karafka 1.2.2 → 1.4.0.rc1

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).freeze
+          end
+        end
+      end
+    end
+  end
+end

data/lib/karafka/params/metadata.rb

@@ -0,0 +1,20 @@
+# frozen_string_literal: true
+
+module Karafka
+  module Params
+    # Single message / params metadata details that can be accessed without the need for the
+    # payload deserialization
+    Metadata = Struct.new(
+      :create_time,
+      :headers,
+      :is_control_record,
+      :key,
+      :offset,
+      :deserializer,
+      :partition,
+      :receive_time,
+      :topic,
+      keyword_init: true
+    )
+  end
+end

data/lib/karafka/params/params.rb

@@ -0,0 +1,50 @@
+# frozen_string_literal: true
+
+module Karafka
+  # Params namespace encapsulating all the logic that is directly related to params handling
+  module Params
+    # It provides lazy loading not only until the first usage, but also allows us to skip
+    # using deserializer until we execute our logic. That way we can operate with
+    # heavy-deserialization data without slowing down the whole application.
+    class Params
+      attr_reader :raw_payload, :metadata
+
+      # @param raw_payload [Object] incoming payload before deserialization
+      # @param metadata [Karafka::Params::Metadata] message metadata object
+      def initialize(raw_payload, metadata)
+        @raw_payload = raw_payload
+        @metadata = metadata
+        @deserialized = false
+        @payload = nil
+      end
+
+      # @return [Object] lazy-deserialized data (deserialized upon first request)
+      def payload
+        return @payload if deserialized?
+
+        @payload = deserialize
+        # We mark deserialization as successful after deserialization, as in case of an error
+        # this won't be falsely set to true
+        @deserialized = true
+        @payload
+      end
+
+      # @return [Boolean] did given params payload were deserialized already
+      def deserialized?
+        @deserialized
+      end
+
+      private
+
+      # @return [Object] tries de-serializes data
+      def deserialize
+        Karafka.monitor.instrument('params.params.deserialize', caller: self) do
+          metadata.deserializer.call(self)
+        end
+      rescue ::StandardError => e
+        Karafka.monitor.instrument('params.params.deserialize.error', caller: self, error: e)
+        raise e
+      end
+    end
+  end
+end

data/lib/karafka/params/params_batch.rb

@@ -3,43 +3,57 @@
 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_fetch 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 params instance
+      # @note Invocation of this method will not cause loading and deserializing each param after
+      #   another.
       def each
-        @params_batch.each { |param| yield(param.retrieve!) }
+        @params_array.each { |param| yield(param) }
       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
-        each(&:itself)
+      #   be deserialized
+      def deserialize!
+        each(&:payload)
       end
 
-      # @return [Karafka::Params::Params] last element after the unparsing process
+      # @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
+        map(&:payload)
+      end
+
+      # @return [Karafka::Params::Params] first element
+      def first
+        @params_array.first
+      end
+
+      # @return [Karafka::Params::Params] last element
       def last
-        @params_batch.last.retrieve!
+        @params_array.last
+      end
+
+      # @return [Integer] number of messages in the batch
+      def size
+        @params_array.size
       end
 
-      # @return [Array<Karafka::Params::Params>] pure array with params (not parsed)
+      # @return [Array<Karafka::Params::Params>] pure array with params
       def to_a
-        @params_batch
+        @params_array
       end
     end
   end

data/lib/karafka/patches/ruby_kafka.rb

@@ -1,34 +1,47 @@
 # frozen_string_literal: true
 
 module Karafka
+  # Namespace for various other libs patches
   module Patches
     # Patches for Ruby Kafka gem
     module RubyKafka
       # This patch allows us to inject business logic in between fetches and before the consumer
       # stop, so we can perform stop commit or anything else that we need since
       # ruby-kafka fetch loop does not allow that directly
-      # We don't wan't to use poll ruby-kafka api as it brings many more problems that we would
+      # We don't won't to use poll ruby-kafka api as it brings many more problems that we would
       # have to take care of. That way, nothing like that ever happens but we get the control
       # over the stopping process that we need (since we're the once that initiate it for each
       # thread)
       def consumer_loop
         super do
-          consumers = Karafka::Persistence::Consumer
-                      .all
+          consumers = Karafka::Persistence::Consumers
+                      .current
                       .values
                       .flat_map(&:values)
-                      .select { |ctrl| ctrl.respond_to?(:run_callbacks) }
+                      .select { |consumer| consumer.class.respond_to?(:after_fetch) }
 
