Testing: Speed up Twitter tests by paring down VCR cassette content

This commit reduces the execution time of the specs for
Channel::Driver::Twitter#fetch by a little over 50%.

The #fetch method calls out to the Twitter API
and receives a list of tweets (serialized as JSON),
then converts them to tickets and articles.
Even with HTTP caching (via VCR), this is an expensive operation:
each transaction may contain dozens of tweets to be processed.

This commit reduces the processing burden
by manually editing the tests' VCR cassettes,
removing the vast majority of incoming tweets
and keeping only those necessary to run meaningful tests.

The result is a set of tests for the #fetch method that,
on a given machine, now take 65 seconds instead of 140.
As an ancillary benefit, the file size of the associated
VCR cassette directory has been reduced from 16MB to under 5MB.

spec/models/channel/driver/twitter_spec.rb

@@ -821,7 +821,7 @@ RSpec.describe Channel::Driver::Twitter do
       context 'with search term configured (at .options[:sync][:search])' do
         it 'creates an article for each recent tweet' do
           expect { channel.fetch }
-            .to change(Ticket, :count).by(8)
+            .to change(Ticket, :count).by(2)
 
           expect(Ticket.last.attributes).to include(
             'title'       => "Come and join our team to bring Zammad even further forward!   It's gonna be ama...",
@@ -831,42 +831,140 @@ RSpec.describe Channel::Driver::Twitter do
           )
         end
 
-        it 'skips retweets' do
-          expect { channel.fetch }
-            .not_to change { Ticket.where('title LIKE ?', 'RT @%').count }.from(0)
-        end
-
-        it 'skips tweets 15+ days older than channel itself' do
-          expect { channel.fetch }
-            .not_to change { Ticket.where('title LIKE ?', 'GitHub Trending Archive, 2_ Nov 2018, Ruby. %').count }.from(0)
-        end
-
-        context 'when fetched tweets have already been imported' do
-          before do
-            tweet_ids.each { |tweet_id| create(:ticket_article, message_id: tweet_id) }
+        context 'for responses to other tweets' do
+          let(:thread) do
+            Ticket.joins(articles: :type).where(ticket_article_types: { name: 'twitter status' })
+              .group('tickets.id').having(
+                case ActiveRecord::Base.connection_config[:adapter]
+                when 'mysql2'
+                  'COUNT("ticket_articles.*") > 1'
+                when 'postgresql'
+                  'COUNT(ticket_articles.*) > 1'
+                end
+              ).first
           end
 
-          let(:tweet_ids) do
-            [1224440380881428480,
-             1224426978557800449,
-             1224427517869809666,
-             1224427776654135297,
-             1224428510225354753,
-             1223188240078909440,
-             1223273797987508227,
-             1223103807283810304,
-             1223121619561799682,
-             1222872891320143878,
-             1222881209384161283,
-             1222896407524212736,
-             1222237955588227075,
-             1222108036795334657,
-             1222126386334388225,
-             1222109934923460608]
+          it 'creates articles for parent tweets as well' do
+            channel.fetch
+
+            expect(thread.articles.last.body).to match(/zammad/i)       # search result
+            expect(thread.articles.first.body).not_to match(/zammad/i)  # parent tweet
+          end
+        end
+
+        context 'and "track_retweets" option' do
+          context 'is false (default)' do
+            it 'skips retweets' do
+              expect { channel.fetch }
+                .not_to change { Ticket.where('title LIKE ?', 'RT @%').count }.from(0)
+            end
           end
 
