activerecord-refined 0.8.1 → 0.10.0

data/test/test_helper.rb

@@ -1,221 +0,0 @@
-$LOAD_PATH.unshift(File.join(File.dirname(__FILE__), '..', 'lib'))
-$LOAD_PATH.unshift(File.dirname(__FILE__))
-
-require 'minitest/autorun'
-Bundler.require
-
-require 'active_record'
-require 'etc'
-require 'json'
-
-# Which database to generate SQL for.  The tests only build queries, so any
-# server reachable without further setup will do; the devcontainer runs
-# PostgreSQL and MariaDB on localhost with a passwordless account named after
-# the container user.
-#
-#   rake test                  # sqlite3
-#   ADAPTER=postgresql rake test
-#   ADAPTER=mysql2 rake test
-#   rake test:all              # all three in turn
-ADAPTER = ENV.fetch('ADAPTER', 'sqlite3')
-
-DATABASE_NAME = 'activerecord_refined_test'.freeze
-
-DATABASE_CONFIG =
-  case ADAPTER
-  when 'sqlite3'
-    {adapter: 'sqlite3', database: ':memory:'}
-  when 'postgresql', 'mysql2'
-    {
-      adapter: ADAPTER,
-      host: ENV.fetch('DB_HOST', '127.0.0.1'),
-      username: ENV.fetch('DB_USERNAME') { Etc.getlogin },
-      password: ENV['DB_PASSWORD'],
-      database: DATABASE_NAME,
-    }
-  else
-    raise ArgumentError, "unknown ADAPTER: #{ADAPTER}"
-  end
-
-# The server databases persist between runs, so create one on first use.
-unless ADAPTER == 'sqlite3'
-  maintenance_database = ADAPTER == 'postgresql' ? 'postgres' : 'mysql'
-  ActiveRecord::Base.establish_connection(
-    DATABASE_CONFIG.merge(database: maintenance_database))
-  begin
-    ActiveRecord::Base.lease_connection.create_database(DATABASE_NAME)
-  rescue ActiveRecord::DatabaseAlreadyExists
-  end
-  ActiveRecord::Base.remove_connection
-end
-
-ActiveRecord::Base.establish_connection(DATABASE_CONFIG)
-
-module SqlAssertions
-  # Arel spells the regexp match differently per adapter, and raises on the
-  # ones missing from this list.
-  REGEXP_OPERATORS = {
-    'postgresql' => ['~', '!~'],
-    'mysql2' => ['REGEXP', 'NOT REGEXP'],
-  }.freeze
-
-  def skip_without_regexp_support
-    skip "#{ADAPTER} has no regexp operator" unless REGEXP_OPERATORS.key?(ADAPTER)
-  end
-
-  # upsert_all wants to be told which unique index it is upserting against,
-  # except on MySQL, which does not accept being told.
-  def upsert_target
-    ADAPTER == 'mysql2' ? {} : {unique_by: :page}
-  end
-
-  # JSON containment: PostgreSQL has @>, MySQL JSON_CONTAINS, SQLite neither.
-  def skip_without_json_containment
-    skip "#{ADAPTER} has no JSON containment" if ADAPTER == 'sqlite3'
-  end
-
-  # MySQL is the one without a FULL OUTER JOIN, and so is MariaDB.
-  def skip_without_full_outer_joins
-    skip "#{ADAPTER} has no full outer join" if ADAPTER == 'mysql2'
-  end
-
-  # MariaDB's json is a checked longtext, which Active Record sees as a string
-  # and does not serialise a hash into -- what would go in is Ruby's inspect,
-  # which the check refuses.  MySQL's json is a type of its own and takes the
-  # hash; handing that one a string would store the document as a JSON string
-  # rather than as the object, and every path would find nothing.  Both answer
-  # to the mysql2 adapter, so the two have to be told apart.
-  def json_document(hash)
-    mariadb? ? JSON.generate(hash) : hash
-  end
-
-  # How each adapter spells a cast to a whole number: MySQL casts to SIGNED
-  # and has no integer at all, PostgreSQL the other way about, SQLite either.
-  def integer_type
-    ADAPTER == 'mysql2' ? 'signed' : 'integer'
-  end
-
-  def mariadb?
-    ADAPTER == 'mysql2' && ActiveRecord::Base.connection.mariadb?
-  end
-
-  # GROUPING SETS, ROLLUP and CUBE are PostgreSQL's; MySQL has only WITH
-  # ROLLUP, which is a different clause, and SQLite has none.
