Why extra API call?

[ajmalab]

Hello all,

Been testing out this library and noticed that we require more API calls than necessary with the approach taken in this library. For eg:, if I'd like to post to the dispatch endpoint, I'd have to do githubClient.getRepository("my-repo").dispatch(). The getRepository() makes an additional GET call, instead of simply constructing the URL to be /repos/owner/repo/dispatches.

Curious, why was this approach taken over simply constructing the required endpoint (considering the restrictive API limits for Github)? Is there any caching being done to avoid multiple getRepository calls for the same repository? Or is the consuming service expected to handle the caching? Perhaps it's somewhere in the docs, but I could not find it.

[bitwiseman]

Not sure:

github-api/src/main/java/org/kohsuke/github/GitHub.java

Lines 682 to 687 in 2e6dcc1

	public GHRepository getRepository(String name) throws IOException {
	String[] tokens = name.split("/");
	if (tokens.length != 2) {
	throw new IllegalArgumentException("Repository name must be in format owner/repo");
	}
	return GHRepository.read(this, tokens[0], tokens[1]);

Not sure how this is working in your code given the above...

[ajmalab]

Hi, not sure I understand what you mean. If I understand that code correctly, getRepository() makes an API call?

[bitwiseman]

GHRepository.read(this, tokens[0], tokens[1]); = one api call.

[ajmalab]

yes, but to do a dispatch() call, I have to do GitHub.getRepository().dispatch(), which is 2 API calls. I do not use the response from getRepository() at all, so could save a round-trip if I can avoid this. At least this is my understanding, could not find a way to call dispatch() without calling getRepository().

[bitwiseman]

Yes, that is an potential enhancement. Please feel free to file an issue for this.