activerecord 4.2.8 → 6.0.0

data/lib/active_record/relation/batches.rb

@@ -1,8 +1,14 @@
+# frozen_string_literal: true
+
+require "active_record/relation/batches/batch_enumerator"
+
 module ActiveRecord
   module Batches
+    ORDER_IGNORE_MESSAGE = "Scoped order is ignored, it's forced to be batch order."
+
     # Looping through a collection of records from the database
-    # (using the +all+ method, for example) is very inefficient
-    # since it will try to instantiate all the objects at once.
+    # (using the Scoping::Named::ClassMethods.all method, for example)
+    # is very inefficient since it will try to instantiate all the objects at once.
     #
     # In that case, batch processing methods allow you to work
     # with the records in batches, thereby greatly reducing memory consumption.
@@ -26,38 +32,52 @@ module ActiveRecord
     #   end
     #
     # ==== Options
-    # * <tt>:batch_size</tt> - Specifies the size of the batch. Default to 1000.
-    # * <tt>:start</tt> - Specifies the starting point for the batch processing.
-    # This is especially useful if you want multiple workers dealing with
-    # the same processing queue. You can make worker 1 handle all the records
-    # between id 0 and 10,000 and worker 2 handle from 10,000 and beyond
-    # (by setting the +:start+ option on that worker).
-    #
-    #   # Let's process for a batch of 2000 records, skipping the first 2000 rows
-    #   Person.find_each(start: 2000, batch_size: 2000) do |person|
+    # * <tt>:batch_size</tt> - Specifies the size of the batch. Defaults to 1000.
+    # * <tt>:start</tt> - Specifies the primary key value to start from, inclusive of the value.
+    # * <tt>:finish</tt> - Specifies the primary key value to end at, inclusive of the value.
+    # * <tt>:error_on_ignore</tt> - Overrides the application config to specify if an error should be raised when
+    #   an order is present in the relation.
+    #
+    # Limits are honored, and if present there is no requirement for the batch
+    # size: it can be less than, equal to, or greater than the limit.
+    #
+    # The options +start+ and +finish+ are especially useful if you want
+    # multiple workers dealing with the same processing queue. You can make
+    # worker 1 handle all the records between id 1 and 9999 and worker 2
+    # handle from 10000 and beyond by setting the +:start+ and +:finish+
+    # option on each worker.
+    #
+    #   # In worker 1, let's process until 9999 records.
+    #   Person.find_each(finish: 9_999) do |person|
+    #     person.party_all_night!
+    #   end
+    #
+    #   # In worker 2, let's process from record 10_000 and onwards.
+    #   Person.find_each(start: 10_000) do |person|
     #     person.party_all_night!
     #   end
     #
     # NOTE: It's not possible to set the order. That is automatically set to
     # ascending on the primary key ("id ASC") to make the batch ordering
-    # work. This also means that this method only works with integer-based
-    # primary keys.
+    # work. This also means that this method only works when the primary key is
+    # orderable (e.g. an integer or string).
     #
-    # NOTE: You can't set the limit either, that's used to control
-    # the batch sizes.
-    def find_each(options = {})
+    # NOTE: By its nature, batch processing is subject to race conditions if
+    # other processes are modifying the database.
+    def find_each(start: nil, finish: nil, batch_size: 1000, error_on_ignore: nil)
       if block_given?
-        find_in_batches(options) do |records|
+        find_in_batches(start: start, finish: finish, batch_size: batch_size, error_on_ignore: error_on_ignore) do |records|
           records.each { |record| yield record }
         end
       else
-        enum_for :find_each, options do
-          options[:start] ? where(table[primary_key].gteq(options[:start])).size : size
+        enum_for(:find_each, start: start, finish: finish, batch_size: batch_size, error_on_ignore: error_on_ignore) do
+          relation = self
+          apply_limits(relation, start, finish).size
         end
       end
     end
 
