Programmatically determine forks of a repo with at least n commits. Does not try to disambiguate repeated force pushes.

forks_n_commits.R

# Q on @GitHub: does anyone now how to easily (ie automatically) list all forks
# of a repository that have more than N commits in the fork?

library(gh) # devtools::install_github("gaborcsardi/gh")
forks_n_commits <- function(owner, repo, n = 1) {

  events <- gh("GET /networks/:owner/:repo/events", owner = owner, repo = repo, .limit = Inf)

  fork_pushes <- Filter(function(x) x$type == "PushEvent" && x$repo$name != paste0(owner, "/", repo),
    events)

  sizes <- vapply(fork_pushes, function(x) as.integer(x$payload$distinct_size), integer(1))
  repos <- vapply(fork_pushes, function(x) x$repo$name, character(1))

  res <- setNames(aggregate(sizes, by = list(repos), FUN = sum), nm = c("repo", "n"))
  res[res$n > n, ]
}

Thanks for this @jimhester

I just tried this and it seems to only returns a very small number of very recent results. For example

forks_n_commits("daattali", "beautiful-jekyll", 1)

Has 25 results which are all recent, but I know of many many older forks that have multiple commits. Am I using this wrong? If you're not sure, no worries, I don't actually expect you to solve this for me if you're busy :)

I got the following to work as I wanted:

library(gh)

# MY_GITHUB_PAC is a personal access token you generate on GitHub https://github.com/settings/tokens
# Supplying this parameter just means your daily rate limit will grow from 60 to 5000
# You can also drop that argument altogether 

forks_new_commits <- function(owner = "daattali", repo = "beautiful-jekyll") {

  # Get all forks of the repo
  all_forks <- gh("GET /repos/:owner/:repo/forks",
                  owner = owner,
                  repo = repo,
                  .limit = Inf,
                  .token = MY_GITHUB_PAC)

  cat("Looks like ", owner, "/", repo, " has ", length(all_forks), " forks\n")

  forks_info <- plyr::ldply(
    all_forks,
    function(fork) {
      fork_owner <- fork$owner$login
      fork_repo <- fork$name
      fork_full_name <- paste0(fork_owner, "/", fork_repo)
      cat(fork_full_name)

      # Get the last 50 commits on the fork
      fork_repo <- gh("GET /repos/:owner/:repo/commits",
                      owner = fork_owner,
                      repo = fork_repo,
                      .limit = 50,
                      .token = MY_GITHUB_PAC)

      # Get the number of commits not by the original owner
      num_new_commits <-
        sum(unlist(lapply(fork_repo, function(x) { x$author$login })) != owner)
      cat(" ", num_new_commits, " commits\n", sep = "")

      data.frame(fork = fork_full_name,
                 num_new_commits = num_new_commits,
                 url = fork$html_url)
    }
  )
  dplyr::arrange(forks_info, desc(num_new_commits))
}

UPDATE: This can be simplified a ton by using the "compare two commits" API, which you can use to directly get the number of commits the fork is ahead/behind the source. Example: https://api.github.com/repos/daattali/beautiful-jekyll/compare/daattali:master...alexwhan:master look at the ahead_by and behind_by return values