dynamoid 3.3.0 → 3.7.0

data/lib/dynamoid/associations/single_association.rb

@@ -1,12 +1,13 @@
 # frozen_string_literal: true
 
-module Dynamoid #:nodoc:
+module Dynamoid
   module Associations
     module SingleAssociation
       include Association
 
       delegate :class, to: :target
 
+      # @private
       def setter(object)
         if object.nil?
           delete
@@ -19,16 +20,37 @@ module Dynamoid #:nodoc:
         object
       end
 
+      # Delete a model from the association.
+      #
+      #   post.logo.delete # => nil
+      #
+      # Saves both models immediately - a source model and a target one so any
+      # unsaved changes will be saved. Doesn't delete an associated model from
+      # DynamoDB.
       def delete
-        target.send(target_association).disassociate(source.hash_key) if target && target_association
+        disassociate_source
         disassociate
         target
       end
 
+      # Create a new instance of the target class, persist it and associate.
+      #
+      #   post.logo.create!(hight: 50, width: 90)
+      #
+      # If the creation fails an exception will be raised.
+      #
+      # @param attributes [Hash] attributes of a model to create
+      # @return [Dynamoid::Document] created model
       def create!(attributes = {})
         setter(target_class.create!(attributes))
       end
 
+      # Create a new instance of the target class, persist it and associate.
+      #
+      #   post.logo.create(hight: 50, width: 90)
+      #
+      # @param attributes [Hash] attributes of a model to create
+      # @return [Dynamoid::Document] created model
       def create(attributes = {})
         setter(target_class.create(attributes))
       end
@@ -44,6 +66,7 @@ module Dynamoid #:nodoc:
 
       # Delegate methods we don't find directly to the target.
       #
+      # @private
       # @since 0.2.0
       def method_missing(method, *args)
         if target.respond_to?(method)
@@ -53,21 +76,25 @@ module Dynamoid #:nodoc:
         end
       end
 
+      # @private
       def nil?
         target.nil?
       end
 
+      # @private
       def empty?
         # This is needed to that ActiveSupport's #blank? and #present?
         # methods work as expected for SingleAssociations.
         target.nil?
       end
 
+      # @private
       def associate(hash_key)
-        target.send(target_association).disassociate(source.hash_key) if target && target_association
+        disassociate_source
         source.update_attribute(source_attribute, Set[hash_key])
       end
 
+      # @private
       def disassociate(_hash_key = nil)
         source.update_attribute(source_attribute, nil)
       end
@@ -82,7 +109,7 @@ module Dynamoid #:nodoc:
       def find_target
         return if source_ids.empty?
 
-        target_class.find(source_ids.first)
+        target_class.find(source_ids.first, raise_error: false)
       end
 
       def target=(object)

data/lib/dynamoid/components.rb

@@ -3,6 +3,7 @@
 module Dynamoid
   # All modules that a Document is composed of are defined in this
   # module, to keep the document class from getting too cluttered.
+  # @private
   module Components
     extend ActiveSupport::Concern
 
@@ -14,6 +15,7 @@ module Dynamoid
 
       before_create :set_created_at
       before_save :set_updated_at
+      before_save :set_expires_field
       after_initialize :set_inheritance_field
     end
 

data/lib/dynamoid/config.rb

@@ -2,14 +2,23 @@
 
 require 'uri'
 require 'logger'
-require 'null_logger'
 require 'dynamoid/config/options'
 require 'dynamoid/config/backoff_strategies/constant_backoff'
 require 'dynamoid/config/backoff_strategies/exponential_backoff'
 
 module Dynamoid
   # Contains all the basic configuration information required for Dynamoid: both sensible defaults and required fields.
+  # @private
   module Config
+    # @since 3.3.1
+    DEFAULT_NAMESPACE = if defined?(Rails)
+                          klass = Rails.application.class
+                          app_name = Rails::VERSION::MAJOR >= 6 ? klass.module_parent_name : klass.parent_name
+                          "dynamoid_#{app_name}_#{Rails.env}"
+                        else
+                          'dynamoid'
+                        end
+
     extend self
 
     extend Options
@@ -17,11 +26,13 @@ module Dynamoid
 
     # All the default options.
     option :adapter, default: 'aws_sdk_v3'
-    option :namespace, default: defined?(Rails) ? "dynamoid_#{Rails.application.class.parent_name}_#{Rails.env}" : 'dynamoid'
+    option :namespace, default: DEFAULT_NAMESPACE
     option :access_key, default: nil
     option :secret_key, default: nil
+    option :credentials, default: nil
     option :region, default: nil
     option :batch_size, default: 100
+    option :capacity_mode, default: nil
     option :read_capacity, default: 100
     option :write_capacity, default: 20
     option :warn_on_scan, default: true
@@ -43,6 +54,7 @@ module Dynamoid
       constant: BackoffStrategies::ConstantBackoff,
       exponential: BackoffStrategies::ExponentialBackoff
     }
+    option :log_formatter, default: nil
     option :http_continue_timeout, default: nil # specify if you'd like to overwrite Aws Configure - default: 1
     option :http_idle_timeout, default: nil     #                                                  - default: 5
     option :http_open_timeout, default: nil     #                                                  - default: 15
@@ -67,7 +79,7 @@ module Dynamoid
     # @since 0.2.0
     def logger=(logger)
       case logger
-      when false, nil then @logger = NullLogger.new
+      when false, nil then @logger = ::Logger.new(nil)
       when true then @logger = default_logger
       else
         @logger = logger if logger.respond_to?(:info)

data/lib/dynamoid/config/backoff_strategies/constant_backoff.rb

@@ -2,6 +2,7 @@
 
 module Dynamoid
   module Config
+    # @private
     module BackoffStrategies
       class ConstantBackoff
         def self.call(sec = 1)

data/lib/dynamoid/config/backoff_strategies/exponential_backoff.rb

@@ -2,6 +2,7 @@
 
 module Dynamoid
   module Config
+    # @private
     module BackoffStrategies
       # Truncated binary exponential backoff algorithm
       # See https://en.wikipedia.org/wiki/Exponential_backoff

data/lib/dynamoid/config/options.rb

@@ -4,6 +4,7 @@
 module Dynamoid
   module Config
     # Encapsulates logic for setting options.
+    # @private
     module Options
       # Get the defaults or initialize a new empty hash.
       #

data/lib/dynamoid/criteria.rb

@@ -7,13 +7,21 @@ module Dynamoid
   module Criteria
     extend ActiveSupport::Concern
 
+    # @private
     module ClassMethods
-      %i[where all first last each record_limit scan_limit batch start scan_index_forward find_by_pages project].each do |meth|
+      %i[where consistent all first last delete_all destroy_all each record_limit scan_limit batch start scan_index_forward find_by_pages project pluck].each do |meth|
         # Return a criteria chain in response to a method that will begin or end a chain. For more information,
         # see Dynamoid::Criteria::Chain.
         #
         # @since 0.2.0
         define_method(meth) do |*args, &blk|
+          # Don't use keywork arguments delegating (with **kw). It works in
+          # different way in different Ruby versions: <= 2.6, 2.7, 3.0 and in some
+          # future 3.x versions. Providing that there are no downstream methods
+          # with keyword arguments in Chain.
+          #
+          # https://eregon.me/blog/2019/11/10/the-delegation-challenge-of-ruby27.html
+
           chain = Dynamoid::Criteria::Chain.new(self)
           if args
             chain.send(meth, *args, &blk)

data/lib/dynamoid/criteria/chain.rb

@@ -22,26 +22,74 @@ module Dynamoid
         @consistent_read = false
         @scan_index_forward = true
 
-        # Honor STI and :type field if it presents
-        type = @source.inheritance_field
-        if @source.attributes.key?(type)
-          @query[:"#{type}.in"] = @source.deep_subclasses.map(&:name) << @source.name
-        end
-
         # we should re-initialize keys detector every time we change query
         @key_fields_detector = KeyFieldsDetector.new(@query, @source)
       end
 
