Boosting Code Review Efficiency with Alibaba Cloud’s LSIF‑Powered Syntax Service

Introduction

In code review (CR), browsing pure‑text code is time‑consuming and hurts efficiency. Alibaba Cloud’s intelligent syntax service offers a cloud‑backed fast code navigation that lets users view definitions and references without cloning repositories, greatly improving review speed and quality.

Technical Foundations

The service is built on LSIF (Language Server Index Format), a persistent graph‑based index that maps code documents to syntax results.

Before LSIF, LSP (Language Server Protocol) required separate adapters for each language‑editor pair, leading to M×N implementations. LSIF, combined with LSP, reduces this to M+N implementations.

LSIF stores pre‑computed syntax analysis in the cloud, trading space for time to serve repeated requests quickly.

Implementation

Index Construction

The service receives code platform events (push, PR creation, merge, etc.) to trigger index building. Scheduling relies on Alibaba’s open‑source distributed scheduler tbschedule with Zookeeper for task management.

Event‑driven index construction

User‑request‑driven syntax service response

For Java, the open‑source Spoon tool parses source code into an AST, captures definitions, references, and comments, and outputs a unified LSIF JSON.

// this is a sample class
public class Sample {
}

A sample LSIF JSON fragment:

{ "id": 1, "type": "vertex", "label": "document", "uri": "file:///abc/sample.java", "languageId": "java" }
{ "id": 2, "type": "vertex", "label": "range", "start": { "line": 0, "character": 13 }, "end": { "line": 0, "character": 18 } }
{ "id": 3, "type": "edge", "label": "contains", "outV": 1, "inVs": [2] }
{ "id": 4, "type": "vertex", "label": "hoverResult", "result": ["this is a sample class"] }
{ "id": 5, "type": "edge", "label": "textDocument/hover", "outV": 2, "inV": 4 }

In real projects the LSIF graph can contain hundreds of thousands of nodes.

Incremental Scheme

After each successful branch index build, the system records the branch version. On a new commit, it diffs the generated LSIF files, extracts affected files, and performs incremental Elasticsearch updates, reducing branch build time by about 45% on average.

Time‑Lock Management

Index build times range from seconds to minutes, while push events can peak at hundreds per minute. A Redis‑based distributed time‑lock ensures that only the latest push for a repository is processed, discarding stale tasks and using heartbeats to recover from failures.

Syntax Service Response

The service handles three main requests:

Retrieve all clickable symbols when a file is opened.

Obtain definition and reference lists for a clicked symbol.

Jump to the selected definition or reference.

For the first request, Elasticsearch queries filter by file path to return symbol coordinates, with front‑end pagination for large files. The second request also returns code snippets via a batch file‑segment API. The third request highlights the target line in‑page or opens a new page for cross‑file jumps. Index construction and service response are fully asynchronous and independently scalable.

Index Cleanup

Since syntax indexes can be several times larger than source files, cleanup tasks run when code reviews are merged or branches are deleted, releasing storage resources.

Future Outlook

Symbol navigation remains a pain point for web‑based code browsing. Emerging standards such as LSIF, Kythe, SARIF, UAST, Tree‑sitter, and ctags aim to improve code analysis. Alibaba Cloud’s syntax service will continue to accelerate index building, support more languages, and broaden syntax scenarios to enhance developer experience.

Related links: [1] https://microsoft.github.io/language-server-protocol/specifications/lsif/0.4.0/specification/ [2] http://spoon.gforge.inria.fr/ [3] https://github.com/INRIA/spoon/pull/3513