chewy 0.8.4 → 5.1.0

data/lib/chewy/type/adapter/orm.rb

@@ -6,16 +6,15 @@ module Chewy
       class Orm < Base
         attr_reader :default_scope
 
-        def initialize *args
-          @options = args.extract_options!
-          class_or_relation = args.first
-          if class_or_relation.is_a?(relation_class)
-            @target = model_of_relation(class_or_relation)
-            @default_scope = class_or_relation
+        def initialize(target, **options)
+          if target.is_a?(relation_class)
+            @target = model_of_relation(target)
+            @default_scope = target
           else
-            @target = class_or_relation
+            @target = target
             @default_scope = all_scope
           end
+          @options = options
           cleanup_default_scope!
         end
 
@@ -23,12 +22,12 @@ module Chewy
           @name ||= (options[:name].presence || target.name).to_s.camelize.demodulize
         end
 
-        def identify collection
+        def identify(collection)
           if collection.is_a?(relation_class)
-            pluck_ids(collection)
+            pluck(collection)
           else
             Array.wrap(collection).map do |entity|
-              entity.is_a?(object_class) ? entity.public_send(primary_key) : entity
+              entity.respond_to?(primary_key) ? entity.public_send(primary_key) : entity
             end
           end
         end
@@ -49,7 +48,7 @@ module Chewy
         #
         # Method handles destroyed objects as well. In case of objects ORM scope
         # or array passed, objects, responding with true to `destroyed?` method will be deleted
-        # from index. In case of ids array passed - documents with missing records ids will be
+        # from index. In case of ids array passed - documents with missing source object ids will be
         # deleted from index:
         #
         #   users = User.all
@@ -73,48 +72,66 @@ module Chewy
         #   # or
         #   UsersIndex::User.import users.map(&:id) # user ids will be deleted from index
         #
-        def import *args, &block
-          import_options = args.extract_options!
-          batch_size = import_options[:batch_size] || BATCH_SIZE
-
-          collection = args.empty? ? default_scope :
-            (args.one? && args.first.is_a?(relation_class) ? args.first : args.flatten.compact)
+        def import(*args, &block)
+          collection, options = import_args(*args)
 
           if collection.is_a?(relation_class)
-            import_scope(collection, batch_size, &block)
+            import_scope(collection, options, &block)
           else
-            import_objects(collection, batch_size, &block)
+            import_objects(collection, options, &block)
           end
         end
 
-        def load *args
-          load_options = args.extract_options!
-          objects = args.flatten
+        def import_fields(*args, &block)
+          return enum_for(:import_fields, *args) unless block_given?
+
+          collection, options = import_args(*args)
+
+          if options[:fields].present? || collection.is_a?(relation_class)
+            collection = all_scope_where_ids_in(identify(collection)) unless collection.is_a?(relation_class)
+            pluck_in_batches(collection, options.slice(:fields, :batch_size, :typecast), &block)
+          else
+            identify(collection).each_slice(options[:batch_size]) do |batch|
+              yield batch
+            end
+          end
+        end
+        alias_method :import_references, :import_fields
 
-          additional_scope = load_options[load_options[:_type].type_name.to_sym].try(:[], :scope) || load_options[:scope]
+        def load(ids, **options)
+          scope = all_scope_where_ids_in(ids)
+          additional_scope = options[options[:_type].type_name.to_sym].try(:[], :scope) || options[:scope]
 
-          scope = all_scope_where_ids_in(objects.map(&primary_key))
-          loaded_objects = load_scope_objects(scope, additional_scope).index_by { |object| object.public_send(primary_key).to_s }
+          loaded_objects = load_scope_objects(scope, additional_scope)
+            .index_by do |object|
+              object.public_send(primary_key).to_s
+            end
 
-          objects.map { |object| loaded_objects[object.public_send(primary_key).to_s] }
+          ids.map { |id| loaded_objects[id.to_s] }
         end
 
       private
 
-        def import_objects(collection, batch_size)
-          hash = Hash[identify(collection).zip(collection)]
+        def import_objects(collection, options)
+          collection_ids = identify(collection)
+          hash = Hash[collection_ids.map(&:to_s).zip(collection)]
+
+          indexed = collection_ids.each_slice(options[:batch_size]).map do |ids|
+            batch = if options[:raw_import]
+              raw_default_scope_where_ids_in(ids, options[:raw_import])
+            else
+              default_scope_where_ids_in(ids)
+            end
 
-          indexed = hash.keys.each_slice(batch_size).map do |ids|
-            batch = default_scope_where_ids_in(ids)
             if batch.empty?
               true
             else
-              identify(batch).each { |id| hash.delete(id) }
+              batch.each { |object| hash.delete(object.send(primary_key).to_s) }
               yield grouped_objects(batch)
             end
           end.all?
 