-      # The workhorse method of the criteria chain. Each key in the passed in hash will become another criteria that the
-      # ultimate query must match. A key can either be a symbol or a string, and should be an attribute name or
-      # an attribute name with a range operator.
+      # Returns a chain which is a result of filtering current chain with the specified conditions.
+      #
+      # It accepts conditions in the form of a hash.
+      #
+      #   Post.where(links_count: 2)
+      #
+      # A key could be either string or symbol.
+      #
+      # In order to express conditions other than equality predicates could be used.
+      # Predicate should be added to an attribute name to form a key +'created_at.gt' => Date.yesterday+
+      #
+      # Currently supported following predicates:
+      # - +gt+ - greater than
+      # - +gte+ - greater or equal
+      # - +lt+ - less than
+      # - +lte+ - less or equal
+      # - +ne+ - not equal
+      # - +between+ - an attribute value is greater than the first value and less than the second value
+      # - +in+ - check an attribute in a list of values
+      # - +begins_with+ - check for a prefix in string
+      # - +contains+ - check substring or value in a set or array
+      # - +not_contains+ - check for absence of substring or a value in set or array
+      # - +null+ - attribute doesn't exists in an item
+      # - +not_null+ - attribute exists in an item
+      #
+      # All the predicates match operators supported by DynamoDB's
+      # {ComparisonOperator}[https://docs.aws.amazon.com/amazondynamodb/latest/APIReference/API_Condition.html#DDB-Type-Condition-ComparisonOperator]
+      #
+      #   Post.where('size.gt' => 1000)
+      #   Post.where('size.gte' => 1000)
+      #   Post.where('size.lt' => 35000)
+      #   Post.where('size.lte' => 35000)
+      #   Post.where('author.ne' => 'John Doe')
+      #   Post.where('created_at.between' => [Time.now - 3600, Time.now])
+      #   Post.where('category.in' => ['tech', 'fashion'])
+      #   Post.where('title.begins_with' => 'How long')
+      #   Post.where('tags.contains' => 'Ruby')
+      #   Post.where('tags.not_contains' => 'Ruby on Rails')
+      #   Post.where('legacy_attribute.null' => true)
+      #   Post.where('optional_attribute.not_null' => true)
+      #
+      # There are some limitations for a sort key. Only following predicates
+      # are supported - +gt+, +gte+, +lt+, +lte+, +between+, +begins_with+.
+      #
+      # +where+ without argument will return the current chain.
+      #
+      # Multiple calls can be chained together and conditions will be merged:
+      #
+      #   Post.where('size.gt' => 1000).where('title' => 'some title')
+      #
+      # It's equivalent to:
+      #
+      #   Post.where('size.gt' => 1000, 'title' => 'some title')
+      #
+      # But only one condition can be specified for a certain attribute. The
+      # last specified condition will override all the others. Only condition
+      # 'size.lt' => 200 will be used in following examples:
       #
-      # @example A simple criteria
-      #   where(:name => 'Josh')
+      #   Post.where('size.gt' => 100, 'size.lt' => 200)
+      #   Post.where('size.gt' => 100).where('size.lt' => 200)
       #
-      # @example A more complicated criteria
-      #   where(:name => 'Josh', 'created_at.gt' => DateTime.now - 1.day)
+      # Internally +where+ performs either +Scan+ or +Query+ operation.
       #
+      # @return [Dynamoid::Criteria::Chain]
       # @since 0.2.0
       def where(args)
         detector = IgnoredConditionsDetector.new(args)
@@ -62,11 +110,18 @@ module Dynamoid
         query.update(args.symbolize_keys)
 
         # we should re-initialize keys detector every time we change query
-        @key_fields_detector = KeyFieldsDetector.new(@query, @source)
+        @key_fields_detector = KeyFieldsDetector.new(@query, @source, forced_index_name: @forced_index_name)
 
         self
       end
 
+      # Turns on strongly consistent reads.
+      #
+      # By default reads are eventually consistent.
+      #
+      #   Post.where('size.gt' => 1000).consistent
+      #
+      # @return [Dynamoid::Criteria::Chain]
       def consistent
         @consistent_read = true
         self
@@ -74,11 +129,39 @@ module Dynamoid
 
       # Returns all the records matching the criteria.
       #
+      # Since +where+ and most of the other methods return a +Chain+
+      # the only way to get a result as a collection is to call the +all+
+      # method. It returns +Enumerator+ which could be used directly or
+      # transformed into +Array+
+      #
+      #   Post.all                            # => Enumerator
+      #   Post.where(links_count: 2).all      # => Enumerator
+      #   Post.where(links_count: 2).all.to_a # => Array
+      #
+      # When the result set is too large DynamoDB divides it into separate
+      # pages.  While an enumerator iterates over the result models each page
+      # is loaded lazily. So even an extra large result set can be loaded and
+      # processed with considerably small memory footprint and throughput
+      # consumption.
+      #
+      # @return [Enumerator::Lazy]
       # @since 0.2.0
       def all
         records
       end
 
