Fix fake_data task; add CI test for it (#1087)

The test is overdue for how brittle fake_data.rb is, but as that's a slow
process that's not critical functionality, I only want it running on CI.

.github/workflows/check.yml

@@ -26,7 +26,9 @@ jobs:
       - name: Prepare database
         run: ./bin/rails db:schema:load
       - name: Run tests
-        run: bundle exec rspec
+        run: |
+          bundle exec rspec
+          bundle exec rspec spec/slow/*_spec.rb
       - name: Run linters
         run: |
           bundle exec rubocop

README.md

@@ -116,6 +116,9 @@ running tests:
 
 * See `config/initializers/production.rb.sample` for GitHub/Twitter integration help.
 
+* You probably want to use [git-imerge](https://lobste.rs/s/dbm2d4) to pull in
+  changes from Lobsters to your site.
+
 #### Administration
 
 Basic moderation happens on-site, but most other administrative tasks require use of the rails console in production.

db/seeds.rb

@@ -31,4 +31,4 @@ puts "  * inactive-user for disowned comments by deleted users"
 puts "  * a test tag"
 puts
 puts "If this is a dev environment, you probably want to run `rails fake_data`"
-puts "If this is production, you want to run `rails console` to rename your admin, category, and tag"
+puts "If this is production, you want to run `rails console` to rename your admin. Edit your category, and tag on-site."

lib/tasks/fake_data.rake

@@ -19,7 +19,7 @@ class FakeDataGenerator
         email: Faker::Internet.email(name: name),
         password: password,
         password_confirmation: password,
-        username: Faker::Internet.user_name(specifier: 5..40, separators: %w(_)),
+        username: Faker::Internet.user_name(specifier: name, separators: %w(_))[..23],
         created_at: (User::NEW_USER_DAYS + 1).days.ago,
         karma: Random.rand(User::MIN_KARMA_TO_FLAG * 2),
         about: Faker::Lorem.sentence(word_count: 7),
@@ -27,14 +27,20 @@ class FakeDataGenerator
         invited_by_user: User.select(&:can_invite?).sample,
       }
       create_args.merge!(is_admin: true) if i % 8 == 0
-      users << User.create!(create_args)
-      if i % 7 == 0
-        users[i].grant_moderatorship_by_user!(mod)
-      end
-      if i % 6 == 0
-        users[i].disable_invite_by_user_for_reason!(mod, Faker::Lorem.sentence(word_count: 5))
+      begin
+        users << User.create!(create_args)
+        if i % 7 == 0
+          users.last.grant_moderatorship_by_user!(mod)
+        end
+        if i % 6 == 0
+          users.last.disable_invite_by_user_for_reason!(mod, Faker::Lorem.sentence(word_count: 5))
+        end
+      rescue ActiveRecord::RecordInvalid => e
+        puts "caught #{e}"
+        next if e.message == 'Validation failed: Username has already been taken'
       end
     end
+    users.compact!
     puts
 
     print 'Categories '
@@ -225,7 +231,7 @@ class FakeDataGenerator
       user.ban_by_user_for_reason!(User.moderators.sample,
                                    Faker::Lorem.sentence(word_count: 5))
       if i.even?
-        user.unban_by_user!(User.moderators.sample)
+        user.unban_by_user!(User.moderators.sample, "reformed")
       end
     end
     puts

spec/slow/README.md

@@ -0,0 +1,2 @@
+Tests in this directory do not run as part of the regular test suite with
+`bundle exec rspec`. They are excluded by config.exclude

spec/slow/fake_data_spec.rb

@@ -0,0 +1,12 @@
+require 'rails_helper'
+
+Rails.application.load_tasks
+
+describe "fake_data" do
+  before { Rails.application.load_seed }
+
+  # basic smoke test, task shouldn't throw exceptions
+  it 'runs' do
+    FakeDataGenerator.new.generate
+  end
+end

spec/spec_helper.rb

@@ -37,6 +37,9 @@ RSpec.configure do |config|
     mocks.verify_partial_doubles = true
   end
 
+  # Some tests are slow, rarely catch a bug, and are not critical functionality.
+  config.exclude_pattern = 'spec/slow/*_spec.rb'
+
   # This option will default to `:apply_to_host_groups` in RSpec 4 (and will
   # have no way to turn it off -- the option exists only for backwards
   # compatibility in RSpec 3). It causes shared context metadata to be