-    # Yields each batch of records that was found by the find +options+ as
+    # Yields each batch of records that was found by the find options as
     # an array.
     #
     #   Person.where("age > 21").find_in_batches do |group|
@@ -76,63 +96,195 @@ module ActiveRecord
     # To be yielded each record one by one, use #find_each instead.
     #
     # ==== Options
-    # * <tt>:batch_size</tt> - Specifies the size of the batch. Default to 1000.
-    # * <tt>:start</tt> - Specifies the starting point for the batch processing.
-    # This is especially useful if you want multiple workers dealing with
-    # the same processing queue. You can make worker 1 handle all the records
-    # between id 0 and 10,000 and worker 2 handle from 10,000 and beyond
-    # (by setting the +:start+ option on that worker).
-    #
-    #   # Let's process the next 2000 records
-    #   Person.find_in_batches(start: 2000, batch_size: 2000) do |group|
+    # * <tt>:batch_size</tt> - Specifies the size of the batch. Defaults to 1000.
+    # * <tt>:start</tt> - Specifies the primary key value to start from, inclusive of the value.
+    # * <tt>:finish</tt> - Specifies the primary key value to end at, inclusive of the value.
+    # * <tt>:error_on_ignore</tt> - Overrides the application config to specify if an error should be raised when
+    #   an order is present in the relation.
+    #
+    # Limits are honored, and if present there is no requirement for the batch
+    # size: it can be less than, equal to, or greater than the limit.
+    #
+    # The options +start+ and +finish+ are especially useful if you want
+    # multiple workers dealing with the same processing queue. You can make
+    # worker 1 handle all the records between id 1 and 9999 and worker 2
+    # handle from 10000 and beyond by setting the +:start+ and +:finish+
+    # option on each worker.
+    #
+    #   # Let's process from record 10_000 on.
+    #   Person.find_in_batches(start: 10_000) do |group|
     #     group.each { |person| person.party_all_night! }
     #   end
     #
     # NOTE: It's not possible to set the order. That is automatically set to
     # ascending on the primary key ("id ASC") to make the batch ordering
-    # work. This also means that this method only works with integer-based
-    # primary keys.
+    # work. This also means that this method only works when the primary key is
+    # orderable (e.g. an integer or string).
     #
-    # NOTE: You can't set the limit either, that's used to control
-    # the batch sizes.
-    def find_in_batches(options = {})
-      options.assert_valid_keys(:start, :batch_size)
-
+    # NOTE: By its nature, batch processing is subject to race conditions if
+    # other processes are modifying the database.
+    def find_in_batches(start: nil, finish: nil, batch_size: 1000, error_on_ignore: nil)
       relation = self
-      start = options[:start]
-      batch_size = options[:batch_size] || 1000
-
       unless block_given?
-        return to_enum(:find_in_batches, options) do
-          total = start ? where(table[primary_key].gteq(start)).size : size
+        return to_enum(:find_in_batches, start: start, finish: finish, batch_size: batch_size, error_on_ignore: error_on_ignore) do
+          total = apply_limits(relation, start, finish).size
           (total - 1).div(batch_size) + 1
         end
       end
 
-      if logger && (arel.orders.present? || arel.taken.present?)
-        logger.warn("Scoped order and limit are ignored, it's forced to be batch order and batch size")
+      in_batches(of: batch_size, start: start, finish: finish, load: true, error_on_ignore: error_on_ignore) do |batch|
+        yield batch.to_a
+      end
+    end
+
+    # Yields ActiveRecord::Relation objects to work with a batch of records.
+    #
+    #   Person.where("age > 21").in_batches do |relation|
+    #     relation.delete_all
+    #     sleep(10) # Throttle the delete queries
+    #   end
+    #
+    # If you do not provide a block to #in_batches, it will return a
+    # BatchEnumerator which is enumerable.
+    #
+    #   Person.in_batches.each_with_index do |relation, batch_index|
+    #     puts "Processing relation ##{batch_index}"
+    #     relation.delete_all
+    #   end
+    #
+    # Examples of calling methods on the returned BatchEnumerator object:
+    #
+    #   Person.in_batches.delete_all
+    #   Person.in_batches.update_all(awesome: true)
+    #   Person.in_batches.each_record(&:party_all_night!)
+    #
+    # ==== Options
+    # * <tt>:of</tt> - Specifies the size of the batch. Defaults to 1000.
+    # * <tt>:load</tt> - Specifies if the relation should be loaded. Defaults to false.
+    # * <tt>:start</tt> - Specifies the primary key value to start from, inclusive of the value.
+    # * <tt>:finish</tt> - Specifies the primary key value to end at, inclusive of the value.
+    # * <tt>:error_on_ignore</tt> - Overrides the application config to specify if an error should be raised when
+    #   an order is present in the relation.
+    #
+    # Limits are honored, and if present there is no requirement for the batch
+    # size, it can be less than, equal, or greater than the limit.
+    #
+    # The options +start+ and +finish+ are especially useful if you want
+    # multiple workers dealing with the same processing queue. You can make
+    # worker 1 handle all the records between id 1 and 9999 and worker 2
+    # handle from 10000 and beyond by setting the +:start+ and +:finish+
+    # option on each worker.
+    #
+    #   # Let's process from record 10_000 on.
+    #   Person.in_batches(start: 10_000).update_all(awesome: true)
+    #
+    # An example of calling where query method on the relation:
+    #
+    #   Person.in_batches.each do |relation|
+    #     relation.update_all('age = age + 1')
+    #     relation.where('age > 21').update_all(should_party: true)
+    #     relation.where('age <= 21').delete_all
+    #   end
+    #
+    # NOTE: If you are going to iterate through each record, you should call
+    # #each_record on the yielded BatchEnumerator:
+    #
+    #   Person.in_batches.each_record(&:party_all_night!)
+    #
+    # NOTE: It's not possible to set the order. That is automatically set to
+    # ascending on the primary key ("id ASC") to make the batch ordering
+    # consistent. Therefore the primary key must be orderable, e.g. an integer
+    # or a string.
+    #
+    # NOTE: By its nature, batch processing is subject to race conditions if
+    # other processes are modifying the database.
+    def in_batches(of: 1000, start: nil, finish: nil, load: false, error_on_ignore: nil)
+      relation = self
+      unless block_given?
+        return BatchEnumerator.new(of: of, start: start, finish: finish, relation: self)
+      end
+
+      if arel.orders.present?
+        act_on_ignored_order(error_on_ignore)
+      end
+
+      batch_limit = of
+      if limit_value
+        remaining   = limit_value
+        batch_limit = remaining if remaining < batch_limit
       end
 
-      relation = relation.reorder(batch_order).limit(batch_size)
-      records = start ? relation.where(table[primary_key].gteq(start)).to_a : relation.to_a
+      relation = relation.reorder(batch_order).limit(batch_limit)
+      relation = apply_limits(relation, start, finish)
+      relation.skip_query_cache! # Retaining the results in the query cache would undermine the point of batching
+      batch_relation = relation
+
+      loop do
+        if load
+          records = batch_relation.records
+          ids = records.map(&:id)
+          yielded_relation = where(primary_key => ids)
+          yielded_relation.load_records(records)
+        else
+          ids = batch_relation.pluck(primary_key)
+          yielded_relation = where(primary_key => ids)
+        end
+
+        break if ids.empty?
 
-      while records.any?
-        records_size = records.size
-        primary_key_offset = records.last.id
-        raise "Primary key not included in the custom select clause" unless primary_key_offset
+        primary_key_offset = ids.last
+        raise ArgumentError.new("Primary key not included in the custom select clause") unless primary_key_offset
 
-        yield records
+        yield yielded_relation
 
-        break if records_size < batch_size
+        break if ids.length < batch_limit
 
-        records = relation.where(table[primary_key].gt(primary_key_offset)).to_a
+        if limit_value
+          remaining -= ids.length
+
+          if remaining == 0
+            # Saves a useless iteration when the limit is a multiple of the
+            # batch size.
+            break
+          elsif remaining < batch_limit
+            relation = relation.limit(remaining)
+          end
+        end
+
+        batch_relation = relation.where(
+          bind_attribute(primary_key, primary_key_offset) { |attr, bind| attr.gt(bind) }
+        )
       end
     end
 
     private
 
-    def batch_order
-      "#{quoted_table_name}.#{quoted_primary_key} ASC"
-    end
+      def apply_limits(relation, start, finish)
+        relation = apply_start_limit(relation, start) if start
+        relation = apply_finish_limit(relation, finish) if finish
+        relation
+      end
+
+      def apply_start_limit(relation, start)
+        relation.where(bind_attribute(primary_key, start) { |attr, bind| attr.gteq(bind) })
+      end
+
+      def apply_finish_limit(relation, finish)
+        relation.where(bind_attribute(primary_key, finish) { |attr, bind| attr.lteq(bind) })
+      end
+
+      def batch_order
+        arel_attribute(primary_key).asc
+      end
+
+      def act_on_ignored_order(error_on_ignore)
+        raise_error = (error_on_ignore.nil? ? klass.error_on_ignored_order : error_on_ignore)
+
+        if raise_error
+          raise ArgumentError.new(ORDER_IGNORE_MESSAGE)
+        elsif logger
+          logger.warn(ORDER_IGNORE_MESSAGE)
+        end
+      end
   end
 end

data/lib/active_record/relation/batches/batch_enumerator.rb

@@ -0,0 +1,69 @@
+# frozen_string_literal: true
+
+module ActiveRecord
+  module Batches
+    class BatchEnumerator
+      include Enumerable
+
+      def initialize(of: 1000, start: nil, finish: nil, relation:) #:nodoc:
+        @of       = of
+        @relation = relation
+        @start = start
+        @finish = finish
+      end
+
+      # Looping through a collection of records from the database (using the
+      # +all+ method, for example) is very inefficient since it will try to
+      # instantiate all the objects at once.
+      #
+      # In that case, batch processing methods allow you to work with the
+      # records in batches, thereby greatly reducing memory consumption.
+      #
+      #   Person.in_batches.each_record do |person|
+      #     person.do_awesome_stuff
+      #   end
+      #
+      #   Person.where("age > 21").in_batches(of: 10).each_record do |person|
+      #     person.party_all_night!
+      #   end
+      #
+      # If you do not provide a block to #each_record, it will return an Enumerator
+      # for chaining with other methods:
+      #
+      #   Person.in_batches.each_record.with_index do |person, index|
+      #     person.award_trophy(index + 1)
+      #   end
+      def each_record
+        return to_enum(:each_record) unless block_given?
+
+        @relation.to_enum(:in_batches, of: @of, start: @start, finish: @finish, load: true).each do |relation|
+          relation.records.each { |record| yield record }
+        end
+      end
+
+      # Delegates #delete_all, #update_all, #destroy_all methods to each batch.
+      #
+      #   People.in_batches.delete_all
+      #   People.where('age < 10').in_batches.destroy_all
+      #   People.in_batches.update_all('age = age + 1')
+      [:delete_all, :update_all, :destroy_all].each do |method|
+        define_method(method) do |*args, &block|
+          @relation.to_enum(:in_batches, of: @of, start: @start, finish: @finish, load: false).each do |relation|
+            relation.send(method, *args, &block)
+          end
+        end
+      end
+
+      # Yields an ActiveRecord::Relation object for each batch of records.
+      #
+      #   Person.in_batches.each do |relation|
+      #     relation.update_all(awesome: true)
+      #   end
+      def each
+        enum = @relation.to_enum(:in_batches, of: @of, start: @start, finish: @finish, load: false)
+        return enum.each { |relation| yield relation } if block_given?
+        enum
+      end
+    end
+  end
+end

data/lib/active_record/relation/calculations.rb

@@ -1,3 +1,5 @@
+# frozen_string_literal: true
+
 module ActiveRecord
   module Calculations
     # Count the records.
@@ -14,127 +16,134 @@ module ActiveRecord
     #   Person.distinct.count(:age)
     #   # => counts the number of different age values
     #
-    # If +count+ is used with +group+, it returns a Hash whose keys represent the aggregated column,
+    # If #count is used with {Relation#group}[rdoc-ref:QueryMethods#group],
+    # it returns a Hash whose keys represent the aggregated column,
     # and the values are the respective amounts:
     #
     #   Person.group(:city).count
     #   # => { 'Rome' => 5, 'Paris' => 3 }
     #
-    # If +count+ is used with +group+ for multiple columns, it returns a Hash whose
+    # If #count is used with {Relation#group}[rdoc-ref:QueryMethods#group] for multiple columns, it returns a Hash whose
     # keys are an array containing the individual values of each column and the value
-    # of each key would be the +count+.
+    # of each key would be the #count.
     #
     #   Article.group(:status, :category).count
     #   # =>  {["draft", "business"]=>10, ["draft", "technology"]=>4,
     #          ["published", "business"]=>0, ["published", "technology"]=>2}
     #
-    # If +count+ is used with +select+, it will count the selected columns:
+    # If #count is used with {Relation#select}[rdoc-ref:QueryMethods#select], it will count the selected columns:
     #
     #   Person.select(:age).count
     #   # => counts the number of different age values
     #
-    # Note: not all valid +select+ expressions are valid +count+ expressions. The specifics differ
+    # Note: not all valid {Relation#select}[rdoc-ref:QueryMethods#select] expressions are valid #count expressions. The specifics differ
     # between databases. In invalid cases, an error from the database is thrown.
-    def count(column_name = nil, options = {})
-      if options.present? && !ActiveRecord.const_defined?(:DeprecatedFinders)
-        raise ArgumentError, "Relation#count does not support finder options anymore. " \
-                             "Please build a scope and then call count on it or use the " \
-                             "activerecord-deprecated_finders gem to enable this functionality."
+    def count(column_name = nil)
+      if block_given?
+        unless column_name.nil?
+          raise ArgumentError, "Column name argument is not supported when a block is passed."
+        end
 
+        super()
+      else
+        calculate(:count, column_name)
       end
-
-      # TODO: Remove options argument as soon we remove support to
-      # activerecord-deprecated_finders.
-      calculate(:count, column_name, options)
     end
 
     # Calculates the average value on a given column. Returns +nil+ if there's
-    # no row. See +calculate+ for examples with options.
+    # no row. See #calculate for examples with options.
     #
     #   Person.average(:age) # => 35.8
-    def average(column_name, options = {})
-      # TODO: Remove options argument as soon we remove support to
-      # activerecord-deprecated_finders.
-      calculate(:average, column_name, options)
+    def average(column_name)
+      calculate(:average, column_name)
     end
 
     # Calculates the minimum value on a given column. The value is returned
     # with the same data type of the column, or +nil+ if there's no row. See
-    # +calculate+ for examples with options.
+    # #calculate for examples with options.
     #
     #   Person.minimum(:age) # => 7
-    def minimum(column_name, options = {})
-      # TODO: Remove options argument as soon we remove support to
-      # activerecord-deprecated_finders.
-      calculate(:minimum, column_name, options)
+    def minimum(column_name)
+      calculate(:minimum, column_name)
     end
 
     # Calculates the maximum value on a given column. The value is returned
     # with the same data type of the column, or +nil+ if there's no row. See
-    # +calculate+ for examples with options.
+    # #calculate for examples with options.
     #
     #   Person.maximum(:age) # => 93
-    def maximum(column_name, options = {})
-      # TODO: Remove options argument as soon we remove support to
-      # activerecord-deprecated_finders.
-      calculate(:maximum, column_name, options)
+    def maximum(column_name)
+      calculate(:maximum, column_name)
     end
 
     # Calculates the sum of values on a given column. The value is returned
-    # with the same data type of the column, 0 if there's no row. See
-    # +calculate+ for examples with options.
+    # with the same data type of the column, +0+ if there's no row. See
+    # #calculate for examples with options.
     #
     #   Person.sum(:age) # => 4562
-    def sum(*args)
-      calculate(:sum, *args)
+    def sum(column_name = nil)
+      if block_given?
+        unless column_name.nil?
+          raise ArgumentError, "Column name argument is not supported when a block is passed."
+        end
+
+        super()
+      else
+        calculate(:sum, column_name)
+      end
     end
 
-    # This calculates aggregate values in the given column. Methods for count, sum, average,
-    # minimum, and maximum have been added as shortcuts.
+    # This calculates aggregate values in the given column. Methods for #count, #sum, #average,
+    # #minimum, and #maximum have been added as shortcuts.
     #
-    # There are two basic forms of output:
+    #   Person.calculate(:count, :all) # The same as Person.count
+    #   Person.average(:age) # SELECT AVG(age) FROM people...
     #
-    #   * Single aggregate value: The single value is type cast to Integer for COUNT, Float
-    #     for AVG, and the given column's type for everything else.
+    #   # Selects the minimum age for any family without any minors
+    #   Person.group(:last_name).having("min(age) > 17").minimum(:age)
     #
-    #   * Grouped values: This returns an ordered hash of the values and groups them. It
-    #     takes either a column name, or the name of a belongs_to association.
+    #   Person.sum("2 * age")
     #
-    #       values = Person.group('last_name').maximum(:age)
-    #       puts values["Drake"]
-    #       # => 43
+    # There are two basic forms of output:
     #
-    #       drake  = Family.find_by(last_name: 'Drake')
-    #       values = Person.group(:family).maximum(:age) # Person belongs_to :family
-    #       puts values[drake]
-    #       # => 43
+    # * Single aggregate value: The single value is type cast to Integer for COUNT, Float
+    #   for AVG, and the given column's type for everything else.
     #
-    #       values.each do |family, max_age|
-    #       ...
-    #       end
+    # * Grouped values: This returns an ordered hash of the values and groups them. It
+    #   takes either a column name, or the name of a belongs_to association.
     #
-    #   Person.calculate(:count, :all) # The same as Person.count
-    #   Person.average(:age) # SELECT AVG(age) FROM people...
+    #      values = Person.group('last_name').maximum(:age)
+    #      puts values["Drake"]
+    #      # => 43
     #
-    #   # Selects the minimum age for any family without any minors
-    #   Person.group(:last_name).having("min(age) > 17").minimum(:age)
+    #      drake  = Family.find_by(last_name: 'Drake')
+    #      values = Person.group(:family).maximum(:age) # Person belongs_to :family
+    #      puts values[drake]
+    #      # => 43
     #
-    #   Person.sum("2 * age")
-    def calculate(operation, column_name, options = {})
-      # TODO: Remove options argument as soon we remove support to
-      # activerecord-deprecated_finders.
-      if column_name.is_a?(Symbol) && attribute_alias?(column_name)
-        column_name = attribute_alias(column_name)
-      end
-
+    #      values.each do |family, max_age|
+    #        ...
+    #      end
+    def calculate(operation, column_name)
       if has_include?(column_name)
-        construct_relation_for_association_calculations.calculate(operation, column_name, options)
+        relation = apply_join_dependency
+
+        if operation.to_s.downcase == "count"
+          unless distinct_value || distinct_select?(column_name || select_for_count)
+            relation.distinct!
+            relation.select_values = [ klass.primary_key || table[Arel.star] ]
+          end
+          # PostgreSQL: ORDER BY expressions must appear in SELECT list when using DISTINCT
+          relation.order_values = []
+        end
+
+        relation.calculate(operation, column_name)
       else
-        perform_calculation(operation, column_name, options)
+        perform_calculation(operation, column_name)
       end
     end
 
-    # Use <tt>pluck</tt> as a shortcut to select one or more attributes without
+    # Use #pluck as a shortcut to select one or more attributes without
     # loading a bunch of records just to grab the attributes you want.
     #
     #   Person.pluck(:name)
@@ -143,19 +152,19 @@ module ActiveRecord
     #
     #   Person.all.map(&:name)
     #
-    # Pluck returns an <tt>Array</tt> of attribute values type-casted to match
+    # Pluck returns an Array of attribute values type-casted to match
     # the plucked column names, if they can be deduced. Plucking an SQL fragment
     # returns String values by default.
     #
-    #   Person.pluck(:id)
-    #   # SELECT people.id FROM people
-    #   # => [1, 2, 3]
+    #   Person.pluck(:name)
+    #   # SELECT people.name FROM people
+    #   # => ['David', 'Jeremy', 'Jose']
     #
     #   Person.pluck(:id, :name)
     #   # SELECT people.id, people.name FROM people
     #   # => [[1, 'David'], [2, 'Jeremy'], [3, 'Jose']]
     #
-    #   Person.pluck('DISTINCT role')
+    #   Person.distinct.pluck(:role)
     #   # SELECT DISTINCT role FROM people
     #   # => ['admin', 'member', 'guest']
     #
@@ -167,27 +176,43 @@ module ActiveRecord
     #   # SELECT DATEDIFF(updated_at, created_at) FROM people
     #   # => ['0', '27761', '173']
     #
+    # See also #ids.
+    #
     def pluck(*column_names)
-      column_names.map! do |column_name|
-        if column_name.is_a?(Symbol) && attribute_alias?(column_name)
-          attribute_alias(column_name)
-        else
-          column_name.to_s
-        end
+      if loaded? && (column_names.map(&:to_s) - @klass.attribute_names - @klass.attribute_aliases.keys).empty?
+        return records.pluck(*column_names)
       end
 
       if has_include?(column_names.first)
-        construct_relation_for_association_calculations.pluck(*column_names)
+        relation = apply_join_dependency
+        relation.pluck(*column_names)
       else
+        klass.disallow_raw_sql!(column_names)
         relation = spawn
-        relation.select_values = column_names.map { |cn|
-          columns_hash.key?(cn) ? arel_table[cn] : cn
-        }
-        result = klass.connection.select_all(relation.arel, nil, relation.arel.bind_values + bind_values)
-        result.cast_values(klass.column_types)
+        relation.select_values = column_names
+        result = skip_query_cache_if_necessary { klass.connection.select_all(relation.arel, nil) }
+        result.cast_values(klass.attribute_types)
       end
     end
 
+    # Pick the value(s) from the named column(s) in the current relation.
+    # This is short-hand for <tt>relation.limit(1).pluck(*column_names).first</tt>, and is primarily useful
+    # when you have a relation that's already narrowed down to a single row.
+    #
+    # Just like #pluck, #pick will only load the actual value, not the entire record object, so it's also
+    # more efficient. The value is, again like with pluck, typecast by the column type.
+    #
+    #   Person.where(id: 1).pick(:name)
+    #   # SELECT people.name FROM people WHERE id = 1 LIMIT 1
+    #   # => 'David'
+    #
+    #   Person.where(id: 1).pick(:name, :email_address)
+    #   # SELECT people.name, people.email_address FROM people WHERE id = 1 LIMIT 1
+    #   # => [ 'David', 'david@loudthinking.com' ]
+    def pick(*column_names)
+      limit(1).pluck(*column_names).first
+    end
+
     # Pluck all the ID's for the relation using the table's primary key
     #
     #   Person.ids # SELECT people.id FROM people
@@ -197,214 +222,203 @@ module ActiveRecord
     end
 
     private
+      def has_include?(column_name)
+        eager_loading? || (includes_values.present? && column_name && column_name != :all)
+      end
 
-    def has_include?(column_name)
-      eager_loading? || (includes_values.present? && ((column_name && column_name != :all) || references_eager_loaded_tables?))
-    end
-
-    def perform_calculation(operation, column_name, options = {})
-      # TODO: Remove options argument as soon we remove support to
-      # activerecord-deprecated_finders.
-      operation = operation.to_s.downcase
-
-      # If #count is used with #distinct / #uniq it is considered distinct. (eg. relation.distinct.count)
-      distinct = self.distinct_value
-
-      if operation == "count"
-        column_name ||= select_for_count
-
-        unless arel.ast.grep(Arel::Nodes::OuterJoin).empty?
-          distinct = true
+      def perform_calculation(operation, column_name)
+        operation = operation.to_s.downcase
+
+        # If #count is used with #distinct (i.e. `relation.distinct.count`) it is
+        # considered distinct.
+        distinct = distinct_value
+
+        if operation == "count"
+          column_name ||= select_for_count
+          if column_name == :all
+            if !distinct
+              distinct = distinct_select?(select_for_count) if group_values.empty?
+            elsif group_values.any? || select_values.empty? && order_values.empty?
+              column_name = primary_key
+            end
+          elsif distinct_select?(column_name)
+            distinct = nil
+          end
         end
 
-        column_name = primary_key if column_name == :all && distinct
-        distinct = nil if column_name =~ /\s*DISTINCT[\s(]+/i
+        if group_values.any?
+          execute_grouped_calculation(operation, column_name, distinct)
+        else
+          execute_simple_calculation(operation, column_name, distinct)
+        end
       end
 
-      if group_values.any?
-        execute_grouped_calculation(operation, column_name, distinct)
-      else
-        execute_simple_calculation(operation, column_name, distinct)
+      def distinct_select?(column_name)
+        column_name.is_a?(::String) && /\bDISTINCT[\s(]/i.match?(column_name)
       end
-    end
 
-    def aggregate_column(column_name)
-      if @klass.column_names.include?(column_name.to_s)
-        Arel::Attribute.new(@klass.unscoped.table, column_name)
-      else
-        Arel.sql(column_name == :all ? "*" : column_name.to_s)
+      def aggregate_column(column_name)
+        return column_name if Arel::Expressions === column_name
+
+        arel_column(column_name.to_s) do |name|
+          Arel.sql(column_name == :all ? "*" : name)
+        end
       end
-    end
 
-    def operation_over_aggregate_column(column, operation, distinct)
-      operation == 'count' ? column.count(distinct) : column.send(operation)
-    end
+      def operation_over_aggregate_column(column, operation, distinct)
+        operation == "count" ? column.count(distinct) : column.send(operation)
+      end
 
-    def execute_simple_calculation(operation, column_name, distinct) #:nodoc:
-      # Postgresql doesn't like ORDER BY when there are no GROUP BY
-      relation = unscope(:order)
+      def execute_simple_calculation(operation, column_name, distinct) #:nodoc:
+        column_alias = column_name
 
-      column_alias = column_name
+        if operation == "count" && (column_name == :all && distinct || has_limit_or_offset?)
+          # Shortcut when limit is zero.
+          return 0 if limit_value == 0
 
-      bind_values = nil
+          query_builder = build_count_subquery(spawn, column_name, distinct)
+        else
+          # PostgreSQL doesn't like ORDER BY when there are no GROUP BY
+          relation = unscope(:order).distinct!(false)
 
-      if operation == "count" && (relation.limit_value || relation.offset_value)
-        # Shortcut when limit is zero.
-        return 0 if relation.limit_value == 0
+          column = aggregate_column(column_name)
 
-        query_builder = build_count_subquery(relation, column_name, distinct)
-        bind_values = query_builder.bind_values + relation.bind_values
-      else
-        column = aggregate_column(column_name)
+          select_value = operation_over_aggregate_column(column, operation, distinct)
+          if operation == "sum" && distinct
+            select_value.distinct = true
+          end
 
-        select_value = operation_over_aggregate_column(column, operation, distinct)
+          column_alias = select_value.alias
+          column_alias ||= @klass.connection.column_name_for_operation(operation, select_value)
+          relation.select_values = [select_value]
 
-        column_alias = select_value.alias
-        column_alias ||= @klass.connection.column_name_for_operation(operation, select_value)
-        relation.select_values = [select_value]
+          query_builder = relation.arel
+        end
 
-        query_builder = relation.arel
-        bind_values = query_builder.bind_values + relation.bind_values
-      end
+        result = skip_query_cache_if_necessary { @klass.connection.select_all(query_builder, nil) }
+        row    = result.first
+        value  = row && row.values.first
+        type   = result.column_types.fetch(column_alias) do
+          type_for(column_name)
+        end
 
-      result = @klass.connection.select_all(query_builder, nil, bind_values)
-      row    = result.first
-      value  = row && row.values.first
-      column = result.column_types.fetch(column_alias) do
-        type_for(column_name)
+        type_cast_calculated_value(value, type, operation)
       end
 
-      type_cast_calculated_value(value, column, operation)
-    end
-
-    def execute_grouped_calculation(operation, column_name, distinct) #:nodoc:
-      group_attrs = group_values
+      def execute_grouped_calculation(operation, column_name, distinct) #:nodoc:
+        group_fields = group_values
 
-      if group_attrs.first.respond_to?(:to_sym)
-        association  = @klass._reflect_on_association(group_attrs.first)
-        associated   = group_attrs.size == 1 && association && association.belongs_to? # only count belongs_to associations
-        group_fields = Array(associated ? association.foreign_key : group_attrs)
-      else
-        group_fields = group_attrs
-      end
-      group_fields = arel_columns(group_fields)
+        if group_fields.size == 1 && group_fields.first.respond_to?(:to_sym)
+          association  = klass._reflect_on_association(group_fields.first)
+          associated   = association && association.belongs_to? # only count belongs_to associations
+          group_fields = Array(association.foreign_key) if associated
+        end
+        group_fields = arel_columns(group_fields)
 
-      group_aliases = group_fields.map { |field|
-        column_alias_for(field)
-      }
-      group_columns = group_aliases.zip(group_fields).map { |aliaz,field|
-        [aliaz, field]
-      }
+        group_aliases = group_fields.map { |field|
+          field = connection.visitor.compile(field) if Arel.arel_node?(field)
+          column_alias_for(field.to_s.downcase)
+        }
+        group_columns = group_aliases.zip(group_fields)
+
+        aggregate_alias = column_alias_for("#{operation} #{column_name.to_s.downcase}")
+
+        select_values = [
+          operation_over_aggregate_column(
+            aggregate_column(column_name),
+            operation,
+            distinct).as(aggregate_alias)
+        ]
+        select_values += self.select_values unless having_clause.empty?
+
+        select_values.concat group_columns.map { |aliaz, field|
+          if field.respond_to?(:as)
+            field.as(aliaz)
+          else
+            "#{field} AS #{aliaz}"
+          end
+        }
 
-      group = group_fields
+        relation = except(:group).distinct!(false)
+        relation.group_values  = group_fields
+        relation.select_values = select_values
 
-      if operation == 'count' && column_name == :all
-        aggregate_alias = 'count_all'
-      else
-        aggregate_alias = column_alias_for([operation, column_name].join(' '))
-      end
+        calculated_data = skip_query_cache_if_necessary { @klass.connection.select_all(relation.arel, nil) }
 
-      select_values = [
-        operation_over_aggregate_column(
-          aggregate_column(column_name),
-          operation,
-          distinct).as(aggregate_alias)
-      ]
-      select_values += select_values unless having_values.empty?
-
-      select_values.concat group_fields.zip(group_aliases).map { |field,aliaz|
-        if field.respond_to?(:as)
-          field.as(aliaz)
-        else
-          "#{field} AS #{aliaz}"
+        if association
+          key_ids     = calculated_data.collect { |row| row[group_aliases.first] }
+          key_records = association.klass.base_class.where(association.klass.base_class.primary_key => key_ids)
+          key_records = Hash[key_records.map { |r| [r.id, r] }]
         end
-      }
-
-      relation = except(:group)
-      relation.group_values  = group
-      relation.select_values = select_values
-
-      calculated_data = @klass.connection.select_all(relation, nil, relation.arel.bind_values + bind_values)
 
