@probablycorey I love that you're working on this 👏 I feel like the current level of boilerplate is just getting out of hand.

This seems simple, but I ran into enough problems that I want to verify the interface before I go any further. I'm proposing this...

func getPr(cmd *cobra.Command, prString string) (*api.PullRequest, error)

Knowing our specific use-cases, I feel like this interface is too simple. For example, as you said, it always results in an API request, but for some cases we don't necessarily need an API request, e.g. if we parse a PR number from a URL, we can just open that URL directly in pr view --web without the overhead of fetching information about the PR that we are not going to use in the client.

On the other hand, some use-cases require more GraphQL fields to be fetched than others. For instance, pr view (non-web) needs a lot of information about a PR, including metadata such as projects and assignees, while pr checkout basically only needs the name of the head branch and the owner/repo pair of the head repository (in case PR comes from a fork).

Finally, I think any helper function that parses PRs from URLs should also return the reference to the baseRepo in a separate argument, because for some commands such as pr checkout we need use that baseRepo value in subsequent logic.

I don't know what a correct solution should be, but it seems to me that we have different needs in each of the gh pr ... commands and that a one-size-fits-all approach would be tricky to achieve. 🤔

Oooh I just had an epiphany about requesting variable level of GraphQL information per different invocation.

How about we initialize a model in the caller and pass that to the query function to tell it what level of information we want?

prArg := args[0]
pr := api.PullRequestLight{}
err := pullRequestFromArgument(prArg, &pr)
if err != nil {
  return err
}
// now we can use all `pr.*` fields as defined by PullRequestLight

// alternatively, `pr view` might request a model with more fields
pr := api.PullRequestHeavy{} // this has many extra fields
err := pullRequestFromArgument(prArg, &pr)
// ...
// `pr.*` now has many extra fields we can peruse

This is assuming that the new queries would be backed by the githubv4 library. I'm happy to pair on this if it helps illustrate it further!

Finally, I think any helper function that parses PRs from URLs should also return the reference to the baseRepo in a separate argument, because for some commands such as pr checkout we need use that baseRepo value in subsequent logic.

One thing I don't get is why we need to get the baseRepo from a PR from a url, but not from a PR that is opened from a fork? Why don't we change the baseRepo in both instances since the baseRepo will be different in both instances?