Refactor/simplify internal link check resolution #795
Conversation
* They are needed in `URL` methods to check internal links, and this way we avoid having to set them as `Runner` attributes while looping over collected links. * This also allows a more consistent definition and usage of internal links metadata.
* So we can extract `line` and `content` from there instead of explicitly passing them every time.
* Example code as in the `check_spec.rb`.
* Using `rubocop-md`
* Most of the logic is legacy from the very early HTMLProofer, and is irrelevant not since `@filename` can only be the file where the link is defined, with `File.join` properly handling both same-directory, nested, and parent links, together with the ultimate `File.expand_path` in `absolute_path. * The legacy logic comes pretty much from gjtorikian#6 and gjtorikian#23. * This way we can also avoid `File.exist`, which can ultimately be delegated to checking the existence, not constructing the path.
* We now construct and maintain a hash of resolved paths, as a way to have a single instance of going through OS for checking the existence of alternative resolved paths, including assumed extension and directory index file.
|
The additional commit d52d828 above provides a larger refactoring aiming at a single-point interaction with the OS for resolving and checking existence of the same |
* To not clash / be confusing with `resolved_path`.
|
@riccardoporreca Would you like to have push rights to this repository? I fear I can no longer devote time to it and your work has been exemplery. I glanced through this PR and the only thing that matters to me is that the tests pass. As it does that, this looks great. |
I don't mind to keep going via PRs from my fork and would in any case go through the PR review process. However I understand your limited availability for this project might benefit from having a second, more active person with push rights and would not mind that. Up to you! |
|
I sent you a collab invitation. Thank you again very much for this PR. |
Following some deep dive into the internal link check resolution from #792 (comment), I am proposing here a improved and simplified approach that also removes some legacy historical logic.
This also includes:
sourceandfilenameinformation asURLattributes, which is more natural given the role they play in checking internal links, compared toRunnerattributes that need to be consistently set and complicate code understanding/maintenance.More details can be found in individual commit for better assessment of individual changes, with commit messages providing more information.