LSIF quickstart guide

This quickstart guide will walk you through installing and generating LSIF data locally on your machine, and then manually uploading the LSIF data to your Sourcegraph instance for your repository. This will let you experiment with the process locally, and test your generated LSIF data on your repository before changing your CI process.

1. Set up your environment

  1. Install the Sourcegraph CLI (src) - used for uploading LSIF data to your Sourcegraph instance.
  2. Install the LSIF indexer for your repository’s language:
    1. Go to https://lsif.dev
    2. Find the LSIF indexer for your language
    3. Install the indexer as a command-line tool using the installation instructions in the indexer’s README

What is an LSIF indexer?

An LSIF indexer is a command line tool that analyzes your project’s source code and generates a file in LSIF format containing all the definitions, references, and hover documentation in your project. That LSIF file is later uploaded to Sourcegraph to provide code intelligence.

2. Generate the LSIF data

You now need to generate the LSIF data for your repository. Each language’s LSIF is unique to that language, so run the command in the generate LSIF data step found in the README of the installed indexer.

3. Upload the LSIF data

For all languages, the upload step is the same. Make sure the current working directory is somewhere inside your repository, then use the Sourcegraph CLI to run:

$ src \
  -endpoint=https://sourcegraph.example.com \
  lsif upload \
  -file=<LSIF file (e.g. ./cmd/dump.lsif)>

The upload command in the Sourcegraph CLI will try to infer the repository and git commit by invoking git commands on your local clone. If git is not installed, is older than version 2.7.0, or you are running on code outside of a git clone, you will need to also specify the -repo and -commit flags explicitly.

If successful, you’ll see the following message:

Repository: <location of repository>
Commit: <40-character commit associated with this LSIF upload>
File: <LSIF data file>
Root: <subdirectory in the repository where this LSIF dump was generated>

LSIF dump successfully uploaded for processing.
View processing status at <link to your Sourcegraph instance LSIF status>.

Possible upload errors include:

  • Unknown repository (404): check your -endpoint and make sure you can view the repository on your Sourcegraph instance
  • Invalid commit (404): try visiting the repository at that commit on your Sourcegraph instance to trigger an update
  • Invalid auth when using Sourcegraph.com or when lsifEnforceAuth is true (401 for an invalid token or 404 if the repository cannot be found on GitHub.com): make sure your GitHub token is valid and that the repository is correct
  • Unexpected errors (500s): file an issue

LSIF processing failures for a repository are listed in Repository settings > Code intelligence > Activity for this repository. Failures can occur if the LSIF data is invalid (e.g., malformed indexer output), or problems were encountered during processing (e.g., system-level bug, flaky connections, etc). Try again or file an issue if the problem persists.

Proving ownership of a GitHub repository

If you’re uploading to Sourcegraph.com or another Sourcegraph instance with lsifEnforceAuth enabled, you must prove that you own the GitHub repository in order to upload LSIF data for it. To do so, authenticate your upload by passing a GitHub App installation access token or a GitHub access token with public_repo scope as -github-token=abc.... You can create one at https://github.com/settings/tokens.

4. Test out code intelligence

Make sure you have enabled LSIF code intelligence on your Sourcegraph instance.

Once the LSIF data is uploaded, navigate to a code file for the targeted language in the repository (or sub-directory) the LSIF index was generated from. LSIF data should now be the source of hover-tooltips, definitions, and references for that file.

To verify that LSIF is correctly enabled, hover over a symbol and ensure that its hover text is not decorated with a tooltip tooltip. This icon indicates the results are from search-based basic code intelligence and should be absent when results are precise (from an LSIF index or a Language Server).

5. Productionize the process

Now that you’re happy with the code intelligence on your repository, you need to make sure that is stays up to date with your repository. This can be done by periodically generating LSIF data, and pushing it to Sourcegraph. You can either add a step to your CI, or run it as a GitHub Action.