Skip to content

Latest commit

 

History

History
141 lines (94 loc) · 7.4 KB

goals.md

File metadata and controls

141 lines (94 loc) · 7.4 KB

Code Intelligence goals and roadmap

Goals

Deliver precise code intelligence to as many users as possible

Progress on adoption and usage is tracked in our Looker dashboard.

Problem:

  • Adoption of precise code intelligence comes with a cost. This cost often exceeds a customers's expectation of the benefits precise code intelligence would provide to their organization. We therefore have low adoption in both private instances and Sourcegraph Cloud.
  • There is no large open source corpus to demonstrate the benefit of precise code intelligence when adopted at scale.
  • Direct delivery to customers prior to having a large, working sample of repositories is difficult as we don't know what issues are caused by our tooling, and what issues are caused by non-standard build configurations.

High level outcomes:

  • CE and Sales have a significant amount of open source repos to showcase and can easily demonstrate the value of setting up precise code intelligence.
  • For languages that are particularly subject to unusual setups and tooling, our indexed open source showcase helps us determine if setup issues are caused by an unusual customer dev environment.

C++ precise code intelligence support

Milestones:

  1. ✅ Sourcegraph Cloud has up-to-date precise code intelligence for C++ repositories.

    • ✅ Increase observability in the code intel executor to make troubleshooting failed auto-index records possible.
    • ✅ Auto-index 5 repositories on Sourcegraph Cloud.

    Outcome: 5 popular open source C++ repositories have up-to-date precise code intelligence on Sourcegraph Cloud.

  2. ✅ Sourcegraph Cloud has precise code intelligence for C++ repositories that use cmake as a build system.

    • ✅ Auto-index 5 repositories on Sourcegraph Cloud that use cmake as a build system.
    • ✅ Research: Add inference rules for cmake projects to auto-index on Cloud.

    Outcome: 5 popular open source cmake repositories have up-to-date precise code intelligence on Sourcegraph Cloud.

  3. 🔄 Build lsif-clang testing infrastructure

    • 🔄 Build a tool which can to semantically compare and diff LSIF dumps.
    • 🔄 Write a test harness for lsif-clang which can perform snapshot tests on small code snippets.
    • Spike: Explore how to perform snapshot tests on large projects for integration testing.

    Outcome: We can confidently make code changes to lsif-clang without worrying about obvious bugs.

  4. Sourcegraph Cloud has precise code intelligence for C++ repositories that use Bazel as a build system.

    • Support Bazel projects in lsif-clang.
    • Auto-index 5 repositories on Sourcegraph Cloud that use Bazel as a build system.
    • Update and improve documentation to ensure a good user experience.

    Outcome: 5 popular open source C++ repositories using Bazel have up-to-date precise code intelligence on Sourcegraph Cloud.

  5. 🔄 Increase C++ precise code intel monthly operations.

    • 🔄 Deliver to N1.
    • Deliver to N2.
    • Deliver to N3.

    Outcome: 3 enterprise customers with C++ precise code intel configured in their repositories.

  6. Sourcegraph Cloud has up-to-date precise code intelligence for C++ repositories that exceed 1mil SLOC.

    • Create benchmarks of lsif-clang memory and CPU usage for various sizes of repositories.
    • Auto-index 5 repositories on Sourcegraph Cloud that exceed 1M SLOC.

    Outcome: 5 popular open source C++ repositories that exceed 1M SLOC have up-to-date precise code intelligence on Sourcegraph Cloud.

  7. Implement standard C++ code navigation features.

    • Jump to declaration.
    • Jump to implementation.
    • Switch between source and header file.

Java precise code intelligence support

Milestones:

  1. 🔄 Sourcegraph Cloud has up-to-date precise code intelligence for 5 Gradle repositories.

    • ✅ Automatic inference for projects that use the Java plugin
    • Automatic inference for projects that use the Android plugin
    • Publish Gradle plugin.
    • 🔄 Update and improve documentation to ensure a good user experience.

    Outcome: 5 popular open source Gradle repositories have up-to-date precise code intelligence on Sourcegraph Cloud.

  2. 🔄 Sourcegraph Cloud has up-to-date precise code intelligence for 5 Maven repositories.

    • ✅ Automatic inference for projects that use the Java compiler plugin
    • Automatic inference for projects that use the Scala/Kotlin plugins
    • Publish Maven plugin.
    • 🔄 Update and improve documentation to ensure a good user experience.

    Outcome: 5 popular open source Maven repositories have up-to-date precise code intelligence on Sourcegraph Cloud.

  3. Sourcegraph Cloud has precise code intelligence for Java repositories that use Bazel as a build system.

    • Manual configuration in a forked repository.
    • Bazel script implementation.
    • Update and improve documentation to ensure a good user experience.

    Outcome: 5 popular open source Java repositories using Bazel have up-to-date precise code intelligence on Sourcegraph Cloud.

  4. Increase Java precise code intel monthly operations.

    • Deliver to N4.
    • Deliver to N5.
    • Deliver to N6.

    Outcome: 3 enterprise customers with Java precise code intel configured in their repositories.

  5. lsif-java supports third party dependency navigation.

  6. ✅ lsif-java supports Java 11-16.

Backend stability

Milestones:

  1. 🔄 Stabilize code intelligence backend.

    • ✅ Rewrite resolvers so they can be optimized.
    • 🔄 Add hard limits to all database queries.
    • 🔄 Implement RFC 311 (enable out-of-band data migrations).
    • Update codeintel data schema to reduce memory pressure in query path.

    Outcome: No code intelligence queries consume compute or unbounded memory.

  2. Reduce code intelligence failure blast radius.

    • Move background processes into isolated container(s).
    • Spike: Explore how to tune Postgres for our database insertion patterns.
    • Add hard limits to code intelligence tables in shared database.

    Outcome: No code intelligence failure will affect the stability of the core search feature.

  3. Decrease backend latency and resource requirements.

Roadmap

The code intel team roadmap is tracked in productboard.

At a glance roadmap as of 2021-02-18:

2021-02-18