Implement locRootPath

[annasong20]

This PR

• implements locRootPath, which calculates the localized path for remote roots
• changes the behavior of locFilePath to include userinfo and port in the path, and adds tests

We include userinfo and port now because otherwise we'd need url.Parse to remove them in locRootPath. Also, as discussed with @natasha41575, because the file scheme contains absolute paths on the local file system instead of url authorities, we put them inside their own sub-directory inside the localize directory.

[annasong20]

/cc @natasha41575

[natasha41575]

Question, what value does including userinfo in the resulting directory structure provide? To me it seems unnecessary to keep it, but I'd love to hear your opinion. How difficult is it to remove with url.Parse?

[annasong20]

Question, what value does including userinfo in the resulting directory structure provide? To me it seems unnecessary to keep it, but I'd love to hear your opinion. How difficult is it to remove with url.Parse?

Different userinfo can technically lead to different remote content. I'm not 100% certain, but I believe the idea is that userinfo indicates the user authenticating the domain, and each user can see a different view of the domain (edge case). For example, if the git repo path contains the working directory, i.e. ssh://user@host.com/~/repo.git, different users will have different working directories with different repos under path ~/repo.

My initial motivation for including userinfo in this PR, despite us having decided to remove it for simplicity and consolidation of https://github.com, ssh://git@github.com, and git@github.com urls, was to avoid parsing it out from remote roots. (I realized my previous approach of blindly parsing based on @ and : characters was flawed.) However, a valid workaround is to use url.Parse to parse userinfo and port for the remote roots too, so we could totally remove userinfo and port if we wanted to.

userinfo is an edge case that shouldn't be very common; we just have to decide whether we want to support it.

[natasha41575]

I'd prefer for now to stick to the KEP and exclude port and userinfo to keep the directory looking simple. It will be easier to add extra support for these edge cases later than it would be to remove them.

[k8s-ci-robot]

@annasong20: This PR has multiple commits, and the default merge method is: merge.
You can request commits to be squashed using the label: tide/merge-method-squash

Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes/test-infra repository.

[annasong20]

/label tide/merge-method-squash

[yuwenma]

PR looks great on my side! Since we have a lot of discussion about this repoSpec behaviors, I would really prefer another eye on the review for the parseHost method. 

[natasha41575]

I'm not that familiar with URL parsing, so I have some questions that may help me understand the goals of this code:

1. Does this ssh you're adding show up in the output of localize?
2. Is this ssh a placeholder just to avoid url.Parse errors, or are we intentionally adding the ssh because we expect by default schemes to be ssh://? If it's the first, I wonder if there is some other placeholder we can use that isn't actually a real scheme to make it extra clear.
3. What are some cases that would hit the default case? Seems like we would hit it for both empty schemes and for other schemes that we don't know about. For the latter (other schemes we don't know about), it seems weird to prepend ssh://. Was that your intention?

[annasong20]

1. No, we strip the scheme when we call url.Hostname: https://github.com/kubernetes-sigs/kustomize/pull/4909/files/b235147079b094ea1dd8c4ebbb1863c2a2c7dfe6#diff-f52a1bf07c678854203647b33ccaa09d945af03c9233b3dec3a947ba27d531deR212
2. Yes, it's just to avoid url.Parse errors. However, the correct scheme to add here is technically ssh. See my response to 3.
3. In this function, git.NewRepoSpecFromURL must have been called successfully. Given this, the only url that hits the default case is the relative ssh case that begins with git@. Relative ssh urls have a different delimiter than full ssh urls (ones with ssh:// header), but we remove this distinction here: https://github.com/kubernetes-sigs/kustomize/pull/4909/files/b235147079b094ea1dd8c4ebbb1863c2a2c7dfe6#diff-f52a1bf07c678854203647b33ccaa09d945af03c9233b3dec3a947ba27d531deR206. Thus, we can safely add the ssh:// scheme header to urls under the default case.

[natasha41575]

/lgtm

[k8s-ci-robot]

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: annasong20, natasha41575, yuwenma

The full list of commands accepted by this bot can be found here.

The pull request process is described here

Needs approval from an approver in each of these files:

• OWNERS [natasha41575]

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment