A way of encapsulate custom sql within objects and of course it's possible to make really clean with more inheritance.

context.rb

# encoding: utf-8

class Context

  def sanitize_sql(array)
    self.class.sanitize_sql(array)
  end

  def select_all(sql)
    self.class.select_all(sql)
  end

  def execute(sql)
    self.class.execute(sql)
  end

  def self.sanitize_sql(array)
    compress ActiveRecord::Base.send(:sanitize_sql_array, array)
  end

  def compress(sql)
    sql.gsub(/\n/,"").squeeze(' ').strip
  end

  def self.select_all(sql)
    ActiveRecord::Base.connection.select_all(sql)
  end

  def self.execute(sql)
    ActiveRecord::Base.connection.execute(sql)
  end

end

orders.rb

# encoding utf-8

class Statistics::Orders < Context
  
  class Result
    include Virtus
    attribute :total,     Integer
    attribute :delivered, Integer
    attribute :year,      Integer
    attribute :month,     Integer
    
    def month_name
      date.strftime("%B")
    end
    
    def date
      @date ||= Date.new(year, month, 1)
    end
  end
  
  def initialize(from = Date.today.beginning_of_month, to = Date.today.end_of_month)
    from.is_a?(Date) or raise "from is a #{from.class} expected Date"
    to.is_a?(Date)   or raise "to is a #{to.class} expected Date"

    @from = from
    @to   = to
  end

  def each &block
    orders.each &block
  end

  def sum
    @sum ||= Statistics::Result.new.tap do |r|
      r.total     = orders.sum(&:total)
      r.delivered = orders.sum(&:delivered)
    end
  end

  protected
  def orders
    @orders ||= select_all(sql).map { |h| Result.new h }
  end
  
  def sql
    sanitize_sql([raw_sql, @from, @to])
  end
  
  def raw_sql
    compress <<-SQL
      SELECT COUNT(*) "total", 
             SUM(CASE WHEN "delivered_at" IS NOT NULL THEN 1 ELSE 0 END) AS "delivered",
             EXTRACT(year from created_at) AS "year",
             EXTRACT(month from created_at) AS "month"
      FROM "orders" 
        INNER JOIN "users" ON "orders"."user_id" = "users"."id"
      WHERE "users"."active" = 1
        AND "created_at" >= ?
        AND "created_at" <= ?
      GROUP BY EXTRACT(year from created_at)
               EXTRACT(month from created_at)
    SQL
  end

end