-  def skip_without_grouping_sets
-    skip "#{ADAPTER} has no GROUPING SETS" unless ADAPTER == 'postgresql'
-  end
-
-  # LATERAL is PostgreSQL's and MySQL 8's; MariaDB and SQLite have none.
-  def skip_without_lateral
-    return if ADAPTER == 'postgresql'
-    skip "#{mariadb? ? 'MariaDB' : ADAPTER} has no LATERAL" if ADAPTER != 'mysql2' || mariadb?
-  end
-
-  # DISTINCT ON is PostgreSQL's; Arel refuses to write it for the others.
-  def skip_without_distinct_on
-    skip "#{ADAPTER} has no DISTINCT ON" unless ADAPTER == 'postgresql'
-  end
-
-  # ANY and ALL over a subquery are PostgreSQL's and MySQL's alike; SQLite
-  # has neither quantifier.
-  def skip_without_quantifiers
-    skip "#{ADAPTER} has no ANY or ALL" if ADAPTER == 'sqlite3'
-  end
-
-  # BIT_AND, BIT_OR and BIT_XOR are PostgreSQL's and MySQL's; SQLite has
-  # none of the three, nor BIT_COUNT.
-  def skip_without_bit_aggregates
-    skip "#{ADAPTER} has no bit aggregates" if ADAPTER == 'sqlite3'
-  end
-
-  def skip_without_array_columns
-    skip "#{ADAPTER} has no array columns" unless ADAPTER == 'postgresql'
-  end
-
-  # MySQL has no NULLS FIRST/LAST; Arel emulates it with a leading IS NULL
-  # ordering, so only the resulting order is portable, not the SQL.
-  def skip_without_nulls_ordering_syntax
-    skip "#{ADAPTER} emulates NULLS FIRST/LAST" if ADAPTER == 'mysql2'
-  end
-
-  def regexp_operator
-    Regexp.escape(REGEXP_OPERATORS.fetch(ADAPTER).first)
-  end
-
-  def not_regexp_operator
-    Regexp.escape(REGEXP_OPERATORS.fetch(ADAPTER).last)
-  end
-
-  # MySQL quotes identifiers with backticks instead of double quotes, and
-  # doubles the backslashes inside string literals because backslash is itself
-  # an escape character there.  Normalising both lets a single expectation
-  # cover every adapter.
-  def normalize_sql(sql)
-    sql.tr('`', '"').gsub("\\\\") { "\\" }
-  end
-
-  # Takes a relation or the SQL string of one.
-  def assert_sql(pattern, relation_or_sql)
-    sql = relation_or_sql.respond_to?(:to_sql) ? relation_or_sql.to_sql : relation_or_sql
-    assert_match(pattern, normalize_sql(sql))
-  end
-end
-
-class Minitest::Test
-  include SqlAssertions
-end
-
-class User < ActiveRecord::Base
-end
-
-class Author < ActiveRecord::Base
-  has_many :posts
-end
-
-class Post < ActiveRecord::Base
-  belongs_to :author
-end
-
-# Self-referencing, for the recursive CTE tests.
-class Node < ActiveRecord::Base
-end
-
-# For the writing statements, which need something to conflict with.
-class Tally < ActiveRecord::Base
-end
-
-# For the JSON tests; see json_document for how the two MySQLs differ.
-class Doc < ActiveRecord::Base
-end
-
-class CreateAllTables < ActiveRecord::Migration[8.1]
-  def up
-    drop_table(:users, if_exists: true)
-    drop_table(:authors, if_exists: true)
-    drop_table(:posts, if_exists: true)
-    drop_table(:nodes, if_exists: true)
-    drop_table(:tallies, if_exists: true)
-    drop_table(:docs, if_exists: true)
-    create_table(:users) do |t|
-      t.string :name
-      t.integer :age
-      t.boolean :active
-      t.integer :flags
-      t.string :tags, array: true if ADAPTER == 'postgresql'
-    end
-    create_table(:authors) {|t| t.string :name}
-    create_table(:posts) {|t| t.string :title; t.integer :author_id}
-    create_table(:nodes) {|t| t.string :name; t.integer :parent_id}
-    create_table(:tallies) do |t|
-      t.string :page
-      t.integer :hits
-      t.index :page, unique: true
-    end
-    create_table(:docs) do |t|
-      t.string :name
-      ADAPTER == 'postgresql' ? t.jsonb(:meta) : t.json(:meta)
-    end
-  end
-end
-ActiveRecord::Migration.verbose = false
-CreateAllTables.new.up