chewy 0.9.0 → 5.2.0

data/lib/chewy/type/import/bulk_builder.rb

@@ -0,0 +1,122 @@
+module Chewy
+  class Type
+    module Import
+      # This class purpose is to build ES client-acceptable bulk
+      # request body from the passed objects for index and deletion.
+      # It handles parent-child relationships as well by fetching
+      # existing documents from ES, taking their `_parent` field and
+      # using it in the bulk body.
+      # If fields are passed - it creates partial update entries except for
+      # the cases when the type has parent and parent_id has been changed.
+      class BulkBuilder
+        # @param type [Chewy::Type] desired type
+        # @param index [Array<Object>] objects to index
+        # @param delete [Array<Object>] objects or ids to delete
+        # @param fields [Array<Symbol, String>] and array of fields for documents update
+        def initialize(type, index: [], delete: [], fields: [])
+          @type = type
+          @index = index
+          @delete = delete
+          @fields = fields.map!(&:to_sym)
+        end
+
+        # Returns ES API-ready bulk requiest body.
+        # @see https://github.com/elastic/elasticsearch-ruby/blob/master/elasticsearch-api/lib/elasticsearch/api/actions/bulk.rb
+        # @return [Array<Hash>] bulk body
+        def bulk_body
+          @bulk_body ||= @index.flat_map(&method(:index_entry)).concat(
+            @delete.flat_map(&method(:delete_entry))
+          )
+        end
+
+        # The only purpose of this method is to cache document ids for
+        # all the passed object for index to avoid ids recalculation.
+        #
+        # @return [Hash[String => Object]] an ids-objects index hash
+        def index_objects_by_id
+          @index_objects_by_id ||= index_object_ids.invert.stringify_keys!
+        end
+
+      private
+
+        def crutches
+          @crutches ||= Chewy::Type::Crutch::Crutches.new @type, @index
+        end
+
+        def parents
+          return unless type_root.parent_id
+
+          @parents ||= begin
+            ids = @index.map do |object|
+              object.respond_to?(:id) ? object.id : object
+            end
+            ids.concat(@delete.map do |object|
+              object.respond_to?(:id) ? object.id : object
+            end)
+            @type.filter(ids: {values: ids}).order('_doc').pluck(:_id, :_parent).to_h
+          end
+        end
+
+        def index_entry(object)
+          entry = {}
+          entry[:_id] = index_object_ids[object] if index_object_ids[object]
+
+          if parents
+            entry[:parent] = type_root.compose_parent(object)
+            parent = entry[:_id].present? && parents[entry[:_id].to_s]
+          end
+
+          if parent && entry[:parent].to_s != parent
+            entry[:data] = @type.compose(object, crutches)
+            [{delete: entry.except(:data).merge(parent: parent)}, {index: entry}]
+          elsif @fields.present?
+            return [] unless entry[:_id]
+            entry[:data] = {doc: @type.compose(object, crutches, fields: @fields)}
+            [{update: entry}]
+          else
+            entry[:data] = @type.compose(object, crutches)
+            [{index: entry}]
+          end
+        end
+
+        def delete_entry(object)
+          entry = {}
+          entry[:_id] = entry_id(object)
+          entry[:_id] ||= object.as_json
+
+          return [] if entry[:_id].blank?
+
+          if parents
+            parent = entry[:_id].present? && parents[entry[:_id].to_s]
+            return [] unless parent
+            entry[:parent] = parent
+          end
+
+          [{delete: entry}]
+        end
+
+        def entry_id(object)
+          if type_root.id
+            type_root.compose_id(object)
+          else
+            id = object.id if object.respond_to?(:id)
+            id ||= object[:id] || object['id'] if object.is_a?(Hash)
+            id = id.to_s if defined?(BSON) && id.is_a?(BSON::ObjectId)
+            id
+          end
+        end
+
+        def index_object_ids
+          @index_object_ids ||= @index.each_with_object({}) do |object, result|
+            id = entry_id(object)
+            result[object] = id if id.present?
+          end
+        end
+
+        def type_root
+          @type_root ||= @type.root
+        end
+      end
+    end
+  end
+end

