Is there a way to github search 'generated' code?

[zeroflucs-adam-bishop]

Select Topic Area

Question

Body

We have 600+ repos for our organisation.
We have created a bunch of code generators to generate code
The code generators output a comment that includes a version
When we update the version of a generator we used to use github to search for the old version and upgrade the projects with the latest generator.
Recently we corrected the version comment in our generator to output it in the correct go format to indicate that a file is generated.
// Code generated by mygenerator 0.0.2-b DO NOT EDIT
Github now treats these files as generated no longer indexes them for search, and so we can't find stale versions via github search.
Is there a way to flag for our organisation to get generated files indexed?
Is there some other smart way of checking for stale versions across an organisation with large number of repos?
My fallback plan is to have all 600+ repos cloned locally and have them automatically pulled and do a file based search.

[shamnxd]

Short answer: No — GitHub does not currently allow searching code that it classifies as “generated”, and this behavior cannot be overridden at the organization or repository level.

Why this happens

GitHub automatically detects generated files using well-known comment patterns such as:

// Code generated by mygenerator 0.0.2-b DO NOT EDIT

Once a file is classified as generated, GitHub:

• Excludes it from code search indexing
• Excludes it from code navigation
• Does this globally and automatically

This is intentional and there is no opt-in flag to force indexing of generated files.

---

Is there a way to flag generated files to still be indexed?

❌ No.
GitHub does not provide:

• An org-level setting
• A repo-level setting
• A .gitattributes option
• A search qualifier

to re-include generated files in search.

---

Practical Workarounds

1. Add a secondary non-standard marker

Keep the standard generated comment, but add an extra searchable comment:

// Code generated by mygenerator 0.0.2-b DO NOT EDIT
// generator-version: [email protected]

You can later search for:

generator-version: [email protected]

⚠️ This may still fail if GitHub ignores the entire file, but sometimes metadata comments outside the canonical header help.

---

2. Emit a machine-readable manifest

Have each repo generate a file like:

.generator-version

or

generator:
  name: mygenerator
  version: 0.0.2-b

These files are not generated code, so they remain searchable.

---

3. Use GitHub APIs instead of search

Use:

• GitHub REST API
• GitHub GraphQL API
• GitHub Code Search API (v2)

to iterate repositories and inspect files directly.
This bypasses the search index limitation.

---

4. Local or CI-based scanning (your fallback plan)

Your fallback is actually the most reliable approach:

• Clone all repos (or shallow clone)
• Run ripgrep / grep
• Detect stale generator versions
• Automate this via CI or scheduled jobs

This is how many large orgs handle this exact problem.

---

Recommendation

For long-term maintainability across 600+ repos:

• Keep the standard Code generated by … DO NOT EDIT comment
• Add a separate version metadata file or comment
• Automate detection via CI or GitHub API scanning

---

[zeroflucs-adam-bishop]

Thank you Shamnxd for the very comprehensive answer.
The manifest is an interesting idea. I think I will take this approach. I just need to figure out if the risk of devs checking in some generated files is real (e.g. check in the updated manifest but the generated files aren't all checked in)