+      # Returns the actual number of items in a table matching the criteria.
+      #
+      #   Post.where(links_count: 2).count
+      #
+      # Internally it uses either `Scan` or `Query` DynamoDB's operation so it
+      # costs like all the matching items were read from a table.
+      #
+      # The only difference is that items are read by DynemoDB but not actually
+      # loaded on the client side. DynamoDB returns only count of items after
+      # filtering.
+      #
+      # @return [Integer]
       def count
         if @key_fields_detector.key_present?
           count_via_query
@@ -87,27 +170,72 @@ module Dynamoid
         end
       end
 
-      # Returns the last fetched record matched the criteria
-      # Enumerable doesn't implement `last`, only `first`
-      # So we have to implement it ourselves
+      # Returns the first item matching the criteria.
+      #
+      #   Post.where(links_count: 2).first
+      #
+      # Applies `record_limit(1)` to ensure only a single record is fetched
+      # when no non-key conditions are present and `scan_limit(1)` when no
+      # conditions are present at all.
+      #
+      # If used without criteria it just returns the first item of some
+      # arbitrary order.
       #
+      #   Post.first
+      #
+      # @return [Model|nil]
+      def first(*args)
+        n = args.first || 1
+
+        return scan_limit(n).to_a.first(*args) if @query.blank?
+        return super if @key_fields_detector.non_key_present?
+
+        record_limit(n).to_a.first(*args)
+      end
+
+      # Returns the last item matching the criteria.
+      #
+      #   Post.where(links_count: 2).last
+      #
+      # DynamoDB doesn't support ordering by some arbitrary attribute except a
+      # sort key. So this method is mostly useful during development and
+      # testing.
+      #
+      # If used without criteria it just returns the last item of some arbitrary order.
+      #
+      #   Post.last
+      #
+      # It isn't efficient from the performance point of view as far as it reads and
+      # loads all the filtered items from DynamoDB.
+      #
+      # @return [Model|nil]
       def last
         all.to_a.last
       end
 
-      # Destroys all the records matching the criteria.
+      # Deletes all the items matching the criteria.
+      #
+      #   Post.where(links_count: 2).delete_all
+      #
+      # If called without criteria then it deletes all the items in a table.
       #
+      #   Post.delete_all
+      #
+      # It loads all the items either with +Scan+ or +Query+ operation and
+      # deletes them in batch with +BatchWriteItem+ operation. +BatchWriteItem+
+      # is limited by request size and items count so it's quite possible the
+      # deletion will require several +BatchWriteItem+ calls.
       def delete_all
         ids = []
         ranges = []
 
         if @key_fields_detector.key_present?
-          Dynamoid.adapter.query(source.table_name, range_query).flat_map{ |i| i }.collect do |hash|
+          Dynamoid.adapter.query(source.table_name, range_query).flat_map { |i| i }.collect do |hash|
             ids << hash[source.hash_key.to_sym]
             ranges << hash[source.range_key.to_sym] if source.range_key
           end
         else
-          Dynamoid.adapter.scan(source.table_name, scan_query, scan_opts).flat_map{ |i| i }.collect do |hash|
+          Dynamoid.adapter.scan(source.table_name, scan_query, scan_opts).flat_map { |i| i }.collect do |hash|
             ids << hash[source.hash_key.to_sym]
             ranges << hash[source.range_key.to_sym] if source.range_key
           end
@@ -117,53 +245,259 @@ module Dynamoid
       end
       alias destroy_all delete_all
 
-      # The record limit is the limit of evaluated records returned by the
-      # query or scan.
+      # Set the record limit.
+      #
+      # The record limit is the limit of evaluated items returned by the
+      # +Query+ or +Scan+. In other words it's how many items should be
+      # returned in response.
+      #
+      #   Post.where(links_count: 2).record_limit(1000) # => 1000 models
+      #   Post.record_limit(1000)                       # => 1000 models
+      #
+      # It could be very inefficient in terms of HTTP requests in pathological
+      # cases. DynamoDB doesn't support out of the box the limits for items
+      # count after filtering. So it's possible to make a lot of HTTP requests
+      # to find items matching criteria and skip not matching. It means that
+      # the cost (read capacity units) is unpredictable.
+      #
+      # Because of such issues with performance and cost it's mostly useful in
+      # development and testing.
+      #
+      # When called without criteria it works like +scan_limit+.
+      #
+      # @return [Dynamoid::Criteria::Chain]
       def record_limit(limit)
         @record_limit = limit
         self
       end
 