-          it 'does not import duplicates' do
-            expect { channel.fetch }.not_to change(Ticket::Article, :count)
+          context 'is true' do
+            subject(:channel) { create(:twitter_channel, custom_options: { sync: { track_retweets: true } }) }
+
+            it 'creates an article for each recent tweet/retweet' do
+              expect { channel.fetch }
+                .to change { Ticket.where('title LIKE ?', 'RT @%').count }.by(1)
+                .and change(Ticket, :count).by(3)
+            end
+          end
+        end
+
+        context 'and "import_older_tweets" option (legacy)' do
+          context 'is false (default)' do
+            it 'skips tweets 15+ days older than channel itself' do
+              expect { channel.fetch }
+                .not_to change { Ticket.where('title LIKE ?', 'GitHub Trending Archive, 29 Nov 2018, Ruby. %').count }.from(0)
+            end
+          end
+
+          context 'is true' do
+            subject(:channel) { create(:twitter_channel, :legacy) }
+
+            it 'creates an article for each tweet' do
+              expect { channel.fetch }
+                .to change { Ticket.where('title LIKE ?', 'GitHub Trending Archive, 29 Nov 2018, Ruby. %').count }.by(1)
+                .and change(Ticket, :count).by(3)
+            end
+          end
+        end
+
+        describe 'duplicate handling' do
+          context 'when fetched tweets have already been imported' do
+            before do
+              tweet_ids.each { |tweet_id| create(:ticket_article, message_id: tweet_id) }
+            end
+
+            let(:tweet_ids) { [1222126386334388225, 1222109934923460608] }  # rubocop:disable Style/NumericLiterals
+
+            it 'does not import duplicates' do
+              expect { channel.fetch }.not_to change(Ticket::Article, :count)
+            end
+          end
+
+          describe 'Race condition: when #fetch finds a half-processed, outgoing tweet' do
+            subject!(:channel) { create(:twitter_channel, custom_options: custom_options) }
+
+            let(:custom_options) do
+              {
+                user: {
+                  # Must match connected Twitter account's user ID (on Twitter)--
+                  # that's how #fetch distinguishes between agents' tweets and other people's.
+                  id: '1205290247124217856'
+                },
+                sync: {
+                  search: [
+                    {
+                      term:     'zammadzammadzammad',
+                      group_id: Group.first.id
+                    }
+                  ]
+                }
+              }
+            end
+
+            let!(:tweet) { create(:twitter_article, body: 'zammadzammadzammad') }
+
+            context '(i.e., after the BG job has posted the article to Twitter…' do
+              # NOTE: This context block cannot be set up programmatically.
+              # Instead, the tweet was posted, fetched, recorded into a VCR cassette,
+              # and then manually copied into the existing VCR cassette for this example.
+
+              context '…but before the BG job has "synced" article.message_id with tweet.id)' do
+                let(:twitter_job) { Delayed::Job.find_by(handler: <<~YML) }
+                  --- !ruby/object:Observer::Ticket::Article::CommunicateTwitter::BackgroundJob
+                  article_id: #{tweet.id}
+                YML
+
+                around do |example|
+                  # This test case requires the use_vcr: :time_sensitive option
+                  # to travel_to(when the VCR cassette was recorded).
+                  #
+                  # This ensures that #fetch doesn't ignore
+                  # the "older" tweets stored in the VCR cassette,
+                  # but it also freezes time,
+                  # which breaks this race condition handling logic:
+                  #
+                  #     break if Delayed::Job.where('created_at < ?', Time.current).none?
+                  #
+                  # So, we unfreeze time here.
+                  travel_back
+
+                  # Run BG job (Why not use Scheduler.worker?
+                  # It led to hangs & failures elsewhere in test suite.)
+                  Thread.new do
+                    sleep 5  # simulate other bg jobs holding up the queue
+                    twitter_job.invoke_job
+                  end.tap { example.run }.join
+                end
+
+                it 'does not import the duplicate tweet (waits up to 60s for BG job to finish)' do
+                  expect { channel.fetch }
+                    .to not_change(Ticket::Article, :count)
+                end
+              end
+            end
           end
         end
 
@@ -888,112 +986,27 @@ RSpec.describe Channel::Driver::Twitter do
 
           let(:twitter_articles) { Ticket::Article.joins(:type).where(ticket_article_types: { name: 'twitter status' }) }
 
