Major updates:

• Java 17 is required
• Update RDF4J to 5.0.2
• Update Jena to 5.1

Minor updates:

• Update to log4j-core to 2.16.0
• Update JUnit to 5
• Update orj.json:json to 20240205
• Update org.apache.httpcomponents:httpclient to 4.5.14
• Update org.apache.commons:commons-pool2 to 2.12.0
• Update commons-codec to 1.15
• Exclude commons-collections:3.2.2 from rdf4j-queryresultio-text
• Pin solr-solrj at 9.4.1
• Pin org.apache.zookeeper at 3.9.2

Migrate from the deprecated Apache (commons) HttpClient library to httpcomponents-HttpClient.

The AllegroGraph server, starting in vesion 7.1.0, may return a 204 (no content) response when it knows that it will not be returning any data. The client will then create a null cursor, rather than creating an http cursor which when used immediately returns 'no data left'

The server-side support for setting query options on a per-user, per-repository basis is available in AG v7.1.0 and later and can be accessed from the Java client via AGRepositoryConnection methods:

• Map<String, String> getQueryOptions();
• String getQueryOption(String name);
• void setQueryOption(String name, String value);
• void removeQueryOption(Sring name).

These methods mimic the namespaces API, the only difference being the AGRepositoryConnection.getQueryOptions() method that returns a Map<String, String> instead of a RepositoryResult for simplicity reasons.

If this property has a value true (the default) then connections are taken out of the pool in a Last In First Out manner. If this property has a value false then connections are taken out of the pool in a First In First Out manner. For a connection pool to a load balancer FIFO is a better choice as it will cause work to be distributed to all backends.

Also add a document a test cases related to handling failures during a transaction.

Use HTTPS to access repo1.maven.org to avoid errors.

The server-side support for loading data into multiple contexts is available since v7.0.0. In order to enable it, the system property "com.franz.agraph.http.AGProtocol.overrideContext" (the name is also accessible via AGProtocol.PROP_OVERRIDE_CONTEXT) must be set to "true", otherwise all but first context arguments are ignored.

The new system property "com.franz.agraph.http.AGProtocol.overrideContext" (the name is also accessible via AGProtocol.PROP_OVERRIDE_CONTEXT) can be used to enable context overriding behaviour according to RDF4J API specification. The server-side support is only available since v7.0.0.

The AGServer.changeUserPassword() method was previously unimplemented. It is now supported.

AGRepositoryConnection.hasStatement() is significantly faster now, because the counting is moved to the server side and does not require parser initialization.

The RDF4J dependency has been upgraded from 2.2.2 to 2.4.0. RDF4J versions prior to 2.4.0 have a moderate severity security vulnerability (CVE-2018-1000644).

AGRepositoryConnection.hasStatement() is now much more efficient.

nD Geospatial Automation can now be enabled, disabled and queried.

New methods:

• AGRepositoryConnection.getNDGeospatialDatatypeAutomation
• AGRepositoryConnection.enableNDGeospatialDatatypeAutomation
• AGRepositoryConnection.disableNDGeospatialDatatypeAutomation

Users must now choose their preferred logging framework as described here: http://docs.rdf4j.org/programming/#_logging_slf4j_initialization

It is now possible to execute the 'warmup' operation through the Java client. This will cause the server to read some internal structures of a repository into memory, thus speeding up processing of future requests in that repository.

Warmup can be executed in two ways:

• by calling the warmup() method of a connection object.
• by creating a connection pool with the WARMUP parameter set to true.

HTTP connections used to not be cleaned up properly if the server returned an error in response to a query. Connections used by Jena select queries were also leaked. This resulted in suboptimal resource usage.

Changed preferred TupleQueryResultFormat for tuple queries to be TSV when possible (as long as the server version is 6.4.2 or higher). Also implemented results streaming for TSV and made the results streaming by default.

All of the above is expected to boost the performance of tuple queries.

The AllegroGraph Java client can now operate as a JTA (Java Transaction API) resource manager. This was specifically tested against Atomikos Transaction Essentials Transaction Manager. v4.0.6.