-      # The scan limit which is the limit of records that DynamoDB will
-      # internally query or scan. This is different from the record limit
-      # as with filtering DynamoDB may look at N scanned records but return 0
-      # records if none pass the filter.
+      # Set the scan limit.
+      #
+      # The scan limit is the limit of records that DynamoDB will internally
+      # read with +Query+ or +Scan+. It's different from the record limit as
+      # with filtering DynamoDB may look at N scanned items but return 0
+      # items if none passes the filter. So it can return less items than was
+      # specified with the limit.
+      #
+      #   Post.where(links_count: 2).scan_limit(1000)   # => 850 models
+      #   Post.scan_limit(1000)                         # => 1000 models
+      #
+      # By contrast with +record_limit+ the cost (read capacity units) and
+      # performance is predictable.
+      #
+      # When called without criteria it works like +record_limit+.
+      #
+      # @return [Dynamoid::Criteria::Chain]
       def scan_limit(limit)
         @scan_limit = limit
         self
       end
 
+      # Set the batch size.
+      #
+      # The batch size is a number of items which will be lazily loaded one by one.
+      # When the batch size is set then items will be loaded batch by batch of
+      # the specified size instead of relying on the default paging mechanism
+      # of DynamoDB.
+      #
+      #   Post.where(links_count: 2).batch(1000).all.each do |post|
+      #     # process a post
+      #   end
+      #
+      # It's useful to limit memory usage or throughput consumption
+      #
+      # @return [Dynamoid::Criteria::Chain]
       def batch(batch_size)
         @batch_size = batch_size
         self
       end
 
+      # Set the start item.
+      #
+      # When the start item is set the items will be loaded starting right
+      # after the specified item.
+      #
+      #   Post.where(links_count: 2).start(post)
+      #
+      # It can be used to implement an own pagination mechanism.
+      #
+      #   Post.where(author_id: author_id).start(last_post).scan_limit(50)
+      #
+      # The specified start item will not be returned back in a result set.
+      #
+      # Actually it doesn't need all the item attributes to start - an item may
+      # have only the primary key attributes (partition and sort key if it's
+      # declared).
+      #
+      #   Post.where(links_count: 2).start(Post.new(id: id))
+      #
+      # It also supports a +Hash+ argument with the keys attributes - a
+      # partition key and a sort key (if it's declared).
+      #
+      #   Post.where(links_count: 2).start(id: id)
+      #
+      # @return [Dynamoid::Criteria::Chain]
       def start(start)
         @start = start
         self
       end
 
+      # Reverse the sort order.
+      #
+      # By default the sort order is ascending (by the sort key value). Set a
+      # +false+ value to reverse the order.
+      #
+      #   Post.where(id: id, 'views_count.gt' => 1000).scan_index_forward(false)
+      #
+      # It works only for queries with a partition key condition e.g. +id:
+      # 'some-id'+ which internally performs +Query+ operation.
+      #
+      # @return [Dynamoid::Criteria::Chain]
       def scan_index_forward(scan_index_forward)
         @scan_index_forward = scan_index_forward
         self
       end
 
-      # Allows you to use the results of a search as an enumerable over the results found.
+      # Force the index name to use for queries.
+      #
+      # By default allows the library to select the most appropriate index.
+      # Sometimes you have more than one index which will fulfill your query's
+      # needs. When this case occurs you may want to force an order. This occurs
+      # when you are searching by hash key, but not specifying a range key.
+      #
+      #  class Comment
+      #    include Dynamoid::Document
+      #
+      #    table key: :post_id
+      #    range_key :author_id
+      #
+      #    field :post_date, :datetime
+      #
+      #    global_secondary_index name: :time_sorted_comments, hash_key: :post_id, range_key: post_date, projected_attributes: :all
+      #  end
+      #
+      #
+      #   Comment.where(post_id: id).with_index(:time_sorted_comments).scan_index_forward(false)
+      #
+      # @return [Dynamoid::Criteria::Chain]
+      def with_index(index_name)
+        raise Dynamoid::Errors::InvalidIndex, "Unknown index #{index_name}" unless @source.find_index_by_name(index_name)
+
+        @forced_index_name = index_name
+        @key_fields_detector = KeyFieldsDetector.new(@query, @source, forced_index_name: index_name)
+        self
+      end
+
+      # Allows to use the results of a search as an enumerable over the results
+      # found.
+      #
+      #   Post.each do |post|
+      #   end
+      #
+      #   Post.all.each do |post|
+      #   end
+      #
+      #   Post.where(links_count: 2).each do |post|
+      #   end
+      #
+      # It works similar to the +all+ method so results are loaded lazily.
       #
       # @since 0.2.0
       def each(&block)
         records.each(&block)
       end
 
