OpenGrok: Oracle's source browser that indexes version history too

A source browser born at Sun in 2006

OpenGrok is a fast and usable source code search and cross reference engine, written in Java, that helps you search, cross-reference and navigate your source tree, with the distinctive capability that it can understand various program file formats and version control histories of many source code management systems, so blame, history and cross references arrive in the same interface as search. The lineage is documented rather than guessed, the project was originally conceived in Sun Microsystems by Chandan B.N., and the copyright now runs to Oracle and its affiliates from 2006 through 2025, making this one of the longest continuously maintained open source developer tools in corporate hands. The official page sits at oracle.github.io/opengrok, with the release line 1.14 moving steadily, 1.14.17 on 2026-08-24, 1.14.18 on 2026-09-02 and 1.14.19 on 2026-09-14.

Semver with an operational gloss

OpenGrok uses semantic versioning, but the README assigns each component an operational meaning beyond compatibility, a table worth pinning above a deployment. A major version change is a major backwards incompatible update requiring a full reindex as well as configuration changes. A minor change needs a full clean reindex of your repositories, the example reason given being that the index format has changed. A micro change requires only redeploying the web application. The closing rule is the one that bites, generally it is possible to go backward only within the micro version, so a minor upgrade is effectively a commitment, and the version number you choose carries a migration plan with it rather than just a changelog. Reading the three rows together yields a deployment strategy, stay on a micro train for stability, schedule minor upgrades as maintenance windows with reindex time budgeted, and treat major jumps as projects with configuration migration work attached, because the index, not the binary, is the expensive artifact to rebuild on a large source tree.

Installation lives on the wiki

The README's installation section is a pointer, see the How to setup OpenGrok wiki page, with developer information likewise routed to wiki pages, and this is a deliberate division, the setup procedure is long enough to be versioned and updated independently of the release. For the container path, the Dockerfile builds in two stages, an ubuntu:jammy builder installing OpenJDK 21 and Python 3 with a Maven world cached in an early layer so incremental builds resume after the pom, a dummy build priming dependencies before the real package step, and a distribution tar.gz lifted from the target directory. The runtime stage is tomcat:10.1.59-jdk21, so the web application deploys into Tomcat with the indexer tooling beside it, the complete search service in one image. The caching strategy in the builder stage is worth copying for any Maven project of this size, copying only the pom files and module descriptors into the first layer and running a dummy package means a one line source change does not redownload the dependency graph, which is the difference between container rebuilds measured in minutes and ones measured in tens of minutes.

The indexer, the web app, and the suggester

The Maven module structure names the architecture, opengrok-indexer as the indexing engine, opengrok-web as the search interface deployed to the servlet container, plugins for extension, suggester for query completion, tools for command line helpers, and distribution for packaging. An openapi.yaml at the root documents the REST API surface, the programmatic side that search integrations and the vscode-style clients consume. Language support is visible in the repository's license files, separate LICENSE texts for adaic.org, eclipse, ecma, mandoc, ruby and scala analyzers, a small archaeology of the file formats the engine understands, each analyzer carrying its own licensing provenance beside the main CDDL-1.0 license the badge references. The suggester module powers the query completion in the web interface, drawing on the indexed identifiers so that partial queries propose real symbols from the codebase, and the plugins module provides the extension surface for site specific analyzers and behavior without forking the core.

Oracle's process, applied to an open source project

The repository carries the full corporate open source apparatus, a CONTRIBUTING.md gating pull requests, a SECURITY.md describing a responsible security vulnerability disclosure process, GitHub Discussions as the contact channel, and SonarCloud badges watching code quality on new code. The Dockerfile itself carries hadolint suppression comments, a linting discipline for the container build, and per-file copyright headers reaching 2026. This is what a two decade old project looks like under sustained corporate maintenance, process heavy, quietly updated, with the master branch pushed 2026-09-21 and JetBrains noted among the powered by credits for development tooling. For teams evaluating self-hosted code search, the age cuts both ways, the feature set is settled and the failure modes are known, while the interface reflects its long history rather than current web fashion, a trade every buyer weighs against the privacy and scale guarantees a self-hosted engine provides.

Editorial conclusion

Run OpenGrok when a team needs private, self-hosted code search across large trees and their version control history, the gap hosted code hosts leave when source cannot leave the building. Note the operational contract before deploying, minor version upgrades require a full clean reindex, major ones also change configuration, and only micro versions allow backward movement, so pin deliberately. Installation instructions live on the wiki rather than the README, the current release line 1.14 moves about weekly, and the container route wraps the indexer and web application in one Tomcat image built from an Ubuntu JDK 21 builder.

Frequently asked questions

Official sources

1. Issues
2. oracle/opengrok on GitHub
3. Project website
4. README
5. Releases