-          deleted = hash.keys.each_slice(batch_size).map do |group|
+          deleted = hash.keys.each_slice(options[:batch_size]).map do |group|
             yield delete: hash.values_at(*group)
           end.all?
 
@@ -137,7 +154,7 @@ module Chewy
           target.where(nil)
         end
 
-        def model_of_relation relation
+        def model_of_relation(relation)
           relation.klass
         end
 
@@ -150,6 +167,25 @@ module Chewy
             scope
           end
         end
+
+        def grouped_objects(objects)
+          options[:delete_if] ? super : {index: objects.to_a}
+        end
+
+        def import_args(*args)
+          options = args.extract_options!
+          options[:batch_size] ||= BATCH_SIZE
+
+          collection = if args.empty?
+            default_scope
+          elsif args.one? && args.first.is_a?(relation_class)
+            args.first
+          else
+            args.flatten.compact
+          end
+
+          [collection, options]
+        end
       end
     end
   end

data/lib/chewy/type/adapter/sequel.rb

@@ -16,45 +16,63 @@ module Chewy
       private
 
         def cleanup_default_scope!
-          if Chewy.logger && @default_scope != @default_scope.unordered.unlimited
-            Chewy.logger.warn('Default type scope order, limit and offset are ignored and will be nullified')
-          end
+          Chewy.logger.warn('Default type scope order, limit and offset are ignored and will be nullified') if Chewy.logger && @default_scope != @default_scope.unordered.unlimited
 
           @default_scope = @default_scope.unordered.unlimited
         end
 
-        def import_scope(scope, batch_size)
-          scope = scope.unordered.order(::Sequel.asc(primary_key)).limit(batch_size)
-
-          ids = pluck_ids(scope)
-          result = true
-
-          while ids.present?
-            result &= yield grouped_objects(default_scope_where_ids_in(ids).all)
-            break if ids.size < batch_size
-            ids = pluck_ids(scope.where { |o| o.__send__(primary_key) > ids.last })
+        def import_scope(scope, options)
+          pluck_in_batches(scope, options.slice(:batch_size)).inject(true) do |result, ids|
+            result & yield(grouped_objects(default_scope_where_ids_in(ids).all))
           end
-
-          result
         end
 
         def primary_key
           target.primary_key
         end
 
+        def full_column_name(column)
+          ::Sequel.qualify(target.table_name, column)
+        end
+
         def all_scope
           target.dataset
         end
 
-        def pluck_ids(scope)
-          scope.distinct.select_map(primary_key)
+        def target_columns
+          @target_columns ||= target.columns.to_set
+        end
+
+        def pluck(scope, fields: [])
+          fields = fields.map(&:to_sym).unshift(primary_key).map do |column|
+            target_columns.include?(column) ? full_column_name(column) : column
+          end
+          scope.distinct.select_map(fields.one? ? fields.first : fields)
+        end
+
+        def pluck_in_batches(scope, fields: [], batch_size: nil, **options)
+          return enum_for(:pluck_in_batches, scope, fields: fields, batch_size: batch_size, **options) unless block_given?
+
+          scope = scope.unordered.order(full_column_name(primary_key).asc).limit(batch_size)
+
+          ids = pluck(scope, fields: fields)
+          count = 0
+
+          while ids.present?
+            yield ids
+            break if ids.size < batch_size
+            last_id = ids.last.is_a?(Array) ? ids.last.first : ids.last
+            ids = pluck(scope.where { |_o| full_column_name(primary_key) > last_id }, fields: fields)
+          end
+
+          count
         end
 
         def scope_where_ids_in(scope, ids)
-          scope.where(primary_key => Array.wrap(ids))
+          scope.where(full_column_name(primary_key) => Array.wrap(ids))
         end
 
-        def model_of_relation relation
+        def model_of_relation(relation)
           relation.model
         end
 

data/lib/chewy/type/crutch.rb

@@ -9,9 +9,10 @@ module Chewy
       end
 
       class Crutches
-        def initialize type, collection
-          @type, @collection = type, collection
-          @type._crutches.keys.each do |name|
+        def initialize(type, collection)
+          @type = type
+          @collection = collection
+          @type._crutches.each_key do |name|
             singleton_class.class_eval <<-METHOD, __FILE__, __LINE__ + 1
               def #{name}
                 @#{name} ||= @type._crutches[:#{name}].call @collection
@@ -22,7 +23,7 @@ module Chewy
       end
 
       module ClassMethods
-        def crutch name, &block
+        def crutch(name, &block)
           self._crutches = _crutches.merge(name.to_sym => block)
         end
       end

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