code-exploration

Compare original and translation side by side

🇺🇸

Original

English
🇨🇳

Translation

Chinese

Cross-Repository Code Exploration via Glean

借助Glean进行跨仓库代码探索

When users need to understand code across internal repositories—beyond the local codebase—use Glean's code search to explore the entire organization's code.
当用户需要了解跨内部仓库的代码(而非仅本地代码库)时,可使用Glean的代码搜索功能探索整个组织的代码。

Tool Naming

工具命名

See the
glean-tools-guide
skill for Glean MCP tool naming conventions. Tools follow the pattern
mcp__glean_[server-name]__[tool]
where the server name is dynamic.
请查看
glean-tools-guide
技能了解Glean MCP工具的命名规范。工具遵循
mcp__glean_[server-name]__[tool]
的命名模式,其中服务器名称是动态的。

When This Applies

适用场景

Use Glean code search when users ask about:
  • How something is implemented (in other repos)
  • Where the code for a system/service lives
  • Who has been working on a codebase
  • Similar implementations across the org
  • Examples of how to use an internal API/library
  • Code patterns used by other teams
当用户询问以下问题时,使用Glean代码搜索:
  • 某功能在其他仓库中的实现方式
  • 某系统/服务的代码存放位置
  • 谁在维护某代码库
  • 组织内的类似实现案例
  • 内部API/库的使用示例
  • 其他团队采用的代码模式

BE SKEPTICAL

保持审慎

Not every code result is worth presenting.
Quality Test
  • Is this good code to reference?
  • ✅ GOOD: Clean, tested, actively maintained
  • ⚠️ ACCEPTABLE: Works but has caveats
  • ❌ POOR: Hacky, deprecated, abandoned
Recency Test
  • Is this code maintained?
  • ✅ ACTIVE: Commits in past 3 months
  • ⚠️ SLOWING: 3-12 months since last commit
  • ❌ STALE: 12+ months - likely outdated patterns
Relevance Test
  • Does this actually answer the question?
  • ✅ RELEVANT: Directly addresses what was asked
  • ⚠️ RELATED: Similar but different context
  • ❌ TANGENTIAL: Keyword match only
Filter Out:
  • Code in
    /deprecated/
    ,
    /old/
    ,
    /legacy/
    paths
  • Abandoned repositories
  • Prototype/experimental code
  • Code with extensive TODO/FIXME comments
Quality over quantity: 3 good examples beat 10 mediocre ones.
并非所有代码结果都值得展示。
质量测试
  • 该代码是否适合作为参考?
  • ✅ 优质:代码简洁、经过测试、持续维护
  • ⚠️ 可接受:能正常运行但存在局限性
  • ❌ 劣质:代码粗糙、已废弃、无人维护
时效性测试
  • 该代码是否仍在维护?
  • ✅ 活跃:近3个月内有提交记录
  • ⚠️ 维护放缓:上次提交距今3-12个月
  • ❌ 过时:上次提交距今12个月以上——代码模式可能已过时
相关性测试
  • 该结果是否能真正回答问题?
  • ✅ 相关:直接回应用户问题
  • ⚠️ 关联:内容相似但场景不同
  • ❌ 无关:仅关键词匹配
过滤内容
  • 位于
    /deprecated/
    /old/
    /legacy/
    路径下的代码
  • 已废弃的仓库
  • 原型/实验性代码
  • 包含大量TODO/FIXME注释的代码
质量优先:3个优质示例胜过10个平庸结果。

Key Differentiator

核心优势

Local tools (grep, glob) search only the current repo. Glean searches across ALL repositories in the organization. This is powerful for:
  • Finding examples: "How do other teams handle authentication?"
  • Understanding systems: "What repos touch the billing service?"
  • Finding owners: "Who's been active in the payments codebase?"
本地工具(grep、glob)仅能搜索当前仓库,而Glean可搜索组织内的所有仓库。这一功能在以下场景中极具价值:
  • 查找示例:“其他团队是如何处理认证的?”
  • 了解系统:“哪些仓库涉及计费服务?”
  • 查找负责人:“谁在活跃维护支付代码库?”

Tool Selection

工具选择

User IntentGlean Tool
Find code by content, pattern, or file
code_search
Find related design docs or specs
search
Identify code owners/contributors
code_search
+
employee_search
Read full file content
read_document
用户意图Glean工具
根据内容、模式或文件查找代码
code_search
查找相关设计文档或规范
search
识别代码所有者/贡献者
code_search
+
employee_search
查看完整文件内容
read_document

Query Patterns

查询模式

Glean's code search understands natural language. Use filters for precision:
undefined
Glean的代码搜索支持自然语言查询,可使用过滤器提高精准度:
undefined

Search by content

按内容搜索

code_search "authentication middleware" code_search "rate limiting implementation"
code_search "authentication middleware" code_search "rate limiting implementation"

Search by contributor

按贡献者搜索

code_search "owner:"John Smith" billing service" code_search "from:me updated:past_week"
code_search "owner:"John Smith" billing service" code_search "from:me updated:past_week"

Search by time

按时间搜索

code_search "after:2024-01-01 payments API"
code_search "after:2024-01-01 payments API"

Search by file pattern

按文件模式搜索

code_search "*.proto user service"
undefined
code_search "*.proto user service"
undefined

Workflow: Exploring a System

系统探索流程

  1. Find the code:
    code_search "[system name]"
  2. Vet results: Filter for quality and recency
  3. Find the docs:
    search "[system name] design doc OR architecture"
  4. Find the people:
    code_search "owner:* [system] updated:past_month"
  5. Read details:
    read_document
    with URLs from vetted results
  1. 查找代码
    code_search "[系统名称]"
  2. 筛选结果:根据质量和时效性过滤结果
  3. 查找文档
    search "[系统名称] design doc OR architecture"
  4. 查找负责人
    code_search "owner:* [系统名称] updated:past_month"
  5. 查看详情:使用经筛选结果中的URL调用
    read_document
    查看完整内容

If No Good Code Found

未找到优质代码时的处理

Don't pad with weak results:
markdown
No high-quality code examples found for [topic].

**What was searched:**
- [Queries attempted]

**What was filtered:**
- [X] matches - [reasons: outdated/poor quality]

**Suggestions:**
- Check external libraries
- Ask in [relevant channel]
- This may need to be built from scratch
请勿用劣质结果填充内容:
markdown
未找到与[主题]相关的高质量代码示例。

**搜索内容:**
- [尝试的查询语句]

**过滤原因:**
- [X] 匹配结果 - [原因:过时/质量低下]

**建议:**
- 查看外部库
- 在[相关频道]中提问
- 该功能可能需要从零开始构建

Relationship to Commands

与命令的关联

For structured workflows, suggest the relevant slash command:
  • /glean-code:codebase-context [system]
    - Get comprehensive context
  • /glean-code:find-examples [API/pattern]
    - Find usage examples
  • /glean-code:code-owners [component]
    - Identify maintainers
  • /glean-code:similar-code [pattern]
    - Find similar implementations
对于结构化工作流,可推荐使用对应的斜杠命令:
  • /glean-code:codebase-context [system]
    - 获取全面的代码库上下文
  • /glean-code:find-examples [API/pattern]
    - 查找使用示例
  • /glean-code:code-owners [component]
    - 识别维护者
  • /glean-code:similar-code [pattern]
    - 查找类似实现