+      # Iterates over the pages returned by DynamoDB.
+      #
+      # DynamoDB has its own paging machanism and divides a large result set
+      # into separate pages. The +find_by_pages+ method provides access to
+      # these native DynamoDB pages.
+      #
+      # The pages are loaded lazily.
+      #
+      #   Post.where('views_count.gt' => 1000).find_by_pages do |posts, options|
+      #     # process posts
+      #   end
+      #
+      # It passes as block argument an +Array+ of models and a Hash with options.
+      #
+      # Options +Hash+ contains only one option +:last_evaluated_key+. The last
+      # evaluated key is a Hash with key attributes of the last item processed by
+      # DynamoDB. It can be used to resume querying using the +start+ method.
+      #
+      #   posts, options = Post.where('views_count.gt' => 1000).find_by_pages.first
+      #   last_key = options[:last_evaluated_key]
+      #
+      #   # ...
+      #
+      #   Post.where('views_count.gt' => 1000).start(last_key).find_by_pages do |posts, options|
+      #   end
+      #
+      # If it's called without a block then it returns an +Enumerator+.
+      #
+      #   enum = Post.where('views_count.gt' => 1000).find_by_pages
+      #
+      #   enum.each do |posts, options|
+      #     # process posts
+      #   end
+      #
+      # @return [Enumerator::Lazy]
       def find_by_pages(&block)
         pages.each(&block)
       end
 
+      # Select only specified fields.
+      #
+      # It takes one or more field names and returns a collection of models with only
+      # these fields set.
+      #
+      #   Post.where('views_count.gt' => 1000).select(:title)
+      #   Post.where('views_count.gt' => 1000).select(:title, :created_at)
+      #   Post.select(:id)
+      #
+      # It can be used to avoid loading large field values and to decrease a
+      # memory footprint.
+      #
+      # @return [Dynamoid::Criteria::Chain]
       def project(*fields)
         @project = fields.map(&:to_sym)
         self
       end
 
+      # Select only specified fields.
+      #
+      # It takes one or more field names and returns an array of either values
+      # or arrays of values.
+      #
+      #   Post.pluck(:id)                   # => ['1', '2']
+      #   Post.pluck(:title, :title)        # => [['1', 'Title #1'], ['2', 'Title#2']]
+      #
+      #   Post.where('views_count.gt' => 1000).pluck(:title)
+      #
+      # There are some differences between +pluck+ and +project+. +pluck+
+      # - doesn't instantiate models
+      # - it isn't chainable and returns +Array+ instead of +Chain+
+      #
+      # It deserializes values if a field type isn't supported by DynamoDB natively.
+      #
+      # It can be used to avoid loading large field values and to decrease a
+      # memory footprint.
+      #
+      # @return [Array]
+      def pluck(*args)
+        fields = args.map(&:to_sym)
+        @project = fields
+
+        if fields.many?
+          items.map do |item|
+            fields.map { |key| Undumping.undump_field(item[key], source.attributes[key]) }
+          end.to_a
+        else
+          key = fields.first
+          items.map { |item| Undumping.undump_field(item[key], source.attributes[key]) }.to_a
+        end
+      end
+
       private
 
       # The actual records referenced by the association.
@@ -172,7 +506,12 @@ module Dynamoid
       #
       # @since 0.2.0
       def records
-        pages.lazy.flat_map { |i| i }
+        pages.lazy.flat_map { |items, _| items }
+      end
+
+      # Raw items like they are stored before type casting
+      def items
+        raw_pages.lazy.flat_map { |items, _| items }
       end
 
