
ContinuousQueryListener<Integer, Person> listener = new ContinuousQueryListener<Integer, Person>() {
    public void resultJoining(Integer key, Person value) {
        matches.put(key, value);

    public void resultUpdated(Integer key, Person value) {
        // We do not process this event.

    public void resultLeaving(Integer key) {

// Add the listener and the query.
continuousQuery.addContinuousQueryListener(query, listener);


// Remove the listener to stop receiving notifications.

6. Monitoring and tuning Infinispan queries

Infinispan exposes statistics for queries and provides attributes that you can adjust to improve query performance.

6.1. Getting query statistics

Collect statistics to gather information about performance of your indexes and queries, including information such as the types of indexes, average time for queries to complete and the number of possible failures on indexing operations.


Do one of the following:

  • Invoke the getSearchStatistics() or getClusteredSearchStatistics() methods for embedded caches.

  • Use GET requests to obtain statistics for remote caches from the REST API.

Embedded caches
// Statistics for the local cluster member
SearchStatistics statistics = Search.getSearchStatistics(cache);

// Consolidated statistics for the whole cluster
CompletionStage<SearchStatisticsSnapshot> statistics = Search.getClusteredSearchStatistics(cache)
Remote caches
GET /rest/v2/caches/{cacheName}/search/stats

6.2. Tuning query performance

Use the following guidelines to help you improve the performance of indexing operations and queries.

Checking index usage statistics

Queries against partially indexed caches return slower results. For instance, if some fields in a schema are not annotated then the resulting index does not include those fields.

Start tuning query performance by checking the time it takes for each type of query to run. If your queries seem to be slow, you should make sure that queries are using the indexes for caches and that all entities and field mappings are indexed.

Adjusting the commit interval for indexes

Indexing can degrade write throughput for Infinispan clusters. The commit-interval attribute defines the interval, in milliseconds, between which index changes that are buffered in memory are flushed to the index storage and a commit is performed.

This operation is costly so you should avoid configuring an interval that is too small. The default is 1000 ms (1 second).

Adjusting the refresh interval for queries

The refresh-interval attribute defines the interval, in milliseconds, between which the index reader is refreshed.

The default value is 0, which returns data in queries as soon as it is written to a cache.

A value greater than 0 results in some stale query results but substantially increases throughput, especially in write-heavy scenarios. If you do not need data returned in queries as soon as it is written, you should adjust the refresh interval to improve query performance.