check for dead links?

[hansvancalster]

The tutorials website contains a lot of hyperlinks to websites. Would it be worthwhile/possible to check for dead links during continuous integration?

[florisvdh]

Great idea @hansvancalster 👍 Who knows how to do this?

[niconoe]

I'll try to find a solution for that!

[niconoe]

Update:

I'm working on this and found a nice tool to do that: https://github.com/stevenvachon/broken-link-checker. Apparently there's already a GitHub action to easily run it: https://github.com/marketplace/actions/broken-link-check

There are a still few things that bother me, to think about:

• The GitHub Action takes an URL, which means we can only run it after the site have been deployed to GitHub pages (at the end of the process: no pre-publish check, no check on PR, ...)
• That tool is very verbose (list all the links, broken or not), and no super easy solution exist for it yet (PR ignored by the maintainer)
• Given the number of links to check, the tool takes quite long to run (14 minutes for a test run on my machine)

to be continued soon!

[hansvancalster]

@niconoe, maybe https://httr.r-lib.org/reference/http_error.html can form the basis of a lightweight check?

[niconoe]

@hansvancalster: thanks! Indeed, we could implement out own checks using R (or other) HTTP libraries. I think I'm however more inclined to use an existing tool that should already handle all the gory details (redirections, possibility to ignore links ignored by robots, make the distinction between internal and external links, ...)

[damianooldoni]

I can try to check this in 2023 based my experience with GitHub Actions.

[damianooldoni]

A small update: the world of GitHub Acions is rapidly evolving and so I founded an action which seems more promising as it has already implemented the functionality of adding issues for any broken link found:
https://github.com/marketplace/actions/broken-link-checker-action

[damianooldoni]

Update part 2: the GitHub Action I mentioned above is however archived and read only. That means no maintenance will follow, so I think it's not the best idea to proceed with that.

I found this other actively maintained GitHub Action which produce reports and will fail if broken links are present. No issues are created, which is maybe even better: I don't want this repo flooded with issues 😄 a failure action (red cross) is enough.

https://github.com/marketplace/actions/broken-web-link-checker

[damianooldoni]

GitHub Action implemented on my dummy test-workflows repo, see (yml file), but I got an error: package aurora is not found. issue reported to the author, see elliotforbes/broken-link-checker#2.