data/lib/chewy/type/import/bulk_request.rb

@@ -0,0 +1,78 @@
+module Chewy
+  class Type
+    module Import
+      # Adds additional features to elasticsearch-api bulk method:
+      # * supports Chewy index suffix if necessary;
+      # * supports bulk_size, devides the passed body in chunks
+      #   and peforms a separate request for each chunk;
+      # * returns only errored document entries from the response
+      #   if any present.
+      #
+      # @see https://github.com/elastic/elasticsearch-ruby/blob/master/elasticsearch-api/lib/elasticsearch/api/actions/bulk.rb
+      class BulkRequest
+        # @param type [Chewy::Type] a type for the request
+        # @param suffix [String] an index name optional suffix
+        # @param bulk_size [Integer] bulk size in bytes
+        # @param bulk_options [Hash] options passed to the elasticsearch-api bulk method
+        def initialize(type, suffix: nil, bulk_size: nil, **bulk_options)
+          @type = type
+          @suffix = suffix
+          @bulk_size = bulk_size - 1.kilobyte if bulk_size # 1 kilobyte for request header and newlines
+          @bulk_options = bulk_options
+
+          raise ArgumentError, '`bulk_size` can\'t be less than 1 kilobyte' if @bulk_size && @bulk_size <= 0
+        end
+
+        # Performs a bulk request with the passed body, returns empty
+        # array if everything is fine and array filled with errored
+        # document entries if something went wrong.
+        #
+        # @param body [Array<Hash>] a standard bulk request body
+        # @return [Array<Hash>] an array of bulk errors
+        def perform(body)
+          return [] if body.blank?
+
+          request_bodies(body).each_with_object([]) do |request_body, results|
+            response = @type.client.bulk request_base.merge(body: request_body) if request_body.present?
+
+            next unless response.try(:[], 'errors')
+
+            response_items = (response.try(:[], 'items') || [])
+              .select { |item| item.values.first['error'] }
+            results.concat(response_items)
+          end
+        end
+
+      private
+
+        def request_base
+          @request_base ||= {
+            index: @type.index_name(suffix: @suffix),
+            type: @type.type_name
+          }.merge!(@bulk_options)
+        end
+
+        def request_bodies(body)
+          if @bulk_size
+            serializer = ::Elasticsearch::API.serializer
+            pieces = body.each_with_object(['']) do |piece, result|
+              operation, meta = piece.to_a.first
+              data = meta.delete(:data)
+              piece = serializer.dump(operation => meta)
+              piece << "\n" << serializer.dump(data) if data.present?
+
+              if result.last.bytesize + piece.bytesize > @bulk_size
+                result.push(piece)
+              else
+                result[-1].blank? ? (result[-1] = piece) : (result[-1] << "\n" << piece)
+              end
+            end
+            pieces.each { |piece| piece << "\n" }
+          else
+            [body]
+          end
+        end
+      end
+    end
+  end
+end

data/lib/chewy/type/import/journal_builder.rb

@@ -0,0 +1,45 @@
+module Chewy
+  class Type
+    module Import
+      class JournalBuilder
+        def initialize(type, index: [], delete: [])
+          @type = type
+          @index = index
+          @delete = delete
+        end
+
+        def bulk_body
+          Chewy::Type::Import::BulkBuilder.new(
+            Chewy::Stash::Journal::Journal,
+            index: [
+              entries(:index, @index),
+              entries(:delete, @delete)
+            ].compact
+          ).bulk_body.each do |item|
+            item.values.first.merge!(
+              _index: Chewy::Stash::Journal.index_name,
+              _type: Chewy::Stash::Journal::Journal.type_name
+            )
+          end
+        end
+
+      private
+
+        def entries(action, objects)
+          return unless objects.present?
+          {
+            index_name: @type.index.derivable_name,
+            type_name: @type.type_name,
+            action: action,
+            references: identify(objects).map { |item| Base64.encode64(::Elasticsearch::API.serializer.dump(item)) },
+            created_at: Time.now.utc
+          }
+        end
+
+        def identify(objects)
+          @type.adapter.identify(objects)
+        end
+      end
+    end
+  end
+end

