sourcegraph-playground

📚 Learning and exploring Sourcegraph.

Sourcegraph makes it easy to read, write, and fix code—even in big, complex codebases.

--https://about.sourcegraph.com

Overview

This repository is my personal playground for learning and exploring Sourcegraph. In it, I have notes, commands and references about Sourcegraph products.

My main interest in Sourcegraph is using it to search my own fleet of many small GitHub repositories. I'm especially sold on precise code navigation. I've also used Sourcegraph to wade through large open source projects and I'm optimistic about making use of features like batch changes, automated integrations to the Sourcegraph APIs (GraphQL + REST), the Sourcegraph browser extension, and now the AI-coding assistant named Cody.

---

Large Language Models entered the chat...

I have to admit... I undersold the last item in that sentence: Cody.

Operationalized "AI power" (read: LLMs) has dwarfed the value proposition of all other software tools/languages/frameworks that I had on my todo-list for learning. It is a volcano 🌋.

I need to figure out how to use AI tools like Cody and the rest of them. In broad strokes, I need to spend my effort learning how to:

• interact with AI
• instruct AI
• think about AI's capabilities/limitations
• map/reduce domain problems into small (8KB, 32KB) puzzles that can be solved by AI

And it's already been fun and fruitful to use these tools and become enthusiastic about what I know they can do.

---

My experience with Sourcegraph is described by an "Awareness, Trial, Productivity, (Repeat)" arc below.

Awareness

I heard about Sourcegraph for the first time from advertisements on the Changelog family of podcasts around 2020 (maybe earlier or later?).

Trial

I got up and running with a local instance of Sourcegraph using the official Docker containers in October 2021. While a success, running via containers was too intensive/slow on my Mac for me to have it up and running all the time.

I got up and running with the "build from source" route. I can't really remember why I didn't use this approach in the long term. Partially because of ctags problems (hard to remember)? I think the main reason was that it was just so easy/effective to use Sourcegraph.com directly because it could index my personal code.

Productivity

my own data + a good search UX = good

My usage of Sourcegraph remained pretty basic—but basics are the best 👑. I would search over my own personal (public) GitHub repositories to find snippets of code or non-code information like my extensive notes across my many small repos. I did this between November 2021 up until personal repo indexing (through the program named personal code beta) was discontinued in August 2022.

Trial Part 2: Failure to launch

After personal repo indexing was discontinued, I wanted an alternative. I'm willing to pay for something, but Sourcegraph only sold "a single-tenant, managed version of Sourcegraph" for teams (makes sense). So, the other option was to run Sourcegraph myself. I came back to this sourcegraph-playground repository to re-visit my notes and get up and running with Sourcegraph locally but I couldn't get it to work. Sourcegraph had grown more sophisticated since I was first able to build and run it. With sophistication comes complexity. I couldn't tackle the complexity. See my extensive notes in archive/old-notes.md.

Trial Part 3: "Sourcegraph App"

It's March 2023 and Sourcegraph launched a desktop app that runs a local version of Sourcegraph from a single binary. This is perfect! See the announcement: Announcing the Sourcegraph app.

I installed it and it's promising although I'm having issues with indexing my repos. My next steps are to capture the issues in detail, and make it clear how to re-produce them. Even if I'm the only audience, that's fine. The Sourcegraph team is very generous in the Sourcegraph Discord.

My working notes for Sourcegraph App are in sourcegraph-app.md.

Wish List

General clean-ups, TODOs and things I wish to implement for this project

• List out more todos. I want to try the intellij plugin, browser extension, etc. I want to use Sourcegraph as the Code Intelligence Platform it's touted as. One thing I've especially been dragging my feet on is implementing automation for markdown list format linting that plagues me everyday. I have the lint rule but never integrated it anywhere. Perfect for batch changes? Or better as a project tool, like when using Cody? I kind of really do not vibe with widespread automated pull requests (despite everyone else and their parents loving it). I still have to validate the change repo-by-repo, because I actually review/QA my software changes (and writing work changes) So I don't grok it, but I'm willing to see the other side of it.

Reference

• Sourcegraph + AI: A Steve Yegge blog post (including rant).

  • There is something legendary and historic happening in software engineering, right now as we speak, and yet most of you don’t realize at all how big it is.

  • The ultra-rare trillion-dollar money volcano

  • In one shot, ChatGPT has produced completely working code from a sloppy English description! With voice input wired up, I could have written this program by asking my computer to do it.

  • Cody is Sourcegraph’s new LLM-backed coding assistant.

• Sourcegraph Discord
• Sourcegraph App docs