karafka 1.1.0 → 1.3.0

data/lib/karafka/instrumentation/stdout_listener.rb

@@ -0,0 +1,138 @@
+# frozen_string_literal: true
+
+module Karafka
+  module Instrumentation
+    # Default listener that hooks up to our instrumentation and uses its events for logging
+    # It can be removed/replaced or anything without any harm to the Karafka app flow
+    class StdoutListener
+      # Log levels that we use in this particular listener
+      USED_LOG_LEVELS = %i[
+        debug
+        info
+        error
+        fatal
+      ].freeze
+
+      # Logs details about incoming batches and with which consumer we will consume them
+      # @param event [Dry::Events::Event] event details including payload
+      def on_connection_batch_delegator_call(event)
+        consumer = event[:consumer]
+        topic = consumer.topic.name
+        kafka_messages = event[:kafka_batch].messages
+        info(
+          <<~MSG.chomp.tr("\n", ' ')
+            #{kafka_messages.count} messages
+            on #{topic} topic
+            delegated to #{consumer.class}
+          MSG
+        )
+      end
+
+      # Logs details about incoming message and with which consumer we will consume it
+      # @param event [Dry::Events::Event] event details including payload
+      def on_connection_message_delegator_call(event)
+        consumer = event[:consumer]
+        topic = consumer.topic.name
+        info "1 message on #{topic} topic delegated to #{consumer.class}"
+      end
+
+      # Logs details about each received message value deserialization
+      # @param event [Dry::Events::Event] event details including payload
+      def on_params_params_deserialize(event)
+        # Keep in mind, that a caller here is a param object not a controller,
+        # so it returns a topic as a string, not a routing topic
+        debug(
+          <<~MSG.chomp.tr("\n", ' ')
+            Params deserialization for #{event[:caller].topic} topic
+            successful in #{event[:time]} ms
+          MSG
+        )
+      end
+
+      # Logs unsuccessful deserialization attempts of incoming data
+      # @param event [Dry::Events::Event] event details including payload
+      def on_params_params_deserialize_error(event)
+        error "Params deserialization error for #{event[:caller].topic} topic: #{event[:error]}"
+      end
+
+      # Logs errors that occurred in a listener fetch loop
+      # @param event [Dry::Events::Event] event details including payload
+      # @note It's an error as we can recover from it not a fatal
+      def on_connection_listener_fetch_loop_error(event)
+        error "Listener fetch loop error: #{event[:error]}"
+      end
+
+      # Logs errors that are related to the connection itself
+      # @param event [Dry::Events::Event] event details including payload
+      # @note Karafka will attempt to reconnect, so an error not a fatal
+      def on_connection_client_fetch_loop_error(event)
+        error "Client fetch loop error: #{event[:error]}"
+      end
+
+      # Logs info about crashed fetcher
+      # @param event [Dry::Events::Event] event details including payload
+      # @note If this happens, Karafka will shutdown as it means a critical error
+      #   in one of the threads
+      def on_fetcher_call_error(event)
+        fatal "Fetcher crash due to an error: #{event[:error]}"
+      end
+
+      # Logs info about processing of a certain dataset with an inline backend
+      # @param event [Dry::Events::Event] event details including payload
+      def on_backends_inline_process(event)
+        count = event[:caller].send(:params_batch).to_a.size
+        topic = event[:caller].topic.name
+        time = event[:time]
+        info "Inline processing of topic #{topic} with #{count} messages took #{time} ms"
+      end
+
+      # Logs info about system signals that Karafka received
+      # @param event [Dry::Events::Event] event details including payload
+      def on_process_notice_signal(event)
+        info "Received #{event[:signal]} system signal"
+      end
+
+      # Logs info about responder usage withing a controller flow
+      # @param event [Dry::Events::Event] event details including payload
+      def on_consumers_responders_respond_with(event)
+        calling = event[:caller]
+        responder = calling.topic.responder
+        data = event[:data]
+        info "Responded from #{calling.class} using #{responder} with following data #{data}"
+      end
+
+      # Logs info that we're initializing Karafka app
+      # @param _event [Dry::Events::Event] event details including payload
+      def on_app_initializing(_event)
+        info "Initializing Karafka server #{::Process.pid}"
+      end
+
+      # Logs info that we're running Karafka app
+      # @param _event [Dry::Events::Event] event details including payload
+      def on_app_running(_event)
+        info "Running Karafka server #{::Process.pid}"
+      end
+
+      # Logs info that we're going to stop the Karafka server
+      # @param _event [Dry::Events::Event] event details including payload
+      def on_app_stopping(_event)
+        # We use a separate thread as logging can't be called from trap context
+        Thread.new { info "Stopping Karafka server #{::Process.pid}" }
+      end
+
+      # Logs an error that Karafka was unable to stop the server gracefully and it had to do a
+      #   forced exit
+      # @param _event [Dry::Events::Event] event details including payload
+      def on_app_stopping_error(_event)
+        # We use a separate thread as logging can't be called from trap context
+        Thread.new { error "Forceful Karafka server #{::Process.pid} stop" }
+      end
+
+      USED_LOG_LEVELS.each do |log_level|
+        define_method log_level do |*args|
+          Karafka.logger.send(log_level, *args)
+        end
+      end
+    end
+  end
+end

