ddtrace 0.40.0 → 0.41.0

data/lib/ddtrace/runtime/metrics.rb

@@ -1,3 +1,4 @@
+require 'ddtrace/ext/integration'
 require 'ddtrace/ext/runtime'
 
 require 'ddtrace/metrics'
@@ -25,8 +26,11 @@ module Datadog
         # Register service as associated with metrics
         register_service(span.service) unless span.service.nil?
 
-        # Tag span with language and runtime ID for association with metrics
-        span.set_tag(Ext::Runtime::TAG_LANG, Runtime::Identity.lang)
+        # Tag span with language and runtime ID for association with metrics.
+        # We only tag spans that performed internal application work.
+        unless span.get_tag(Datadog::Ext::Integration::TAG_PEER_SERVICE)
+          span.set_tag(Ext::Runtime::TAG_LANG, Runtime::Identity.lang)
+        end
       end
 
       # Associate service with runtime metrics

data/lib/ddtrace/sampler.rb

@@ -61,11 +61,11 @@ module Datadog
 
     def sample_rate=(sample_rate)
       @sample_rate = sample_rate
-      @sampling_id_threshold = sample_rate * Span::MAX_ID
+      @sampling_id_threshold = sample_rate * Span::EXTERNAL_MAX_ID
     end
 
     def sample?(span)
-      ((span.trace_id * KNUTH_FACTOR) % Datadog::Span::MAX_ID) <= @sampling_id_threshold
+      ((span.trace_id * KNUTH_FACTOR) % Datadog::Span::EXTERNAL_MAX_ID) <= @sampling_id_threshold
     end
 
     def sample!(span)

data/lib/ddtrace/span.rb

@@ -1,3 +1,5 @@
+# frozen_string_literal: true
+
 require 'time'
 require 'thread'
 
@@ -8,6 +10,7 @@ require 'ddtrace/environment'
 require 'ddtrace/analytics'
 require 'ddtrace/forced_tracing'
 require 'ddtrace/diagnostics/health'
+require 'ddtrace/utils/time'
 
 module Datadog
   # Represents a logical unit of work in the system. Each trace consists of one or more spans.
@@ -24,24 +27,26 @@ module Datadog
     # The max value for a \Span identifier.
     # Span and trace identifiers should be strictly positive and strictly inferior to this limit.
     #
-    # Limited to 63-bit positive integers, as some other languages might be limited to this,
-    # and IDs need to be easy to port across various languages and platforms.
-    MAX_ID = 2**63
+    # Limited to +2<<62-1+ positive integers, as Ruby is able to represent such numbers "inline",
+    # inside a +VALUE+ scalar, thus not requiring memory allocation.
+    #
+    # The range of IDs also has to consider portability across different languages and platforms.
+    RUBY_MAX_ID = (1 << 62) - 1
 
     # While we only generate 63-bit integers due to limitations in other languages, we support
     # parsing 64-bit integers for distributed tracing since an upstream system may generate one
-    EXTERNAL_MAX_ID = 2**64
+    EXTERNAL_MAX_ID = 1 << 64
 
     # This limit is for numeric tags because uint64 could end up rounded.
-    NUMERIC_TAG_SIZE_RANGE = (-2**53..2**53)
+    NUMERIC_TAG_SIZE_RANGE = (-1 << 53..1 << 53)
 
     attr_accessor :name, :service, :resource, :span_type,
-                  :start_time, :end_time,
                   :span_id, :trace_id, :parent_id,
                   :status, :sampled,
-                  :tracer, :context
+                  :tracer, :context, :duration, :start_time, :end_time
 
     attr_reader :parent
+
     # Create a new span linked to the given tracer. Call the \Tracer method <tt>start_span()</tt>
     # and then <tt>finish()</tt> once the tracer operation is over.
     #
@@ -72,11 +77,20 @@ module Datadog
       @parent = nil
       @sampled = true
 
-      @start_time = nil # set by Tracer.start_span
-      @end_time = nil # set by Span.finish
-
       @allocation_count_start = now_allocations
       @allocation_count_finish = @allocation_count_start
+
+      # start_time and end_time track wall clock. In Ruby, wall clock
+      # has less accuracy than monotonic clock, so if possible we look to only use wall clock
+      # to measure duration when a time is supplied by the user, or if monotonic clock
+      # is unsupported.
+      @start_time = nil
+      @end_time = nil
+
+      # duration_start and duration_end track monotonic clock, and may remain nil in cases where it
+      # is known that we have to use wall clock to measure duration.
+      @duration_start = nil
+      @duration_end = nil
     end
 