-          if Karafka::App.stopped?
-            consumers.each { |ctrl| ctrl.run_callbacks :before_stop }
+          if Karafka::App.stopping?
+            publish_event(consumers, 'before_stop')
             Karafka::Persistence::Client.read.stop
           else
-            consumers.each { |ctrl| ctrl.run_callbacks :before_poll }
+            publish_event(consumers, 'before_poll')
             yield
-            consumers.each { |ctrl| ctrl.run_callbacks :after_poll }
+            publish_event(consumers, 'after_poll')
           end
         end
       end
+
+      private
+
+      # Notifies consumers about particular events happening
+      # @param consumers [Array<Object>] all consumers that want to be notified about an event
+      # @param event_name [String] name of the event that happened
+      def publish_event(consumers, event_name)
+        consumers.each do |consumer|
+          key = "consumers.#{Helpers::Inflector.map(consumer.class.to_s)}.#{event_name}"
+          Karafka::App.monitor.instrument(key, context: consumer)
+        end
+      end
     end
   end
 end

data/lib/karafka/persistence/client.rb

@@ -7,18 +7,22 @@ module Karafka
       # Thread.current key under which we store current thread messages consumer client
       PERSISTENCE_SCOPE = :client
 
-      # @param client [Karafka::Connection::Client] messages consumer client of
-      #   a current thread
-      # @return [Karafka::Connection::Client] persisted messages consumer client
-      def self.write(client)
-        Thread.current[PERSISTENCE_SCOPE] = client
-      end
+      private_constant :PERSISTENCE_SCOPE
+
+      class << self
+        # @param client [Karafka::Connection::Client] messages consumer client of
+        #   a current thread
+        # @return [Karafka::Connection::Client] persisted messages consumer client
+        def write(client)
+          Thread.current[PERSISTENCE_SCOPE] = client
+        end
 
-      # @return [Karafka::Connection::Client] persisted messages consumer client
-      # @raise [Karafka::Errors::MissingConsumer] raised when no thread messages consumer
-      #   client but we try to use it anyway
-      def self.read
-        Thread.current[PERSISTENCE_SCOPE] || raise(Errors::MissingClient)
+        # @return [Karafka::Connection::Client] persisted messages consumer client
+        # @raise [Karafka::Errors::MissingClientError] raised when no thread messages consumer
+        #   client but we try to use it anyway
+        def read
+          Thread.current[PERSISTENCE_SCOPE] || raise(Errors::MissingClientError)
+        end
       end
     end
   end

data/lib/karafka/persistence/{consumer.rb → consumers.rb}

@@ -7,30 +7,37 @@ module Karafka
     # Module used to provide a persistent cache across batch requests for a given
     # topic and partition to store some additional details when the persistent mode
     # for a given topic is turned on
-    class Consumer
+    class Consumers
       # Thread.current scope under which we store consumers data
       PERSISTENCE_SCOPE = :consumers
 
+      private_constant :PERSISTENCE_SCOPE
+
       class << self
-        # @return [Hash] current thread persistence scope hash with all the consumers
-        def all
-          # @note This does not need to be threadsafe (Hash) as it is always executed in a
-          # current thread context
-          Thread.current[PERSISTENCE_SCOPE] ||= Hash.new { |hash, key| hash[key] = {} }
+        # @return [Hash] current thread's persistence scope hash with all the consumers
+        def current
+          Thread.current[PERSISTENCE_SCOPE] ||= Concurrent::Hash.new do |hash, key|
+            hash[key] = Concurrent::Hash.new
+          end
         end
 