data/lib/karafka/params/builders/metadata.rb

@@ -0,0 +1,33 @@
+# frozen_string_literal: true
+
+module Karafka
+  module Params
+    module Builders
+      # Builder for creating metadata object based on the message or batch informations
+      # @note We have 2 ways of creating metadata based on the way ruby-kafka operates
+      module Metadata
+        class << self
+          # Creates metadata based on the kafka batch data
+          # @param kafka_batch [Kafka::FetchedBatch] kafka batch details
+          # @param topic [Karafka::Routing::Topic] topic for which we've fetched the batch
+          # @return [Karafka::Params::Metadata] metadata object
+          def from_kafka_batch(kafka_batch, topic)
+            Karafka::Params::Metadata
+              .new
+              .merge!(
+                'batch_size' => kafka_batch.messages.count,
+                'first_offset' => kafka_batch.first_offset,
+                'highwater_mark_offset' => kafka_batch.highwater_mark_offset,
+                'last_offset' => kafka_batch.last_offset,
+                'offset_lag' => kafka_batch.offset_lag,
+                'deserializer' => topic.deserializer,
+                'partition' => kafka_batch.partition,
+                'topic' => kafka_batch.topic,
+                'unknown_last_offset' => kafka_batch.unknown_last_offset?
+              )
+          end
+        end
+      end
+    end
+  end
+end

data/lib/karafka/params/builders/params.rb

@@ -0,0 +1,36 @@
+# frozen_string_literal: true
+
+module Karafka
+  module Params
+    # Due to the fact, that we create params related objects in couple contexts / places
+    # plus backends can build up them their own way we have this namespace.
+    # It allows to isolate actual params objects from their building process that can be
+    # context dependent.
+    module Builders
+      # Builder for params
+      module Params
+        class << self
+          # @param kafka_message [Kafka::FetchedMessage] message fetched from Kafka
+          # @param topic [Karafka::Routing::Topic] topic for which this message was fetched
+          # @return [Karafka::Params::Params] params object
+          def from_kafka_message(kafka_message, topic)
+            Karafka::Params::Params
+              .new
+              .merge!(
+                'create_time' => kafka_message.create_time,
+                'headers' => kafka_message.headers || {},
+                'is_control_record' => kafka_message.is_control_record,
+                'key' => kafka_message.key,
+                'offset' => kafka_message.offset,
+                'deserializer' => topic.deserializer,
+                'partition' => kafka_message.partition,
+                'receive_time' => Time.now,
+                'topic' => kafka_message.topic,
+                'payload' => kafka_message.value
+              )
+          end
+        end
+      end
+    end
+  end
+end

data/lib/karafka/params/builders/params_batch.rb

@@ -0,0 +1,25 @@
+# frozen_string_literal: true
+
+module Karafka
+  module Params
+    module Builders
+      # Builder for creating params batch instances
+      module ParamsBatch
+        class << self
+          # Creates params batch with params inside based on the incoming messages
+          # and the topic from which it comes
+          # @param kafka_messages [Array<Kafka::FetchedMessage>] raw fetched messages
+          # @param topic [Karafka::Routing::Topic] topic for which we're received messages
+          # @return [Karafka::Params::ParamsBatch<Karafka::Params::Params>] batch with params
+          def from_kafka_messages(kafka_messages, topic)
+            params_array = kafka_messages.map! do |message|
+              Karafka::Params::Builders::Params.from_kafka_message(message, topic)
+            end
+
+            Karafka::Params::ParamsBatch.new(params_array)
+          end
+        end
+      end
+    end
+  end
+end

data/lib/karafka/params/metadata.rb

@@ -0,0 +1,35 @@
+# frozen_string_literal: true
+
+module Karafka
+  module Params
+    # Simple metadata object that stores all non-message information received from Kafka cluster
+    # while fetching the data
+    class Metadata < Hash
+      # Attributes that should be accessible as methods as well (not only hash)
+      METHOD_ATTRIBUTES = %w[
+        batch_size
+        first_offset
+        highwater_mark_offset
+        last_offset
+        offset_lag
+        deserializer
+        partition
+        topic
+      ].freeze
+
+      private_constant :METHOD_ATTRIBUTES
+
+      METHOD_ATTRIBUTES.each do |attr|
+        # Defines a method call accessor to a particular hash field.
+        define_method(attr) do
+          self[attr]
+        end
+      end
+
+      # @return [Boolean] is the last offset known or unknown
+      def unknown_last_offset?
+        self['unknown_last_offset']
+      end
+    end
+  end
+end

