cmta new command #237
cmta new command #237
Conversation
There was a problem hiding this comment.
The only issue I see is that the name of the folder of the new repo has some extra bit appended to it.
So for example, lunatech-labs-lunatech-scala-2-to-scala3-course-20510a2 instead of lunatech-labs-lunatech-scala-2-to-scala3-course.
I also like the prefixing with the GitHub org which makes it distinct from a studentified artifact that gets/was generated via cmtc install.
The postfix string is the hash of the commit of the release... it's not added by me but by github... it's easy enough to remove and i'll get to that. I'll create a separate directory for 'cmta' courses, then, i'll have a think about what naming/location makes most sense |
Sorry for missing that in your original message! Yes, separating the main repos from the installed (studentified) courses make a lot of sense. |
- Refactor `ctma new` and `cmtc install` implementations to not utilise the Github API & access token to retrieve artefacts and [release] tags - test code needs to be refactored as it doesn't compile after the refactoring of the main code - Left a couple TODO that require follow-up
| val tag = githubProject.tag | ||
| val cloneGit = Process(Seq("git", "clone", s"[email protected]:$organisation/$project.git"), tmpDir) | ||
| val cloneGh = Process(Seq("gh", "repo", "clone", s"$organisation/$project"), tmpDir) | ||
| // TODO CHECK: I wonder if we should support cloning via HTTP... |
There was a problem hiding this comment.
I don't see why we shouldn't - i think there are many people NOT using ssh so this would be necessary for them to use CMT
There was a problem hiding this comment.
Agreed; There's corner case where the clone is done from a private repo. In that case, cloning via HTTP triggers a request from git for user credentials (via stdin/stdout). Then git reports that support for password authentication on HTTP is no longer supported. In any case, it should work without a problem for public repo's which covers a large chunk of the use cases. Also, the order in which the cloning is tried is ssh, gh (command), and http.
I'll run a few more experiments to figure out more details.
| // TODO This may be brittle; if the user doesn't have its git credentials set, we | ||
| // may want to retry using HTTP | ||
| // if it fails, we may still attempt to download the artefact |
There was a problem hiding this comment.
Yeah, this is a good point but I don't think it's something we should worry about for the 2.0 release - let's just create an issue for it and link to the issue in the TODO
There was a problem hiding this comment.
We may take another route. Instead of trying to figure out if there's an error in the template parameter, we could limit the implementation to simply try the download. We wouldn't be able to provide detailed error info in case of an error (such a non-existing release tag), but I don't think that's a big deal as the template parameter will be supplied by the course organiser, so if it's copy/pasted, there shouldn't be an issue.
If this approach is taken, the whole retrieval of the release tags would be removed from the code.
There was a problem hiding this comment.
+1 for creating an issue for changing the approach later.
Co-authored-by: Trevor Burton-McCreadie <[email protected]>
- Implement `cmtc install` from a folder - Fix errors in `cmta new` - Added checks for pre-existing targets
Testing with
cmta/run new -t lunatech-scala-2-to-scala3-courseThe
newcommand just prepares the correct options for, and delegates to, theinstallcommand...