[FIX] Find draft GitHub releases using tag

[MattSturgeon]

Draft releases are not found at the releases/tags/<tag> endpoint, but are still associated with the tag name.

If we can't find a release at releases/tags/<tag>, then we iterate through all releases looking for one with a matching tag name.

It may be better to come up with a GraphQL query instead of iterating through all releases via the REST API?

The GH CLI implements release fetching as: https://github.com/cli/cli/blob/30066b0042d0c5928d959e288144300cb28196c9/pkg/cmd/release/shared/fetch.go#L134-L161

The draft release lookup uses a GraphQL query: https://github.com/cli/cli/blob/30066b0042d0c5928d959e288144300cb28196c9/pkg/cmd/release/shared/fetch.go#L169-L200

Currently I'm unsure of what the best way to make a GraphQL request is in kohsuke github, although I wonder if this is fine as a stopgap here and fetching draft releases should later be implemented upstream using GraphQL?

I've opened hub4j/github-api#1983 upstream.

[hypherionmc]

Looks good :)

Currently I'm unsure of what the best way to make a GraphQL request is in kohsuke github

I think there is a whole different library required to work with GraphQL in java. I could be completely wrong about this. Will have to do some research on that

EDIT: Turns out it is possible to query GraphQL with standard HTTP libraries, but there are purpose built libraries for it

[hypherionmc]

Available on our snapshots maven as version 2.1.7+snapshot.0

[MattSturgeon]

Thanks for merging and releasing so promptly 🚀

I'll try to look into getting a graphql implementation upstream, I've already documented a gql query for getting the release id in hub4j/github-api#1983.

If upstream ends up being slow or resistant to the change, then I'll look at adding a graphql implementation here instead.

If we get an upstream fix, I'll look at bumping the dependency and dropping the workaround here.

---

On a mostly unrelated note, maybe it would be better to use graphql for all GitHub requests and drop the dependency on hub4j/github-api? That way we end up only fetching the data we actually need. 🤔

[hypherionmc]

On a mostly unrelated note, maybe it would be better to use graphql for all GitHub requests and drop the dependency on hub4j/github-api? That way we end up only fetching the data we actually need. 🤔

Hmmmm. Might not be a bad idea. We don't need all the features provided by github-api, so doing GraphQL would not only result in 1 less dependency (or well, a smaller dependency size), but giving us more control over what we want and how we want it.

I actually started working on version 3.0.0 a couple of days ago, rewriting most of the code base and implementing some of the stuff we talked about, so would be a perfect time to switch over