vernier 1.10.1 → 1.12.0

data/lib/vernier/output/cpuprofile.rb

@@ -1,8 +1,12 @@
 require "json"
 
+require_relative "../output_helpers"
+
 module Vernier
   module Output
     class Cpuprofile
+      include OutputHelpers
+
       def initialize(profile)
         @profile = profile
       end
@@ -98,7 +102,9 @@ module Vernier
         line = stack_table.frame_line_no(frame_idx) - 1
 
         func_name = stack_table.func_name(func_idx)
+        func_name = sanitize_string(func_name) if func_name
         filename = stack_table.func_filename(func_idx)
+        filename = sanitize_string(filename) if filename
 
         node_id = nodes.length
         node = {

data/lib/vernier/output/file_listing.rb

@@ -1,23 +1,19 @@
 # frozen_string_literal: true
 
 require_relative "filename_filter"
+require_relative "../output_helpers"
 require "cgi/escape"
 
 module Vernier
   module Output
     class FileListing
+      include OutputHelpers
+
       class SamplesByLocation
         attr_accessor :self, :total
         def initialize
           @self = @total = 0
         end
-
-        def +(other)
-          ret = SamplesByLocation.new
-          ret.self = @self + other.self
-          ret.total = @total + other.total
-          ret
-        end
       end
 
       def initialize(profile)
@@ -25,6 +21,8 @@ module Vernier
       end
 
       def samples_by_file
+        return @samples_by_file if defined?(@samples_by_file)
+
         thread = @profile.main_thread
         if Hash === thread
           # live profile
@@ -38,22 +36,9 @@ module Vernier
         weights = thread[:weights]
         samples = thread[:samples]
 
-        self_samples_by_frame = Hash.new do |h, k|
-          h[k] = SamplesByLocation.new
-        end
-
-        samples.zip(weights).each do |stack_idx, weight|
-          # self time
-          top_frame_index = stack_table.stack_frame_idx(stack_idx)
-          self_samples_by_frame[top_frame_index].self += weight
+        stack_weights = collapse_stack_weights(samples, weights)
 
-          # total time
-          while stack_idx
-            frame_idx = stack_table.stack_frame_idx(stack_idx)
-            self_samples_by_frame[frame_idx].total += weight
-            stack_idx = stack_table.stack_parent_idx(stack_idx)
-          end
-        end
+        self_by_frame, total_by_frame = frame_weights(stack_table, stack_weights)
 
         samples_by_file = Hash.new do |h, k|
           h[k] = Hash.new do |h2, k2|
@@ -61,19 +46,64 @@ module Vernier
           end
         end
 
-        self_samples_by_frame.each do |frame, samples|
+        self_by_frame.each_with_index do |self_weight, frame|
+          total_weight = total_by_frame[frame]
+          next if self_weight == 0 && total_weight == 0
+
           line = stack_table.frame_line_no(frame)
           func_index = stack_table.frame_func_idx(frame)
           filename = stack_table.func_filename(func_index)
 
-          samples_by_file[filename][line] += samples
+          location = samples_by_file[filename][line]
+          location.self += self_weight
+          location.total += total_weight
         end
 
-        samples_by_file.transform_keys! do |filename|
+        @samples_by_file = samples_by_file.transform_keys! do |filename|
           filename_filter.call(filename)
         end
       end
 
+      # Returns [self_weight, total_weight] arrays indexed by frame, from a
+      # hash of per-stack weights.
+      #
+      # The stack table is a prefix tree whose parent is always interned
+      # before its children (parent idx < child idx), so total weights are
+      # accumulated bottom-up in a single reverse pass.
+      private def frame_weights(stack_table, stack_weights)
+        # Scatter the sparse weights into an array indexed by stack.
+        stack_totals = Array.new(stack_table.stack_count, 0)
+        stack_weights.each do |stack_idx, weight|
+          stack_totals[stack_idx] = weight
+        end
+
+        # Accumulate each stack's total into its parent, children first,
+        # so a single descending pass completes every subtree's total.
+        (stack_totals.length - 1).downto(0) do |stack_idx|
+          parent_idx = stack_table.stack_parent_idx(stack_idx)
+          next unless parent_idx
+
+          if parent_idx >= stack_idx
+            raise "Invalid profile: stack table is not parent-before-child ordered"
+          end
+
+          stack_totals[parent_idx] += stack_totals[stack_idx]
+        end
+
+        self_by_frame = Array.new(stack_table.frame_count, 0)
+        total_by_frame = Array.new(stack_table.frame_count, 0)
+        stack_weights.each do |stack_idx, weight|
+          self_by_frame[stack_table.stack_frame_idx(stack_idx)] += weight
+        end
+        stack_totals.each_with_index do |total, stack_idx|
+          next if total == 0
+
+          total_by_frame[stack_table.stack_frame_idx(stack_idx)] += total
+        end
+
+        [self_by_frame, total_by_frame]
+      end
+
       def output(template: nil)
         output = +""
 