         # Used to build (if block given) and/or fetch a current consumer instance that will be
         #   used to process messages from a given topic and partition
-        # @return [Karafka::BaseConsumer] base consumer descendant
         # @param topic [Karafka::Routing::Topic] topic instance for which we might cache
         # @param partition [Integer] number of partition for which we want to cache
+        # @return [Karafka::BaseConsumer] base consumer descendant
         def fetch(topic, partition)
-          # We always store a current instance for callback reasons
-          if topic.persistent
-            all[topic][partition] ||= topic.consumer.new
-          else
-            all[topic][partition] = topic.consumer.new
-          end
+          current[topic][partition] ||= topic.consumer.new(topic)
+        end
+
+        # Removes all persisted instances of consumers from the consumer cache
+        # @note This is used to reload consumers instances when code reloading in development mode
+        #   is present. This should not be used in production.
+        def clear
+          Thread
+            .list
+            .select { |thread| thread[PERSISTENCE_SCOPE] }
+            .each { |thread| thread[PERSISTENCE_SCOPE].clear }
         end
       end
     end

data/lib/karafka/persistence/topics.rb

@@ -0,0 +1,48 @@
+# frozen_string_literal: true
+
+module Karafka
+  module Persistence
+    # Local cache for routing topics
+    # We use it in order not to build string instances and remap incoming topic upon each
+    # message / message batches received
+    class Topics
+      # Thread.current scope under which we store topics data
+      PERSISTENCE_SCOPE = :topics
+
+      private_constant :PERSISTENCE_SCOPE
+
+      class << self
+        # @return [Concurrent::Hash] hash with all the topics from given groups
+        def current
+          Thread.current[PERSISTENCE_SCOPE] ||= Concurrent::Hash.new do |hash, key|
+            hash[key] = Concurrent::Hash.new
+          end
+        end
+
+        # @param group_id [String] group id for which we fetch a topic representation
+        # @param raw_topic_name [String] raw topic name (before remapping) for which we fetch a
+        #   topic representation
+        # @return [Karafka::Routing::Topics] remapped topic representation that can be used further
+        #   on when working with given parameters
+        def fetch(group_id, raw_topic_name)
+          current[group_id][raw_topic_name] ||= begin
+            # We map from incoming topic name, as it might be namespaced, etc.
+            # @see topic_mapper internal docs
+            mapped_topic_name = Karafka::App.config.topic_mapper.incoming(raw_topic_name)
+            Routing::Router.find("#{group_id}_#{mapped_topic_name}")
+          end
+        end
+
+        # Clears the whole topics cache for all the threads
+        # This is used for in-development code reloading as we need to get rid of all the
+        # preloaded and cached instances of objects to make it work
+        def clear
+          Thread
+            .list
+            .select { |thread| thread[PERSISTENCE_SCOPE] }
+            .each { |thread| thread[PERSISTENCE_SCOPE].clear }
+        end
+      end
+    end
+  end
+end

data/lib/karafka/process.rb

@@ -4,8 +4,6 @@ module Karafka
   # Class used to catch signals from ruby Signal class in order to manage Karafka stop
   # @note There might be only one process - this class is a singleton
   class Process
-    include Singleton
-
     # Signal types that we handle
     HANDLED_SIGNALS = %i[
       SIGINT
@@ -34,10 +32,8 @@ module Karafka
 
     # Method catches all HANDLED_SIGNALS and performs appropriate callbacks (if defined)
     # @note If there are no callbacks, this method will just ignore a given signal that was sent
-    # @yield [Block] block of code that we want to execute and supervise
     def supervise
       HANDLED_SIGNALS.each { |signal| trap_signal(signal) }
-      yield
     end
 
     private

data/lib/karafka/responders/builder.rb

@@ -3,7 +3,7 @@
 module Karafka
   # Responders namespace encapsulates all the internal responder implementation parts
   module Responders
-    # Responders builder is used to finding (based on the consumer class name) a responder
+    # Responders builder is used for finding (based on the consumer class name) a responder
     # that match the consumer. We use it when user does not provide a responder inside routing,
     # but he still names responder with the same convention (and namespaces) as consumer
     #

data/lib/karafka/responders/topic.rb

@@ -7,8 +7,6 @@ module Karafka
     #   Karafka::Responders::Topic.new(:topic_name, {}) #=> #<Karafka::Responders::Topic...
     # @example Define optional topic
     #   Karafka::Responders::Topic.new(:topic_name, required: false)
-    # @example Define topic that on which we want to respond multiple times
-    #   Karafka::Responders::Topic.new(:topic_name, multiple_usage: true)
     class Topic
       # Name of the topic on which we want to respond
       attr_reader :name
@@ -26,16 +24,16 @@ module Karafka
         @options.key?(:required) ? @options[:required] : true
       end
 