data/lib/chewy/type/import/routine.rb

@@ -0,0 +1,138 @@
+module Chewy
+  class Type
+    module Import
+      # This class performs the import routine for the options and objects given.
+      #
+      # 0. Create target and journal indexes if needed.
+      # 1. Iterate over all the passed objects in batches.
+      # 2. For each batch {#process} method is called:
+      #   * creates a bulk request body;
+      #   * appends journal entries for the current batch to the request body;
+      #   * prepends a leftovers bulk to the request body, which is calculated
+      #     basing on the previous iteration errors;
+      #   * performs the bulk request;
+      #   * composes new leftovers bulk for the next iteration basing on the response errors if `update_failover` is true;
+      #   * appends the rest of unfixable errors to the instance level errors array.
+      # 4. Perform the request for the last leftovers bulk if present using {#extract_leftovers}.
+      # 3. Return the result errors array.
+      #
+      # At the moment, it tries to restore only from the partial document update errors in cases
+      # when the document doesn't exist only if `update_failover` option is true. In order to
+      # restore, it indexes such an objects completely on the next iteration.
+      #
+      # @see Chewy::Type::Import::ClassMethods#import
+      class Routine
+        BULK_OPTIONS = %i[
+          suffix bulk_size
+          refresh timeout fields pipeline
+          consistency replication
+          wait_for_active_shards routing _source _source_exclude _source_include
+        ].freeze
+
+        DEFAULT_OPTIONS = {
+          refresh: true,
+          update_fields: [],
+          update_failover: true,
+          batch_size: Chewy::Type::Adapter::Base::BATCH_SIZE
+        }.freeze
+
+        attr_reader :options, :parallel_options, :errors, :stats, :leftovers
+
+        # Basically, processes passed options, extracting bulk request specific options.
+        # @param type [Chewy::Type] chewy type
+        # @param options [Hash] import options, see {Chewy::Type::Import::ClassMethods#import}
+        def initialize(type, **options)
+          @type = type
+          @options = options
+          @options.reverse_merge!(@type._default_import_options)
+          @options.reverse_merge!(journal: Chewy.configuration[:journal])
+          @options.reverse_merge!(DEFAULT_OPTIONS)
+          @bulk_options = @options.slice(*BULK_OPTIONS)
+          @parallel_options = @options.delete(:parallel)
+          if @parallel_options && !@parallel_options.is_a?(Hash)
+            @parallel_options = if @parallel_options.is_a?(Integer)
+              {in_processes: @parallel_options}
+            else
+              {}
+            end
+          end
+          @errors = []
+          @stats = {}
+          @leftovers = []
+        end
+
+        # Creates the journal index and the type corresponding index if necessary.
+        # @return [Object] whatever
+        def create_indexes!
+          Chewy::Stash::Journal.create if @options[:journal]
+          return if Chewy.configuration[:skip_index_creation_on_import]
+          @type.index.create!(**@bulk_options.slice(:suffix)) unless @type.index.exists?
+        end
+
+        # The main process method. Converts passed objects to thr bulk request body,
+        # appends journal entires, performs this request and handles errors performing
+        # failover procedures if applicable.
+        #
+        # @param index [Array<Object>] any acceptable objects for indexing
+        # @param delete [Array<Object>] any acceptable objects for deleting
+        # @return [true, false] the result of the request, true if no errors
+        def process(index: [], delete: [])
+          bulk_builder = BulkBuilder.new(@type, index: index, delete: delete, fields: @options[:update_fields])
+          bulk_body = bulk_builder.bulk_body
+
+          if @options[:journal]
+            journal_builder = JournalBuilder.new(@type, index: index, delete: delete)
+            bulk_body.concat(journal_builder.bulk_body)
+          end
+
+          bulk_body.unshift(*flush_leftovers)
+
+          perform_bulk(bulk_body) do |response|
+            @leftovers = extract_leftovers(response, bulk_builder.index_objects_by_id)
+            @stats[:index] = @stats[:index].to_i + index.count if index.present?
+            @stats[:delete] = @stats[:delete].to_i + delete.count if delete.present?
+          end
+        end
+
+        # Performs a bulk request for the passed body.
+        #
+        # @param body [Array<Hash>] a standard bulk request body
+        # @return [true, false] the result of the request, true if no errors
+        def perform_bulk(body)
+          response = bulk.perform(body)
+          yield response if block_given?
+          Chewy.wait_for_status
+          @errors.concat(response)
+          response.blank?
+        end
+
+      private
+
+        def flush_leftovers
+          leftovers = @leftovers
+          @leftovers = []
+          leftovers
+        end
+
+        def extract_leftovers(errors, index_objects_by_id)
+          return [] unless @options[:update_fields].present? && @options[:update_failover] && errors.present?
+
+          failed_partial_updates = errors.select do |item|
+            item.keys.first == 'update' && item.values.first['error']['type'] == 'document_missing_exception'
+          end
+          failed_ids_hash = failed_partial_updates.index_by { |item| item.values.first['_id'].to_s }
+          failed_ids_for_reimport = failed_ids_hash.keys & index_objects_by_id.keys
+          errors_to_cleanup = failed_ids_hash.values_at(*failed_ids_for_reimport)
+          errors_to_cleanup.each { |error| errors.delete(error) }
+
+          failed_objects = index_objects_by_id.values_at(*failed_ids_for_reimport)
+          BulkBuilder.new(@type, index: failed_objects).bulk_body
+        end
+
+        def bulk
+          @bulk ||= BulkRequest.new(@type, **@bulk_options)
+        end
+      end
+    end
+  end
+end

