Configure GHA to scan for spelling errors and broken links

[kaijli]

PR for spell check and link check github actions.

The link checker used performs (what I believe) is recursive checking through the compiled documents (since Eric mentions in this issue that lychee is not recursive). I found this specific checker by looking through the link checkers listed in this table generated by lychee.

The spell checker used is based off my list in the original issue. I don't remember the reasons for the order of the list, but if this does not serve us as well down the line, there are other options to look into.

Both actions are listed in one yml to be grabbed by the deployment action after the build step. I'm not sure if the move is to keep it as is, or move the build step to its own file to be called by the deploy action. I think it would be good for this check action to be run with every PR and / or deployment because it doesn't hurt.

[eecavanna]

Here's a link to the issue opened by the most recent GHA workflow run on this PR branch: #175. That issue contains a report generated by Lychee.

[eecavanna]

Last night, I neglected to explain my decision to switch to Lychee. My rationale was that we already had a working Lychee scanning/issue creation GHA setup working in a different repo (nmdc-schema), which I could—for the most part—copy/paste into this repo.

Regarding recursion

I had said in December that Lychee's link checking was not recursive. By that, I meant that Lychee does not scan "pages that it learns about by following links on pages it was told about up front." It only scans the pages it is told about up front.

That was a deal breaker for scanning a live website (where we don't have a way to specify all of its pages' URLs to Lychee up front). It isn't a deal breaker for scanning a file tree, since we do have a way to specify all of its pages' file paths to Lychee up front (we can do it with the [...]**/*.html glob).

[eecavanna]

[...] the deployment action after the build step. I'm not sure if the move is to keep it as is, or move the build step to its own file to be called by the deploy action.

The more I plan my next steps on this branch, the more I want to separate the build step and deploy steps into separate workflows. I'll do that on main now and then merge main into this branch.

[eecavanna]

Another option we have is to remove the spell checking from this PR and merge in the link checking, and then introduce spell checking as a follow-on task (some day).

[eecavanna]

The spell check is happening and its results are being formatted and posted in the "Summary" section of the run details page (it's below the link check report on that same page). Here's a screenshot showing the top part of the spell check report:

Source

Outstanding issues:

• It is only scanning 20 files
• It is flagging HTML tags as though they are misspelled English words (e.g. DOCTYPE)

[eecavanna]

Also, I don't want the link checker (or spell checker) to create Issues unless it is being run on the main branch. This is a reminder for myself (and a note to anyone else that may work on this PR before I do again).

Edit: Done!