@@ -98,8 +128,7 @@ module Vernier
       end
 
       def total
-        thread = @profile.main_thread
-        thread[:weights].sum
+        @total ||= @profile.main_thread[:weights].sum
       end
 
       def format_file(output, filename, all_samples, total:)

data/lib/vernier/output/firefox.rb

@@ -4,6 +4,7 @@ require "json"
 require "rbconfig"
 
 require_relative "filename_filter"
+require_relative "../output_helpers"
 
 module Vernier
   module Output
@@ -120,7 +121,7 @@ module Vernier
       end
 
       def output(gzip: false)
-        result = ::JSON.generate(data)
+        result = ::JSON.generate(data(profile.meta[:hide_anonymous_threads]))
         if gzip
           require "zlib"
           result = Zlib.gzip(result)
@@ -132,7 +133,7 @@ module Vernier
 
       attr_reader :profile
 
-      def data
+      def data(hide_anonymous_threads)
         #markers_by_thread = profile.markers.group_by { |marker| marker[0] }
 
         threads = profile.threads.map do |ruby_thread_id, thread_info|
@@ -146,6 +147,12 @@ module Vernier
           )
         end
 
+        initially_visible_threads = if hide_anonymous_threads
+          threads.each_index.reject { |i| threads[i].anonymous? && !threads[i].is_main && !threads[i].is_start }.to_a
+        else
+          threads.each_index.to_a
+        end
+
         {
           meta: {
             interval: 1, # FIXME: memory vs wall
@@ -171,7 +178,7 @@ module Vernier
               }
             end,
             sourceCodeIsNotOnSearchfox: true,
-            initialVisibleThreads: threads.each_index.to_a,
+            initialVisibleThreads: initially_visible_threads,
             initialSelectedThreads: Array(threads.find_index(&:is_start)),
             vernierUserMetadata: profile.meta[:user_metadata],
             extra: [
@@ -267,20 +274,24 @@ module Vernier
       end
 
       class Thread
+        include OutputHelpers
+
         SAMPLE_CATEGORY_NAMES = {
           1 => "Idle",
           2 => "Stalled"
         }.freeze
 
-        attr_reader :profile, :is_start
+        attr_reader :profile, :is_start, :anonymous, :is_main
+        alias :anonymous? :anonymous
 
-        def initialize(ruby_thread_id, profile, categorizer, name:, tid:, samples:, weights:, timestamps: nil, sample_categories: nil, markers:, started_at:, stopped_at: nil, allocations: nil, is_main: nil, is_start: nil)
+        def initialize(ruby_thread_id, profile, categorizer, name:, tid:, samples:, weights:, timestamps: nil, sample_categories: nil, markers:, started_at:, anonymous:, stopped_at: nil, allocations: nil, is_main: nil, is_start: nil)
           @ruby_thread_id = ruby_thread_id
           @profile = profile
           @categorizer = categorizer
           @tid = tid
           @name = name
           @is_main = is_main
+          @anonymous = anonymous
           if is_main.nil?
             @is_main = @ruby_thread_id == ::Thread.main.object_id
           end
@@ -438,7 +449,6 @@ module Vernier
             funcTable: func_table,
             nativeSymbols: {},
             samples: samples_table,
-            jsAllocations: allocations_table,
             stackTable: stack_table,
             resourceTable: {
               length: 0,
@@ -449,7 +459,10 @@ module Vernier
             },
             markers: markers_table,
             stringArray: string_table
-          }.compact
+          }.tap do |thread|
+            allocations = allocations_table
+            thread[:jsAllocations] = allocations if allocations
+          end
         end
 