data/lib/karafka/params/params.rb

@@ -3,125 +3,65 @@
 module Karafka
   # Params namespace encapsulating all the logic that is directly related to params handling
   module Params
-    # Class-wrapper for hash with indifferent access with additional lazy loading feature
     # It provides lazy loading not only until the first usage, but also allows us to skip
-    # using parser until we execute our logic. That way we can operate with
-    # heavy-parsing data without slowing down the whole application.
-    class Params < HashWithIndifferentAccess
-      # Kafka::FetchedMessage attributes that we want to use inside of params
-      KAFKA_MESSAGE_ATTRIBUTES = %i[
-        value
-        partition
-        offset
-        key
-        create_time
-      ].freeze
-
+    # using deserializer until we execute our logic. That way we can operate with
+    # heavy-deserialization data without slowing down the whole application.
+    class Params < Hash
       # Params attributes that should be available via a method call invocation for Kafka
       # client compatibility.
       # Kafka passes internally Kafka::FetchedMessage object and the ruby-kafka consumer
       # uses those fields via method calls, so in order to be able to pass there our params
       # objects, have to have same api.
-      PARAMS_METHOD_ATTRIBUTES = %i[
-        topic
-        partition
-        offset
-        key
+      METHOD_ATTRIBUTES = %w[
         create_time
+        headers
+        is_control_record
+        key
+        offset
+        deserializer
+        deserialized
+        partition
+        receive_time
+        topic
+        payload
       ].freeze
 
-      class << self
-        # We allow building instances only via the #build method
-
-        # @param message [Kafka::FetchedMessage, Hash] message that we get out of Kafka
-        #   in case of building params inside main Karafka process in
-        #   Karafka::Connection::Consumer, or a hash when we retrieve data that is already parsed
-        # @param parser [Class] parser class that we will use to unparse data
-        # @return [Karafka::Params::Params] Karafka params object not yet used parser for
-        #   retrieving data that we've got from Kafka
-        # @example Build params instance from a hash
-        #   Karafka::Params::Params.build({ key: 'value' }) #=> params object
-        # @example Build params instance from a Kafka::FetchedMessage object
-        #   Karafka::Params::Params.build(message) #=> params object
-        def build(message, parser)
-          # Hash case happens inside backends that interchange data
-          if message.is_a?(Hash)
-            new(parser: parser).send(:merge!, message)
-          else
-            # This happens inside Kafka::FetchedProcessor
-            new(
-              parser: parser,
-              parsed: false,
-              received_at: Time.now
-            ).tap do |instance|
-              KAFKA_MESSAGE_ATTRIBUTES.each do |attribute|
-                instance[attribute] = message.send(attribute)
-              end
-
-              # When we get raw messages, they might have a topic, that was modified by a
-              # topic mapper. We need to "reverse" this change and map back to the non-modified
-              # format, so our internal flow is not corrupted with the mapping
-              instance[:topic] = Karafka::App.config.topic_mapper.incoming(message.topic)
-            end
-          end
-        end
+      private_constant :METHOD_ATTRIBUTES
 
+      METHOD_ATTRIBUTES.each do |attr|
         # Defines a method call accessor to a particular hash field.
         # @note Won't work for complex key names that contain spaces, etc
         # @param key [Symbol] name of a field that we want to retrieve with a method call
         # @example
         #   key_attr_reader :example
-        #   params.example #=> 'my example value'
-        def key_attr_reader(key)
-          define_method key do
-            self[key]
-          end
+        #   params.example #=> 'my example payload'
+        define_method(attr) do
+          self[attr]
         end
       end
 
-      # @return [Karafka::Params::Params] this will trigger parser execution. If we decide to
-      #   retrieve data, parser will be executed to parse data. Output of parsing will be merged
-      #   to the current object. This object will be also marked as already parsed, so we won't
-      #   parse it again.
-      def retrieve!
-        return self if self[:parsed]
+      # @return [Karafka::Params::Params] This method will trigger deserializer execution. If we
+      #   decide to retrieve data, deserializer will be executed to get data. Output of that will
+      #   be merged to the current object. This object will be also marked as already deserialized,
+      #   so we won't deserialize it again.
+      def deserialize!
+        return self if self['deserialized']
 
-        merge!(parse(delete(:value)))
+        self['deserialized'] = true
+        self['payload'] = deserialize
+        self
       end
 