     # Set the given key / value tag pair on the span. Keys and values
@@ -160,6 +174,28 @@ module Datadog
       set_tag(Ext::Errors::STACK, e.backtrace) unless e.backtrace.empty?
     end
 
+    # Mark the span started at the current time.
+    def start(start_time = nil)
+      # A span should not be started twice. However, this is existing
+      # behavior and so we maintain it for backward compatibility for those
+      # who are using async manual instrumentation that may rely on this
+
+      @start_time = start_time || Time.now.utc
+      @duration_start = start_time.nil? ? duration_marker : nil
+
+      self
+    end
+
+    # for backwards compatibility
+    def start_time=(time)
+      time.tap { start(time) }
+    end
+
+    # for backwards compatibility
+    def end_time=(time)
+      time.tap { finish(time) }
+    end
+
     # Mark the span finished at the current time and submit it.
     def finish(finish_time = nil)
       # A span should not be finished twice. Note that this is not thread-safe,
@@ -170,12 +206,15 @@ module Datadog
 
       @allocation_count_finish = now_allocations
 
-      # Provide a default start_time if unset, but this should have been set by start_span.
-      # Using now here causes 0-duration spans, still, this is expected, as we never
-      # explicitely say when it started.
-      @start_time ||= Time.now.utc
+      now = Time.now.utc
+
+      # Provide a default start_time if unset.
+      # Using `now` here causes duration to be 0; this is expected
+      # behavior when start_time is unknown.
+      start(finish_time || now) unless started?
 
-      @end_time = finish_time.nil? ? Time.now.utc : finish_time # finish this
+      @end_time = finish_time || now
+      @duration_end = finish_time.nil? ? duration_marker : nil
 
       # Finish does not really do anything if the span is not bound to a tracer and a context.
       return self if @tracer.nil? || @context.nil?
@@ -195,11 +234,6 @@ module Datadog
       self
     end
 
-    # Return whether the span is finished or not.
-    def finished?
-      !@end_time.nil?
-    end
-
     # Return a string representation of the span.
     def to_s
       "Span(name:#{@name},sid:#{@span_id},tid:#{@trace_id},pid:#{@parent_id})"
@@ -246,19 +280,76 @@ module Datadog
         error: @status
       }
 
-      if !@start_time.nil? && !@end_time.nil?
-        h[:start] = (@start_time.to_f * 1e9).to_i
-        h[:duration] = ((@end_time - @start_time) * 1e9).to_i
+      if finished?
+        h[:start] = start_time_nano
+        h[:duration] = duration_nano
       end
 
       h
     end
 
+    # MessagePack serializer interface. Making this object
+    # respond to `#to_msgpack` allows it to be automatically
+    # serialized by MessagePack.
+    #
+    # This is more efficient than doing +MessagePack.pack(span.to_hash)+
+    # as we don't have to create an intermediate Hash.
+    #
+    # @param packer [MessagePack::Packer] serialization buffer, can be +nil+ with JRuby
+    def to_msgpack(packer = nil)
+      # As of 1.3.3, JRuby implementation doesn't pass an existing packer
+      packer ||= MessagePack::Packer.new
+
+      if finished?
+        packer.write_map_header(13) # Set header with how many elements in the map
+
+        packer.write('start')
+        packer.write(start_time_nano)
+
+        packer.write('duration')
+        packer.write(duration_nano)
+      else
+        packer.write_map_header(11) # Set header with how many elements in the map
+      end
+
+      # DEV: We use strings as keys here, instead of symbols, as
+      # DEV: MessagePack will ultimately convert them to strings.
+      # DEV: By providing strings directly, we skip this indirection operation.
+      packer.write('span_id')
+      packer.write(@span_id)
+      packer.write('parent_id')
+      packer.write(@parent_id)
+      packer.write('trace_id')
+      packer.write(@trace_id)
+      packer.write('name')
+      packer.write(@name)
+      packer.write('service')
+      packer.write(@service)
+      packer.write('resource')
+      packer.write(@resource)
+      packer.write('type')
+      packer.write(@span_type)
+      packer.write('meta')
+      packer.write(@meta)
+      packer.write('metrics')
+      packer.write(@metrics)
+      packer.write('allocations')
+      packer.write(allocations)
+      packer.write('error')
+      packer.write(@status)
+      packer
+    end
+
+    # JSON serializer interface.
+    # Used by older version of the transport.
+    def to_json(*args)
+      to_hash.to_json(*args)
+    end
+
     # Return a human readable version of the span
     def pretty_print(q)
-      start_time = (@start_time.to_f * 1e9).to_i rescue '-'
-      end_time = (@end_time.to_f * 1e9).to_i rescue '-'
-      duration = ((@end_time - @start_time) * 1e9).to_i rescue 0
+      start_time = (self.start_time.to_f * 1e9).to_i
+      end_time = (self.end_time.to_f * 1e9).to_i
       q.group 0 do
         q.breakable
         q.text "Name: #{@name}\n"
@@ -271,7 +362,7 @@ module Datadog
         q.text "Error: #{@status}\n"
         q.text "Start: #{start_time}\n"
         q.text "End: #{end_time}\n"
-        q.text "Duration: #{duration}\n"
+        q.text "Duration: #{duration.to_f if finished?}\n"
         q.text "Allocations: #{allocations}\n"
         q.group(2, 'Tags: [', "]\n") do
           q.breakable
@@ -288,8 +379,30 @@ module Datadog
       end
     end
 
