2.7.x release notes

Updates not required.

Not required. Lingo4G 2.7.0 works with indices created by Lingo4G 2.6.x.

Updates not required. Analysis requests written for Lingo4G 2.6.x work unchanged with Lingo4G 2.7.0.

The new documents:​from​Search​Service stage fetches the set of documents to analyze from an external search engine. This release comes with the search​Service:​solr component, which retrieves search results from an Apache Solr instance and maps them to documents in the Lingo4G index.

Each Lingo4G project can now include a repository of analysis requests. Use the repository to keep, organize and share the analysis requests you develop for your project.

Sensitive component properties, such as search service credentials or connection URLs, are now masked when Lingo4G outputs the effective request or project configuration JSONs.

The dataset-wikipedia example project now reads the sharded Wikipedia dumps published at dumps.wikimedia.org/other/cirrus_search_index/ (the previous cirrussearch dumps are no longer updated). All shards found in the input directory are indexed, bzip2-compressed input is read directly and redirect-page stubs are skipped. See datasets/dataset-wikipedia/​R​E​A​D​M​E.txt for download instructions.

All internal HTTP communication now uses the HTTP client built into the Java runtime. This removes the dependency on the Apache HttpComponents library and fixes a number of corner cases in URI and query parameter handling, SSL certificate chain validation and HTTP redirect processing.

Lingo4G now fails more gracefully on startup if the JVM does not provide the jdk.incubator.vector module and includes the offending URL in exception messages thrown when remote requests fail.

When the --host option of the server command is given a host name that resolves to more than one address (for example, localhost resolving to 127.0.0.1 and :​:​1 on dual-stack systems), the server now binds to all the resolved addresses instead of just the system-preferred one. The startup log prints service URLs for each bound address.

The 2.7.0 release fixes a Null​Pointer​Exception thrown by the labels:​all stage when the list of terms to process was empty and corrects a number of warnings and build errors on Windows.

The example USPTO data set descriptor and documentation now reflect the changed layout of the USPTO bulk data download sites.

Lingo4G 2.7.0 adds support for Java 25.

Third-party dependencies have been updated, including Apache Lucene (indices created by earlier Lingo4G versions remain readable), Eclipse Jetty, the embedded HTTP server, now at version 12, and Apache Tika, used by the dataset-autoindex example project, now at version 4.