-      PARAMS_METHOD_ATTRIBUTES.each(&method(:key_attr_reader))
-
       private
 
-      # Overwritten merge! method - it behaves differently for keys that are the same in our hash
-      #  and in a other_hash - it will not replace keys that are the same in our hash
-      #  and in the other one
-      # @param other_hash [Hash, HashWithIndifferentAccess] hash that we want to merge into current
-      # @return [Karafka::Params::Params] our parameters hash with merged values
-      # @example Merge with hash without same keys
-      #   new(a: 1, b: 2).merge!(c: 3) #=> { a: 1, b: 2, c: 3 }
-      # @example Merge with hash with same keys (symbol based)
-      #   new(a: 1).merge!(a: 2) #=> { a: 1 }
-      # @example Merge with hash with same keys (string based)
-      #   new(a: 1).merge!('a' => 2) #=> { a: 1 }
-      # @example Merge with hash with same keys (current string based)
-      #   new('a' => 1).merge!(a: 2) #=> { a: 1 }
-      def merge!(other_hash)
-        super(other_hash) { |_key, base_value, _new_value| base_value }
-      end
-
-      # @param value [String] Raw data that we want to parse using controller's parser
-      # @note If something goes wrong, it will return raw data in a hash with a message key
-      # @return [Hash] parsed data or a hash with message key containing raw data if something
-      #   went wrong during parsing
-      def parse(value)
-        self[:parser].parse(value)
-        # We catch both of them, because for default JSON - we use JSON parser directly
-      rescue ::Karafka::Errors::ParserError => e
-        Karafka.monitor.notice_error(self.class, e)
+      # @return [Object] deserialized data
+      def deserialize
+        Karafka.monitor.instrument('params.params.deserialize', caller: self) do
+          self['deserializer'].call(self)
+        end
+      rescue ::StandardError => e
+        Karafka.monitor.instrument('params.params.deserialize.error', caller: self, error: e)
         raise e
-      ensure
-        self[:parsed] = true
       end
     end
   end

data/lib/karafka/params/params_batch.rb

@@ -3,38 +3,58 @@
 module Karafka
   module Params
     # Params batch represents a set of messages received from Kafka.
-    # @note Params internally are lazy loaded before first use. That way we can skip parsing
-    #   process if we have after_fetched that rejects some incoming messages without using params
-    #   It can be also used when handling really heavy data (in terms of parsing).
+    # @note Params internally are lazy loaded before first use. That way we can skip
+    #   deserialization process if we have after_fetch that rejects some incoming messages
+    #   without using params It can be also used when handling really heavy data.
     class ParamsBatch
       include Enumerable
 
-      # Builds up a params batch based on raw kafka messages
-      # @param messages_batch [Array<Kafka::FetchedMessage>] messages batch
-      # @param topic_parser [Class] topic parser for unparsing messages values
-      def initialize(messages_batch, topic_parser)
-        @params_batch = messages_batch.map do |message|
-          Karafka::Params::Params.build(message, topic_parser)
-        end
+      # @param params_array [Array<Karafka::Params::Params>] array with karafka params
+      # @return [Karafka::Params::ParamsBatch] lazy evaluated params batch object
+      def initialize(params_array)
+        @params_array = params_array
       end
 
-      # @yieldparam [Karafka::Params::Params] each parsed and loaded params instance
-      # @note Invocation of this method will cause loading and parsing each param after another.
-      #   If you want to get access without parsing, please access params_batch directly
+      # @yieldparam [Karafka::Params::Params] each deserialized and loaded params instance
+      # @note Invocation of this method will cause loading and deserializing each param after
+      #   another. If you want to get access without deserializing, please access params_array
+      #   directly
       def each
-        @params_batch.each { |param| yield(param.retrieve!) }
+        @params_array.each { |param| yield(param.deserialize!) }
       end
 
       # @return [Array<Karafka::Params::Params>] returns all the params in a loaded state, so they
       #   can be used for batch insert, etc. Without invoking all, up until first use, they won't
-      #   be parsed
-      def parsed
+      #   be deserialized
+      def deserialize!
         each(&:itself)
       end
 
-      # @return [Array<Karafka::Params::Params>] pure array with params (not parsed)
+      # @return [Array<Object>] array with deserialized payloads. This method can be useful when
+      #   we don't care about metadata and just want to extract all the data payloads from the
+      #   batch
+      def payloads
+        deserialize!.map(&:payload)
+      end
+
+      # @return [Karafka::Params::Params] first element after the deserialization process
+      def first
+        @params_array.first.deserialize!
+      end
+
+      # @return [Karafka::Params::Params] last element after the deserialization process
+      def last
+        @params_array.last.deserialize!
+      end
+
+      # @return [Array<Karafka::Params::Params>] pure array with params (not deserialized)
       def to_a
-        @params_batch
+        @params_array
+      end
+
+      # @return [Integer] number of messages in the batch
+      def size
+        @params_array.size
       end
     end
   end