       # Arrays of records, sized based on the actual pages produced by DynamoDB
@@ -181,11 +520,19 @@ module Dynamoid
       #
       # @since 3.1.0
       def pages
+        raw_pages.lazy.map do |items, options|
+          models = items.map { |i| source.from_database(i) }
+          [models, options]
+        end.each
+      end
+
+      # Pages of items before type casting
+      def raw_pages
         if @key_fields_detector.key_present?
-          pages_via_query
+          raw_pages_via_query
         else
           issue_scan_warning if Dynamoid::Config.warn_on_scan && query.present?
-          pages_via_scan
+          raw_pages_via_scan
         end
       end
 
@@ -194,13 +541,12 @@ module Dynamoid
       # @return [Enumerator] an iterator of the found pages. An array of records
       #
       # @since 3.1.0
-      def pages_via_query
+      def raw_pages_via_query
         Enumerator.new do |y|
           Dynamoid.adapter.query(source.table_name, range_query).each do |items, metadata|
-            page = items.map { |h| source.from_database(h) }
             options = metadata.slice(:last_evaluated_key)
 
-            y.yield page, options
+            y.yield items, options
           end
         end
       end
@@ -210,13 +556,12 @@ module Dynamoid
       # @return [Enumerator] an iterator of the found pages. An array of records
       #
       # @since 3.1.0
-      def pages_via_scan
+      def raw_pages_via_scan
         Enumerator.new do |y|
           Dynamoid.adapter.scan(source.table_name, scan_query, scan_opts).each do |items, metadata|
-            page = items.map { |h| source.from_database(h) }
             options = metadata.slice(:last_evaluated_key)
 
-            y.yield page, options
+            y.yield items, options
           end
         end
       end
@@ -300,6 +645,13 @@ module Dynamoid
 
       def range_query
         opts = {}
+        query = self.query
+
+        # Honor STI and :type field if it presents
+        if @source.attributes.key?(@source.inheritance_field) &&
+           @key_fields_detector.hash_key.to_sym != @source.inheritance_field.to_sym
+          query.update(sti_condition)
+        end
 
         # Add hash key
         opts[:hash_key] = @key_fields_detector.hash_key
@@ -307,15 +659,7 @@ module Dynamoid
 
         # Add range key
         if @key_fields_detector.range_key
-          opts[:range_key] = @key_fields_detector.range_key
-          if query[@key_fields_detector.range_key].present?
-            value = type_cast_condition_parameter(@key_fields_detector.range_key, query[@key_fields_detector.range_key])
-            opts.update(range_eq: value)
-          end
-
-          query.keys.select { |k| k.to_s =~ /^#{@key_fields_detector.range_key}\./ }.each do |key|
-            opts.merge!(range_hash(key))
-          end
+          add_range_key_to_range_query(query, opts)
         end
 
         (query.keys.map(&:to_sym) - [@key_fields_detector.hash_key.to_sym, @key_fields_detector.range_key.try(:to_sym)])
@@ -332,7 +676,19 @@ module Dynamoid
         opts.merge(query_opts).merge(consistent_opts)
       end
 
-      # TODO casting should be operator aware
+      def add_range_key_to_range_query(query, opts)
+        opts[:range_key] = @key_fields_detector.range_key
+        if query[@key_fields_detector.range_key].present?
+          value = type_cast_condition_parameter(@key_fields_detector.range_key, query[@key_fields_detector.range_key])
+          opts.update(range_eq: value)
+        end
+
+        query.keys.select { |k| k.to_s =~ /^#{@key_fields_detector.range_key}\./ }.each do |key|
+          opts.merge!(range_hash(key))
+        end
+      end
+
+      # TODO: casting should be operator aware
       # e.g. for NULL operator value should be boolean
       # and isn't related to an attribute own type
       def type_cast_condition_parameter(key, value)
@@ -393,6 +749,13 @@ module Dynamoid
       end
 
       def scan_query
+        query = self.query
+
+        # Honor STI and :type field if it presents
+        if sti_condition
+          query.update(sti_condition)
+        end
+
         {}.tap do |opts|
           query.keys.map(&:to_sym).each do |key|
             if key.to_s.include?('.')
@@ -415,6 +778,18 @@ module Dynamoid
         opts[:project] = @project
         opts
       end
+
+      def sti_condition
+        condition = {}
+        type = @source.inheritance_field
+
+        if @source.attributes.key?(type)
+          class_names = @source.deep_subclasses.map(&:name) << @source.name
+          condition[:"#{type}.in"] = class_names
+        end
+
+        condition
+      end
     end
   end
 end