+    # Return whether the duration is started or not
+    def started?
+      !@start_time.nil?
+    end
+
+    # Return whether the duration is finished or not.
+    def finished?
+      !@end_time.nil?
+    end
+
+    def duration
+      if @duration_end.nil? || @duration_start.nil?
+        @end_time - @start_time
+      else
+        @duration_end - @duration_start
+      end
+    end
+
     private
 
+    def duration_marker
+      Utils::Time.get_time
+    end
+
     if defined?(JRUBY_VERSION) || Gem::Version.new(RUBY_VERSION) < Gem::Version.new(VERSION::MINIMUM_RUBY_VERSION)
       def now_allocations
         0
@@ -303,5 +416,17 @@ module Datadog
         GC.stat(:total_allocated_objects)
       end
     end
+
+    # Used for serialization
+    # @return [Integer] in nanoseconds since Epoch
+    def start_time_nano
+      @start_time.to_i * 1000000000 + @start_time.nsec
+    end
+
+    # Used for serialization
+    # @return [Integer] in nanoseconds since Epoch
+    def duration_nano
+      (duration * 1e9).to_i
+    end
   end
 end

data/lib/ddtrace/tracer.rb

@@ -1,4 +1,3 @@
-require 'pp'
 require 'thread'
 require 'logger'
 require 'pathname'
@@ -186,8 +185,7 @@ module Datadog
     # * +start_time+: when the span actually starts (defaults to \now)
     # * +tags+: extra tags which should be added to the span.
     def start_span(name, options = {})
-      start_time = options.fetch(:start_time, Time.now.utc)
-
+      start_time = options[:start_time]
       tags = options.fetch(:tags, {})
 
       span_options = options.select do |k, _v|
@@ -212,9 +210,10 @@ module Datadog
         # child span
         span.parent = parent # sets service, trace_id, parent_id, sampled
       end
+
       span.set_tags(@tags) unless @tags.empty?
       span.set_tags(tags) unless tags.empty?
-      span.start_time = start_time
+      span.start(start_time)
 
       # this could at some point be optional (start_active_span vs start_manual_span)
       ctx.add_span(span) unless ctx.nil?

data/lib/ddtrace/transport/http/adapters/net.rb

@@ -20,8 +20,14 @@ module Datadog
           end
 
           def open
-            # Open connection
-            ::Net::HTTP.start(hostname, port, open_timeout: timeout, read_timeout: timeout) do |http|
+            # DEV Initializing +Net::HTTP+ directly help us avoid expensive
+            # options processing done in +Net::HTTP.start+:
+            # https://github.com/ruby/ruby/blob/b2d96abb42abbe2e01f010ffc9ac51f0f9a50002/lib/net/http.rb#L614-L618
+            req = ::Net::HTTP.new(hostname, port, nil)
+
+            req.open_timeout = req.read_timeout = timeout
+
+            req.start do |http|
               yield(http)
             end
           end

data/lib/ddtrace/transport/http/statistics.rb

@@ -18,11 +18,24 @@ module Datadog
               # Add status code tag to api.responses metric
               if metrics.key?(:api_responses)
                 (metrics[:api_responses].options[:tags] ||= []).tap do |tags|
-                  tags << "status_code:#{response.code}"
+                  tags << metrics_tag_value(response.code)
                 end
               end
             end
           end
+
+          private
+
+          # The most common status code on a healthy tracer
+          STATUS_CODE_200 = 'status_code:200'.freeze
+
+          def metrics_tag_value(status_code)
+            if status_code == 200
+              STATUS_CODE_200 # DEV Saves string concatenation/creation for common case
+            else
+              "status_code:#{status_code}"
+            end
+          end
         end
       end
     end