data/lib/chewy/type/mapping.rb

@@ -8,6 +8,8 @@ module Chewy
         class_attribute :_templates
         class_attribute :_agg_defs
         self._agg_defs = {}
+        class_attribute :outdated_sync_field
+        self.outdated_sync_field = :updated_at
       end
 
       module ClassMethods
@@ -15,10 +17,11 @@ module Chewy
         # definition. Use it only if you need to pass options for root
         # object mapping, such as `date_detection` or `dynamic_date_formats`
         #
+        # @example
         #   class UsersIndex < Chewy::Index
         #     define_type User do
         #       # root object defined implicitly and optionless for current type
-        #       field :full_name, type: 'string'
+        #       field :full_name, type: 'keyword'
         #     end
         #   end
         #
@@ -26,32 +29,37 @@ module Chewy
         #     define_type Car do
         #       # explicit root definition with additional options
         #       root dynamic_date_formats: ['yyyy-MM-dd'] do
-        #         field :model_name, type: 'string'
+        #         field :model_name, type: 'keyword'
         #       end
         #     end
         #   end
         #
-        def root(options = {}, &block)
-          raise 'Root is already defined' if root_object
-          build_root(options, &block)
+        def root(**options)
+          self.root_object ||= Chewy::Fields::Root.new(type_name, **Chewy.default_root_options.merge(options))
+          root_object.update_options!(**options)
+          yield if block_given?
+          root_object
         end
 
         # Defines mapping field for current type
         #
+        # @example
         #   class UsersIndex < Chewy::Index
         #     define_type User do
         #       # passing all the options to field definition:
-        #       field :full_name, type: 'string', analyzer: 'special'
+        #       field :full_name, analyzer: 'special'
         #     end
         #   end
         #
         # The `type` is optional and defaults to `string` if not defined:
         #
+        # @example
         #   field :full_name
         #
         # Also, multiple fields might be defined with one call and
         # with the same options:
         #
+        # @example
         #   field :first_name, :last_name, analyzer: 'special'
         #
         # The only special option in the field definition
@@ -59,31 +67,35 @@ module Chewy
         # method will be called for the indexed object. Also
         # `:value` might be a proc or indexed object method name:
         #
+        # @example
         #   class User < ActiveRecord::Base
         #     def user_full_name
         #       [first_name, last_name].join(' ')
         #     end
         #   end
         #
-        #   field :full_name, type: 'string', value: :user_full_name
+        #   field :full_name, type: 'keyword', value: :user_full_name
         #
         # The proc evaluates inside the indexed object context if
         # its arity is 0 and in present contexts if there is an argument:
         #
-        #   field :full_name, type: 'string', value: -> { [first_name, last_name].join(' ') }
+        # @example
+        #   field :full_name, type: 'keyword', value: -> { [first_name, last_name].join(' ') }
         #
         #   separator = ' '
-        #   field :full_name, type: 'string', value: ->(user) { [user.first_name, user.last_name].join(separator) }
+        #   field :full_name, type: 'keyword', value: ->(user) { [user.first_name, user.last_name].join(separator) }
         #
         # If array was returned as value - it will be put in index as well.
         #
-        #   field :tags, type: 'string', value: -> { tags.map(&:name) }
+        # @example
+        #   field :tags, type: 'keyword', value: -> { tags.map(&:name) }
         #
         # Fields supports nesting in case of `object` field type. If
         # `user.quiz` will return an array of objects, then result index content
         # will be an array of hashes, if `user.quiz` is not a collection association
         # then just values hash will be put in the index.
         #
+        # @example
         #   field :quiz do
         #     field :question, :answer
         #     field :score, type: 'integer'
@@ -91,6 +103,7 @@ module Chewy
         #
         # Nested fields are composed from nested objects:
         #
+        # @example
         #   field :name, value: -> { name_translations } do
         #     field :ru, value: ->(name) { name['ru'] }
         #     field :en, value: ->(name) { name['en'] }
@@ -99,32 +112,31 @@ module Chewy
         # Of course it is possible to define object fields contents dynamically
         # but make sure evaluation proc returns hash:
         #
+        # @example
         #   field :name, type: 'object', value: -> { name_translations }
         #
         # The special case is multi_field. If type options and block are
         # both present field is treated as a multi-field. In that case field
         # composition changes satisfy elasticsearch rules:
         #
-        #   field :full_name, type: 'string', analyzer: 'name', value: ->{ full_name.try(:strip) } do
+        # @example
+        #   field :full_name, type: 'text', analyzer: 'name', value: ->{ full_name.try(:strip) } do
         #     field :sorted, analyzer: 'sorted'
         #   end
         #
-        def field(*args, &block)
-          options = args.extract_options!
-          build_root
-
+        def field(*args, **options, &block)
           if args.size > 1
-            args.map { |name| field(name, options) }
+            args.map { |name| field(name, **options) }
           else
-            expand_nested(Chewy::Fields::Base.new(args.first, options), &block)
+            expand_nested(Chewy::Fields::Base.new(args.first, **options), &block)
           end
         end
 
