Exact Code Search

1 post · Mar 2019 – Mar 2026

What this trail covers

GitLab's Exact Code Search, powered by Zoekt, provides lightning-fast, precise code searching across massive codebases (48TB+). It offers Exact Match mode for zero false positives, Regular Expression mode for complex pattern matching, and multi-line results for contextual understanding. The system is built on a distributed, self-registering node architecture with intelligent load balancing and seamless integration with GitLab's permission system, enabling rapid debugging, code exploration, refactoring, and security auditing.

All GitLab trails

Mar 2026Navigate repositories faster with the file tree browserThis post introduces the file tree browser feature for GitLab repositories. It adds a collapsible, resizable panel alongside file and directory views to keep project structure visible while reading code. Key features include synchronized navigation with the current file, filename filtering via a global search dialog, keyboard-first navigation using ARIA treeview patterns, responsiveness across different viewports (desktop, smaller viewports, mobile), and pagination for large repositories. The goal is to improve developer productivity by providing an IDE-like navigation experience within the GitLab UI.
Jun 2025Exact Code Search: Find code faster across repositoriesThis post details the technical implementation of GitLab's Exact Code Search, which integrates Zoekt for fast and precise code searching. It explains Zoekt's use of positional trigrams for exact matches, its regex performance optimizations, and the distributed system architecture including self-registering nodes, dynamic shard assignment, and intelligent load balancing. It also highlights the seamless integration with GitLab's permission system and the engineering challenges overcome in building an orchestration layer for Zoekt at enterprise scale.
Nov 2019Native code intelligence is coming to GitLabThis post details the integration of Sourcegraph's code intelligence capabilities directly into the GitLab codebase. It describes the addition of features like 'go-to-definition' and 'find references' within hover tooltips in code views, file views, merge requests, and code diffs. The integration was a collaboration between GitLab and Sourcegraph engineering teams, resulting in a browser-based developer platform without requiring external plugins for GitLab.com users. The technical implementation involved changes to the GitLab codebase to support Sourcegraph's functionality.
Mar 2019Lessons from implementing global code search on GitLab.comThis post details the challenges and attempted solutions for implementing global code search on GitLab.com using Elasticsearch. It covers efforts to reduce Elasticsearch index size through force merges and adjusting nGram sizes, and explores the benefits and drawbacks of separate indexes. It also outlines improvements needed for Elasticsearch administration, including better cluster visibility, graceful recovery from indexing failures, selective/progressive indexing, making code indexing optional, and sharding Elasticsearch per group. The post indicates a shift in engineering focus from index size reduction to administration capabilities.
Advertisement