2PC operation requires AllegroGraph server version 6.5.0 or better.

See the AllegroGraph Two Phase Commit (2PC) documentation for details.

AGVirtualRepository.federate() uses getSpec() on each supplied AGRepository to construct the triple store spec to use for the federation. getSpec() returns something like <reponame> or <catname:reponame>. This is missing information that is necessary to federate against a repo that is not local to the AG server performing the federation.

Missing info: scheme, host, port, user, password.

This has been fixed.

In case a GET request (e.g. a repository query) fails due to a connection problem, it will now be automatically retried once. In particular this covers "connection reset" socket errors and HTTP 408 "Request Timeout" server responses. For now this is limited to GET requests only.

New property:

• com.franz.agraph.http.numRetries (default: 1)

transactionLatencyTimeout has been added to the TransactionSettings object. This setting determines how long a commit operation should wait for the transaction latency count to be satisfied before throwing an error.

Connection pooling now uses Commons Pool 2.4.2 instead of 1.6. The only change to the public interface of AG connection pools is that closing a pool no longer shuts down connections that have been borrowed from a pool, but not returned. Such connections will now be closed as soon as they are returned to the pool.

Inside a transaction add() statements can optionally be buffered, then sent to the backend in batches to improve efficiency.

New methods:

• AGRepositoryConnection.setAddStatementBufferEnabled
• AGRepositoryConnection.setAddStatementBufferMaxSize

New properties:

• com.franz.agraph.repository.AGRepositoryConnection.useAddStatementBuffer
• com.franz.agraph.repository.AGRepositoryConnection.addStatementBufferMaxSize

It is now possible to specify distributed transaction behavior by:

• using the new setTransactionSettings method of the connection object

• passing a TransactionSettings instance to the commit method.

• Wrapping code in a try-with-resources statement using the object returned by the transactionSettingsCtx method of the connection object.

Note that use of these settings will only have effect when using multi-master replication on AllegroGraph 7 (which is not yet released). If an older version of AllegroGraph is used, or if you're connected to a repository that is not a member of a replica set, the settings will be ignored.

The Sesame version used by the Java client has been updated from Sesame 2.9.0 to RDF4J 2.2.2. Since the Java client is an extension of the Sesame (now RDF4J) API, users should follow the migration guide at http://docs.rdf4j.org/migration/

Performance of AGGraph.clear() and AGGraph.remove() operations has been significantly improved.

The Jena client is now based on Jena 3.3.0 instead of Jena 2. Users should follow the migration guide at https://jena.apache.org/documentation/migrate_jena2_jena3.html

Various resource classes, such as AGServer and AGRepositoryConnection, now implement the AutoCloseable interface introduced in Java 7. This makes it possible to use instances of these classes in try-with-resources statements:

try (AGServer server = ...; AGRepository repo = ...; AGRepositoryConnection conn = ...; TupleQueryResult result = ...) { while (result.hasNext()) { ... } }

The com.franz.util.Closeable interface has been removed, since it is identical to AutoCloseable.

The Sesame library that the client is based on has been updated to version 2.9.0. There have been no changes to the client's API.

A new tutorial program has been added and can be found in the tutorials directory. This program can be used to send SPARQL queries to the server and output results in any format supported by Sesame.

No changes

Depending on the connection pool configuration, many unnecessary requests could be made to the AllegroGraph server when borrowing/returning connections to/from a connection pool.

These calls have been cleaned up, resulting in improved connection pool performance.

The Java client now contains methods to download and save query results to a file. The new methods are:

• AGRepositoryConnection.downloadStatements
• AGRepositoryConnection.streamStatements
• AGQuery.download
• AGQuery.stream

The methods have multiple overloads, allowing for the output path and desired output format to be passed in a variety of ways.

• Queries are now sent in request bodies instead of query strings. This results in faster processing on the server.
• bug24648:inferredGraph parameter missing from AGMaterializer:

When materializing triples through the Java API it is now possible to specify the graph into which the generated triples will be added.

No changes

Tutorial updates