-      if association
-        key_ids     = calculated_data.collect { |row| row[group_aliases.first] }
-        key_records = association.klass.base_class.find(key_ids)
-        key_records = Hash[key_records.map { |r| [r.id, r] }]
+        Hash[calculated_data.map do |row|
+          key = group_columns.map { |aliaz, col_name|
+            type = type_for(col_name) do
+              calculated_data.column_types.fetch(aliaz, Type.default_value)
+            end
+            type_cast_calculated_value(row[aliaz], type)
+          }
+          key = key.first if key.size == 1
+          key = key_records[key] if associated
+
+          type = calculated_data.column_types.fetch(aggregate_alias) { type_for(column_name) }
+          [key, type_cast_calculated_value(row[aggregate_alias], type, operation)]
+        end]
       end
 
-      Hash[calculated_data.map do |row|
-        key = group_columns.map { |aliaz, col_name|
-          column = calculated_data.column_types.fetch(aliaz) do
-            type_for(col_name)
-          end
-          type_cast_calculated_value(row[aliaz], column)
-        }
-        key = key.first if key.size == 1
-        key = key_records[key] if associated
-
-        column_type = calculated_data.column_types.fetch(aggregate_alias) { type_for(column_name) }
-        [key, type_cast_calculated_value(row[aggregate_alias], column_type, operation)]
-      end]
-    end
-
-    # Converts the given keys to the value that the database adapter returns as
-    # a usable column name:
-    #
-    #   column_alias_for("users.id")                 # => "users_id"
-    #   column_alias_for("sum(id)")                  # => "sum_id"
-    #   column_alias_for("count(distinct users.id)") # => "count_distinct_users_id"
-    #   column_alias_for("count(*)")                 # => "count_all"
-    #   column_alias_for("count", "id")              # => "count_id"
-    def column_alias_for(keys)
-      if keys.respond_to? :name
-        keys = "#{keys.relation.name}.#{keys.name}"
+      # Converts the given field to the value that the database adapter returns as
+      # a usable column name:
+      #
+      #   column_alias_for("users.id")                 # => "users_id"
+      #   column_alias_for("sum(id)")                  # => "sum_id"
+      #   column_alias_for("count(distinct users.id)") # => "count_distinct_users_id"
+      #   column_alias_for("count(*)")                 # => "count_all"
+      def column_alias_for(field)
+        column_alias = +field
+        column_alias.gsub!(/\*/, "all")
+        column_alias.gsub!(/\W+/, " ")
+        column_alias.strip!
+        column_alias.gsub!(/ +/, "_")
+
+        connection.table_alias_for(column_alias)
       end
 
-      table_name = keys.to_s.downcase
-      table_name.gsub!(/\*/, 'all')
-      table_name.gsub!(/\W+/, ' ')
-      table_name.strip!
-      table_name.gsub!(/ +/, '_')
-
-      @klass.connection.table_alias_for(table_name)
-    end
-
-    def type_for(field)
-      field_name = field.respond_to?(:name) ? field.name.to_s : field.to_s.split('.').last
-      @klass.type_for_attribute(field_name)
-    end
+      def type_for(field, &block)
+        field_name = field.respond_to?(:name) ? field.name.to_s : field.to_s.split(".").last
+        @klass.type_for_attribute(field_name, &block)
+      end
 
-    def type_cast_calculated_value(value, type, operation = nil)
-      case operation
-        when 'count'   then value.to_i
-        when 'sum'     then type.type_cast_from_database(value || 0)
-        when 'average' then value.respond_to?(:to_d) ? value.to_d : value
-        else type.type_cast_from_database(value)
+      def type_cast_calculated_value(value, type, operation = nil)
+        case operation
+        when "count"   then value.to_i
+        when "sum"     then type.deserialize(value || 0)
+        when "average" then value&.respond_to?(:to_d) ? value.to_d : value
+        else type.deserialize(value)
+        end
       end
-    end
 
-    # TODO: refactor to allow non-string `select_values` (eg. Arel nodes).
-    def select_for_count
-      if select_values.present?
-        select_values.join(", ")
-      else
-        :all
+      def select_for_count
+        if select_values.present?
+          return select_values.first if select_values.one?
+          select_values.join(", ")
+        else
+          :all
+        end
       end
-    end
 
-    def build_count_subquery(relation, column_name, distinct)
-      column_alias = Arel.sql('count_column')
-      subquery_alias = Arel.sql('subquery_for_count')
+      def build_count_subquery(relation, column_name, distinct)
+        if column_name == :all
+          column_alias = Arel.star
+          relation.select_values = [ Arel.sql(FinderMethods::ONE_AS_ONE) ] unless distinct
+        else
+          column_alias = Arel.sql("count_column")
+          relation.select_values = [ aggregate_column(column_name).as(column_alias) ]
+        end
 
-      aliased_column = aggregate_column(column_name == :all ? 1 : column_name).as(column_alias)
-      relation.select_values = [aliased_column]
-      arel = relation.arel
-      subquery = arel.as(subquery_alias)
+        subquery_alias = Arel.sql("subquery_for_count")
+        select_value = operation_over_aggregate_column(column_alias, "count", false)
 
-      sm = Arel::SelectManager.new relation.engine
-      sm.bind_values = arel.bind_values
-      select_value = operation_over_aggregate_column(column_alias, 'count', distinct)
-      sm.project(select_value).from(subquery)
-    end
+        relation.build_subquery(subquery_alias, select_value)
+      end
   end
 end