-          it 'stops importing threads after 120 new articles' do
+          # NOTE: Ordinarily, RSpec examples should be kept as small as possible.
+          # In this case, we bundle these examples together because
+          # separating them would duplicate expensive setup:
+          # even with HTTP caching, this single example takes nearly a minute.
+          it 'imports max. ~120 articles every 15 minutes' do
             channel.fetch
 
-            expect((twitter_articles - Ticket.last.articles).count).to be < 120
-            expect(twitter_articles.count).to be >= 120
-          end
+            expect((twitter_articles - Ticket.last.articles).count).to be <= 120
+            expect(twitter_articles.count).to be > 120
 
-          it 'refuses to import any other tweets for the next 15 minutes' do
-            channel.fetch
             travel(14.minutes)
 
             expect { create(:twitter_channel).fetch }
               .not_to change(Ticket::Article, :count)
-          end
 
-          it 'resumes importing again after 15 minutes' do
-            channel.fetch
-            travel(15.minutes)
+            travel(1.minute)
 
             expect { create(:twitter_channel).fetch }
               .to change(Ticket::Article, :count)
           end
         end
-
-        context 'and "track_retweets" option' do
-          subject(:channel) { create(:twitter_channel, custom_options: { sync: { track_retweets: true } }) }
-
-          it 'creates an article for each recent tweet/retweet' do
-            expect { channel.fetch }
-              .to change { Ticket.where('title LIKE ?', 'RT @%').count }
-              .and change(Ticket, :count).by(21)
-          end
-        end
-
-        context 'and "import_older_tweets" option (legacy)' do
-          subject(:channel) { create(:twitter_channel, :legacy) }
-
-          it 'creates an article for each tweet' do
-            expect { channel.fetch }
-              .to change { Ticket.where('title LIKE ?', 'GitHub Trending Archive, 2_ Nov 2018, Ruby. %').count }
-              .and change(Ticket, :count).by(21)
-          end
-        end
-
-        describe 'Race condition: when #fetch finds a half-processed, outgoing tweet' do
-          subject!(:channel) { create(:twitter_channel, custom_options: custom_options) }
-
-          let(:custom_options) do
-            {
-              user: {
-                # Must match outgoing tweet author Twitter user ID
-                id: '1205290247124217856',
-              },
-              sync: {
-                search: [
-                  {
-                    term:     'zammadzammadzammad',
-                    group_id: Group.first.id
-                  }
-                ]
-              }
-            }
-          end
-
-          let!(:tweet) { create(:twitter_article, body: 'zammadzammadzammad') }
-
-          context '(i.e., after the BG job has posted the article to Twitter…' do
-            # NOTE: This context block cannot be set up programmatically.
-            # Instead, the tweet was posted, fetched, recorded into a VCR cassette,
-            # and then manually copied into the existing VCR cassette for this example.
-
-            context '…but before the BG job has "synced" article.message_id with tweet.id)' do
-              let(:twitter_job) { Delayed::Job.find_by(handler: <<~YML) }
-                --- !ruby/object:Observer::Ticket::Article::CommunicateTwitter::BackgroundJob
-                article_id: #{tweet.id}
-              YML
-
-              around do |example|
-                # This test case requires the use_vcr: :time_sensitive option
-                # to travel_to(when the VCR cassette was recorded).
-                #
-                # This ensures that #fetch doesn't ignore
-                # the "older" tweets stored in the VCR cassette,
-                # but it also freezes time,
-                # which breaks this race condition handling logic:
-                #
-                #     break if Delayed::Job.where('created_at < ?', Time.current).none?
-                #
-                # So, we unfreeze time here.
-                travel_back
-
-                # Run BG job (Why not use Scheduler.worker?
-                # It led to hangs & failures elsewhere in test suite.)
-                Thread.new do
-                  sleep 5  # simulate other bg jobs holding up the queue
-                  twitter_job.invoke_job
-                end.tap { example.run }.join
-              end
-
-              it 'does not import the duplicate tweet (waits up to 60s for BG job to finish)' do
-                expect { channel.fetch }
-                  .to not_change(Ticket::Article, :count)
-              end
-            end
-          end
-        end
       end
     end
   end