Create tool for getting DCO sign off emails

This tool digs through commit messages and parses out names and email
addresses from the `Signed-off-by:` tags, collects all unique email
addresses and outputs a CSV of name/email pairs.

Signed-off-by: Andrew Ross <andrross@amazon.com>

bin/commands/contributors.rb

@@ -32,4 +32,12 @@
       end
     end
   end
+
+  g.desc 'Create name,email CSV list of all DCO signers'
+  g.command 'dco-csv' do |c|
+    c.action do |_global_options, options, _args|
+      org = GitHub::Organization.new(options)
+      puts org.commits(options).unique_dco_signers_csv
+    end
+  end
 end

lib/github/commit.rb

@@ -0,0 +1,17 @@
+# frozen_string_literal: true
+
+# frozen_string_literal: true
+
+module GitHub
+  class Commit < Item
+    # Association list of all name/email pairs extracted from the DCO sign off
+    # in the commit message
+    def dco_signoff_names_and_mails
+      commit.message.scan(/Signed-off-by: (.+) <([A-Za-z0-9._%+-]+@[A-Za-z0-9.-]+\.[A-Za-z]+)>/)
+    end
+
+    def to_s
+      "#{repository.full_name} - #{commit.author.email} - #{sha}"
+    end
+  end
+end

lib/github/commits.rb

@@ -0,0 +1,58 @@
+# frozen_string_literal: true
+
+module GitHub
+  class Commits < Items
+    def initialize(arr_or_options)
+      super arr_or_options, GitHub::Commit
+    end
+
+    def unique_dco_signers_csv
+      # Create an association list of all name->email pairs, e.g:
+      # [['John Lennon', 'john@beatles.com'], ['Paul McCartney', 'paul@beatles.com']]
+      signers_list = each.map(&:dco_signoff_names_and_mails).flatten(1)
+
+      # De-dupe all entries by email address into an email=>[name1,name2] hash
+      email_to_names = {}
+      signers_list.each do |signer|
+        name = signer[0]
+        email = signer[1]
+        email_to_names[email] = [] unless email_to_names.include?(email)
+        email_to_names[email].push(name)
+      end
+
+      email_to_names.to_a
+                    # For each email pick the best name and then reverse the association
+                    .map { |e| [single_best_name(e[1]), e[0]] }
+                    # Sort all "noreply" email addresses to the bottom (for manual curation), then sort by name
+                    .sort_by { |e| [e[1].include?('noreply') ? 1 : 0, e[0].downcase] }
+                    .map { |e| e.join(',') }
+                    .join("\n")
+    end
+
+    def page(options)
+      data = $github.search_commits(query(options), per_page: 1000).items
+      raise 'There are 1000+ commits returned from a single query, reduce --page.' if data.size >= 1000
+
+      data.reject do |commit|
+        commit.commit.author.email.include?('[bot]')
+      end
+    end
+
+    def query(options = {})
+      GitHub::Searchables.new(options).to_a.concat(
+        [
+          "committer-date:#{options[:from]}..#{options[:to]}"
+        ]
+      ).compact.join(' ')
+    end
+
+    private
+
+    # This is a simple heuristic for picking the "best" name by choosing the
+    # one with the most words. For example, if we find both "paul" and
+    # "Paul McCartney" then we'll choose "Paul McCartney".
+    def single_best_name(names)
+      names.max { |a, b| a.split.length <=> b.split.length }
+    end
+  end
+end

lib/github/organization.rb

@@ -28,6 +28,10 @@ def pull_requests(options = {})
       @pull_requests ||= GitHub::PullRequests.new({ org: name, status: :merged }.merge(options))
     end
 
+    def commits(options = {})
+      @commits ||= GitHub::Commits.new({ org: name }.merge(options))
+    end
+
     def issues(options = {})
       @issues ||= GitHub::Issues.new({ org: name }.merge(options))
     end

lib/tools.rb

@@ -19,6 +19,8 @@
 require_relative 'github/repos'
 require_relative 'github/pull_requests'
 require_relative 'github/pull_request'
+require_relative 'github/commits'
+require_relative 'github/commit'
 require_relative 'github/contributors'
 require_relative 'github/contributor'
 require_relative 'github/maintainers'