-      # @return [Boolean] do we expect to use it multiple times in a single respond flow
-      def multiple_usage?
-        @options[:multiple_usage] || false
-      end
-
       # @return [Boolean] was usage of this topic registered or not
       def registered?
         @options[:registered] == true
       end
 
+      # @return [Class] Class to use to serialize messages for this topic
+      def serializer
+        @options[:serializer]
+      end
+
       # @return [Boolean] do we want to use async producer. Defaults to false as the sync producer
       #   is safer and introduces less problems
       def async?
@@ -46,9 +44,9 @@ module Karafka
       def to_h
         {
           name: name,
-          multiple_usage: multiple_usage?,
           required: required?,
           registered: registered?,
+          serializer: serializer,
           async: async?
         }
       end

data/lib/karafka/routing/builder.rb

@@ -9,26 +9,39 @@ module Karafka
     #       consumer NewVideosConsumer
     #     end
     #   end
-    class Builder < Array
-      include Singleton
+    class Builder < Concurrent::Array
+      # Consumer group consistency checking contract
+      CONTRACT = Karafka::Contracts::ConsumerGroup.new.freeze
+
+      private_constant :CONTRACT
+
+      def initialize
+        @draws = Concurrent::Array.new
+      end
 
       # Used to draw routes for Karafka
+      # @param block [Proc] block we will evaluate within the builder context
+      # @yield Evaluates provided block in a builder context so we can describe routes
+      # @raise [Karafka::Errors::InvalidConfigurationError] raised when configuration
+      #   doesn't match with the config contract
       # @note After it is done drawing it will store and validate all the routes to make sure that
       #   they are correct and that there are no topic/group duplications (this is forbidden)
-      # @yield Evaluates provided block in a builder context so we can describe routes
       # @example
       #   draw do
       #     topic :xyz do
       #     end
       #   end
       def draw(&block)
+        @draws << block
+
         instance_eval(&block)
 
         each do |consumer_group|
           hashed_group = consumer_group.to_h
-          validation_result = Karafka::Schemas::ConsumerGroup.call(hashed_group)
-          return if validation_result.success?
-          raise Errors::InvalidConfiguration, validation_result.errors
+          validation_result = CONTRACT.call(hashed_group)
+          next if validation_result.success?
+
+          raise Errors::InvalidConfigurationError, validation_result.errors.to_h
         end
       end
 
@@ -39,18 +52,33 @@ module Karafka
         select(&:active?)
       end
 
+      # Clears the builder and the draws memory
+      def clear
+        @draws.clear
+        super
+      end
+
+      # Redraws all the routes for the in-process code reloading.
+      # @note This won't allow registration of new topics without process restart but will trigger
+      #   cache invalidation so all the classes, etc are re-fetched after code reload
+      def reload
+        draws = @draws.dup
+        clear
+        draws.each { |block| draw(&block) }
+      end
+
       private
 
       # Builds and saves given consumer group
       # @param group_id [String, Symbol] name for consumer group
-      # @yield Evaluates a given block in a consumer group context
+      # @param block [Proc] proc that should be executed in the proxy context
       def consumer_group(group_id, &block)
         consumer_group = ConsumerGroup.new(group_id.to_s)
         self << Proxy.new(consumer_group, &block).target
       end
 
       # @param topic_name [String, Symbol] name of a topic from which we want to consumer
-      # @yield Evaluates a given block in a topic context
+      # @param block [Proc] proc we want to evaluate in the topic context
       def topic(topic_name, &block)
         consumer_group(topic_name) do
           topic(topic_name, &block).tap(&:build)