Allow running in place on an existing checkout #53
Conversation
| default: ".github/workflows" | ||
| default: '[".github/workflows"]' |
There was a problem hiding this comment.
Hey @segiddins thank you for your contributio!
I'm not an expert here, but this looks like a breaking change, how would this impact existing users of the action?
cc @evankanderson
There was a problem hiding this comment.
I like the idea. A couple requests:
- Rather than a true/false value, let's make the in-place enablement use a path, where the empty value uses the current behavior.
- It seems like the in-place behavior implicitly disables the PR path today. It feels like we should make those separate -- e.g. create the
githubClientif the token is present, and pass that in to the action whether it's working on a local checkout or remote. (Unless there's a good reason to make these arguments exclusive.)
| in_place: | ||
| description: "Update the files in place" | ||
| required: false | ||
| default: "false" |
There was a problem hiding this comment.
So, I agree that the existing action is a bit weird -- what would you think about making this be a directory path argument, where the default "" value means to do the existing fetch-from-github behavior? i.e.
| in_place: | |
| description: "Update the files in place" | |
| required: false | |
| default: "false" | |
| repo_root: | |
| description: "Operate on files in the specified filesystem location. If unspecified, check out files from the current repo." | |
| required: false | |
| default: "" |
| var repo *git.Repository | ||
| var fs billy.Filesystem | ||
| var githubClient *github.Client | ||
| var token string |
There was a problem hiding this comment.
It looks like a bunch of these (particularly githubClient and token) don't get initialized on the "directory" path. Should we remove them?
| ActionsPath: os.Getenv("INPUT_ACTIONS"), | ||
| ActionsPaths: envToStrings("INPUT_ACTIONS"), |
There was a problem hiding this comment.
I'd prefer to add a second argument which covers the multiple-paths option (which could be exclusive with the single argument), OR to change valToStrings to also interpret the input foo as equivalent to ["foo"]. (To be honest, I'd rather have had comma-delimited strings over JSON elsewhere, but that ship has sailed.)
There was a problem hiding this comment.
Thinking further, I converting foo to ["foo"] is dangerous, because we might end up with ["[\"foo\",\n]" depending on how strict the JSON parser is and how lazy the user is. Maybe add a second argument for multiple GitHub Actions directories?
| @@ -53,7 +53,6 @@ type FrizbeeAction struct { | |||
| ImagesReplacer *replacer.Replacer | |||
| BFS billy.Filesystem | |||
| Repo *git.Repository | |||
| bodyBuilder *strings.Builder | |||
There was a problem hiding this comment.
Removing bodyBuilder seems like a (nice) simple refactor which doesn't change behavior. I could approve that change real fast if it wasn't entangled in the behavior change around ActionsPaths changing contents.
| @@ -1,4 +1,4 @@ | |||
| FROM golang:alpine3.19@sha256:0466223b8544fb7d4ff04748acc4d75a608234bf4e79563bff208d2060c0dd79 | |||
| FROM golang:1.23.1-alpine3.20@sha256:ac67716dd016429be8d4c2c53a248d7bcdf06d34127d3dc451bda6aa5a87bc06 | |||
There was a problem hiding this comment.
Can we include this change in another PR? I think it's a good change.
|
Hello @segiddins! I'm excited about this PR -- anything we can do to help you out in revisitng this? |
|
I'll try to circle back next week! |
Ran into this when I was setting up rubygems/rubygems.org#5019 -- I want to just run frizbee on the current checkout, so it can run on every PR