data/lib/ddtrace/transport/traces.rb

@@ -57,7 +57,12 @@ module Datadog
         # @return [Enumerable[Array[Bytes,Integer]]] list of encoded chunks: each containing a byte array and
         #   number of traces
         def encode_in_chunks(traces)
-          encoded_traces = traces.map { |t| encode_one(t) }.reject(&:nil?)
+          encoded_traces = if traces.respond_to?(:filter_map)
+                             # DEV Supported since Ruby 2.7, saves an intermediate object creation
+                             traces.filter_map { |t| encode_one(t) }
+                           else
+                             traces.map { |t| encode_one(t) }.reject(&:nil?)
+                           end
 
           Datadog::Chunker.chunk_by_size(encoded_traces, max_size).map do |chunk|
             [encoder.join(chunk), chunk.size]
@@ -86,7 +91,7 @@ module Datadog
         module_function
 
         def encode_trace(encoder, trace)
-          encoder.encode(trace.map(&:to_hash))
+          encoder.encode(trace)
         end
       end
 

data/lib/ddtrace/utils.rb

@@ -3,7 +3,7 @@ require 'ddtrace/utils/database'
 module Datadog
   # Utils contains low-level utilities, typically to provide pseudo-random trace IDs.
   module Utils
-    STRING_PLACEHOLDER = ''.encode(::Encoding::UTF_8).freeze
+    EMPTY_STRING = ''.encode(::Encoding::UTF_8).freeze
     # We use a custom random number generator because we want no interference
     # with the default one. Using the default prng, we could break code that
     # would rely on srand/rand sequences.
@@ -12,7 +12,7 @@ module Datadog
     def self.next_id
       reset! if was_forked?
 
-      @rnd.rand(Datadog::Span::MAX_ID)
+      @rnd.rand(Datadog::Span::RUBY_MAX_ID)
     end
 
     def self.reset!
@@ -53,13 +53,17 @@ module Datadog
         str.encode('UTF-8', 'binary', invalid: :replace, undef: :replace, replace: '')
       elsif str.encoding == ::Encoding::UTF_8
         str
+      elsif str.empty?
+        # DEV Optimization as `nil.to_s` is a very common source for an empty string,
+        # DEV but it comes encoded as US_ASCII.
+        EMPTY_STRING
       else
         str.encode(::Encoding::UTF_8)
       end
     rescue => e
       Datadog.logger.debug("Error encoding string in UTF-8: #{e}")
 
-      options.fetch(:placeholder, STRING_PLACEHOLDER)
+      options.fetch(:placeholder, EMPTY_STRING)
     end
   end
 end

data/lib/ddtrace/version.rb

@@ -1,7 +1,7 @@
 module Datadog
   module VERSION
     MAJOR = 0
-    MINOR = 40
+    MINOR = 41
     PATCH = 0
     PRE = nil
 

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: ddtrace
 version: !ruby/object:Gem::Version
-  version: 0.40.0
+  version: 0.41.0
 platform: ruby
 authors:
 - Datadog, Inc.
 autorequire: 
 bindir: exe
 cert_chain: []
-date: 2020-09-08 00:00:00.000000000 Z
+date: 2020-09-30 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: msgpack
@@ -52,6 +52,20 @@ dependencies:
     - - ">="
       - !ruby/object:Gem::Version
         version: 0.4.1
+- !ruby/object:Gem::Dependency
+  name: concurrent-ruby
+  requirement: !ruby/object:Gem::Requirement
+    requirements:
+    - - ">="
+      - !ruby/object:Gem::Version
+        version: '0'
+  type: :development
+  prerelease: false
+  version_requirements: !ruby/object:Gem::Requirement
+    requirements:
+    - - ">="
+      - !ruby/object:Gem::Version
+        version: '0'
 - !ruby/object:Gem::Dependency
   name: rake
   requirement: !ruby/object:Gem::Requirement
@@ -716,6 +730,7 @@ files:
 - lib/ddtrace/ext/errors.rb
 - lib/ddtrace/ext/forced_tracing.rb
 - lib/ddtrace/ext/http.rb
+- lib/ddtrace/ext/integration.rb
 - lib/ddtrace/ext/manual_tracing.rb
 - lib/ddtrace/ext/metrics.rb
 - lib/ddtrace/ext/net.rb