         # Defines an aggregation that can be bound to a query or filter
         #
-        #   Suppose that a user has posts and each post has ratings
-        #   avg_post_rating is the mean of all ratings
-        #
+        # @example
+        #   # Suppose that a user has posts and each post has ratings
+        #   # avg_post_rating is the mean of all ratings
         #   class UsersIndex < Chewy::Index
         #     define_type User do
         #       field :posts do
@@ -137,64 +149,68 @@ module Chewy
         #     end
         #   end
         def agg(name, &block)
-          build_root
           self._agg_defs = _agg_defs.merge(name => block)
         end
         alias_method :aggregation, :agg
 
         # Defines dynamic template in mapping root objects
         #
+        # @example
         #   class CarsIndex < Chewy::Index
         #     define_type Car do
-        #       template 'model.*', type: 'string', analyzer: 'special'
+        #       template 'model.*', type: 'text', analyzer: 'special'
         #       field 'model', type: 'object' # here we can put { de: 'Der Mercedes', en: 'Mercedes' }
         #                                     # and template will be applyed to this field
         #     end
         #   end
         #
         # Name for each template is generated with the following
-        # rule: "template_#{dynamic_templates.size + 1}".
+        # rule: `template_#!{dynamic_templates.size + 1}`.
         #
+        # @example Templates
         #   template 'tit*', mapping_hash
         #   template 'title.*', mapping_hash # dot in template causes "path_match" using
         #   template /tit.+/, mapping_hash # using "match_pattern": "regexp"
         #   template /title\..+/, mapping_hash # "\." - escaped dot causes "path_match" using
-        #   template /tit.+/, 'string', mapping_hash # "match_mapping_type" as the optionsl second argument
+        #   template /tit.+/, type: 'text', mapping_hash # "match_mapping_type" as the optionsl second argument
         #   template template42: {match: 'hello*', mapping: {type: 'object'}} # or even pass a template as is
         #
         def template(*args)
-          build_root.dynamic_template(*args)
+          root.dynamic_template(*args)
         end
         alias_method :dynamic_template, :template
 
         # Returns compiled mappings hash for current type
         #
         def mappings_hash
-          root_object ? root_object.mappings_hash : {}
+          root.mappings_hash[type_name.to_sym].present? ? root.mappings_hash : {}
+        end
+
+        # Check whether the type has outdated_sync_field defined with a simple value.
+        #
+        # @return [true, false]
+        def supports_outdated_sync?
+          updated_at_field = root.child_hash[outdated_sync_field] if outdated_sync_field
+          !!updated_at_field && updated_at_field.value.nil?
         end
 
       private
 
-        def expand_nested(field, &block)
+        def expand_nested(field)
+          @_current_field ||= root
+
           if @_current_field
             field.parent = @_current_field
             @_current_field.children.push(field)
           end
 
-          return unless block
+          return unless block_given?
 
           previous_field = @_current_field
           @_current_field = field
           yield
           @_current_field = previous_field
         end
-
-        def build_root(options = {}, &block)
-          return root_object if root_object
-          self.root_object = Chewy::Fields::Root.new(type_name, options)
-          expand_nested(root_object, &block)
-          @_current_field = root_object
-        end
       end
     end
   end