         def markers_table
@@ -638,18 +651,7 @@ module Vernier
 
         def string_table
           @strings.keys.map do |string|
-            if string.ascii_only?
-              string
-            elsif string.encoding == Encoding::UTF_8
-              if string.valid_encoding?
-                string
-              else
-                string.scrub
-              end
-            else
-              # TODO: We might want to guess UTF-8 and escape the binary more explicitly
-              string.dup.force_encoding("UTF-8").scrub
-            end
+            sanitize_string(string)
           end
         end
 

data/lib/vernier/output/top.rb

@@ -1,8 +1,12 @@
 # frozen_string_literal: true
 
+require_relative "../output_helpers"
+
 module Vernier
   module Output
     class Top
+      include OutputHelpers
+
       def initialize(profile, row_limit)
         @profile = profile
         @row_limit = row_limit
@@ -57,10 +61,10 @@ module Vernier
             @profile._stack_table
           end
 
-        stack_weights = Hash.new(0)
-        thread[:samples].zip(thread[:weights]) do |stack_idx, weight|
-          stack_weights[stack_idx] += weight
-        end
+        samples = thread[:samples]
+        weights = thread[:weights]
+
+        stack_weights = collapse_stack_weights(samples, weights)
 
         total = stack_weights.values.sum
 

data/lib/vernier/output_helpers.rb

@@ -0,0 +1,33 @@
+# frozen_string_literal: true
+
+module Vernier
+  module OutputHelpers
+    # Collapses parallel samples/weights arrays into a hash of total weight
+    # per stack index. Retained-mode profiles have one sample per object and
+    # many samples share the same stack, so collapsing first keeps later
+    # aggregation proportional to the number of unique stacks.
+    def collapse_stack_weights(samples, weights)
+      stack_weights = Hash.new(0)
+      samples.each_with_index do |stack_idx, idx|
+        stack_weights[stack_idx] += weights[idx]
+      end
+      stack_weights
+    end
+
+    # Returns a string safe to embed in JSON output: valid UTF-8 with any
+    # invalid bytes replaced. Method names and paths can carry arbitrary
+    # encodings (e.g. BINARY or Shift_JIS source), which would otherwise
+    # raise JSON::GeneratorError.
+    def sanitize_string(string)
+      if string.ascii_only?
+        string
+      elsif string.encoding == Encoding::UTF_8
+        string.valid_encoding? ? string : string.scrub
+      else
+        # We don't know the real encoding; interpret the bytes as UTF-8
+        # and replace anything invalid.
+        string.dup.force_encoding(Encoding::UTF_8).scrub
+      end
+    end
+  end
+end

data/lib/vernier/thread_names.rb

@@ -3,6 +3,7 @@ module Vernier
   class ThreadNames
     def initialize
       @names = {}
+      @anonymous = {}
       @tp = TracePoint.new(:thread_end) do |e|
         collect_thread(e.self)
       end
@@ -13,6 +14,10 @@ module Vernier
       @names[object_id] || "thread obj_id:#{object_id}"
     end
 
+    def anonymous?(object_id)
+      @anonymous[object_id]
+    end
+
     def finish
       collect_running
       @tp.disable
@@ -27,6 +32,7 @@ module Vernier
     end
 
     def collect_thread(th)
+      @anonymous[th.object_id] = !th.name
       @names[th.object_id] = pretty_name(th)
     end
 

data/lib/vernier/version.rb

@@ -1,5 +1,5 @@
 # frozen_string_literal: true
 
 module Vernier
-  VERSION = "1.10.1"
+  VERSION = "1.12.0"
 end

data/vernier.gemspec

@@ -9,12 +9,12 @@ Gem::Specification.new do |spec|
   spec.email = ["john@hawthorn.email"]
 
   spec.summary = "A next generation CRuby profiler"
-  spec.description = "Next-generation Ruby 3.2.1+ sampling profiler. Tracks multiple threads, GVL activity, GC pauses, idle time, and more."
+  spec.description = "Next-generation Ruby 3.3+ sampling profiler. Tracks multiple threads, GVL activity, GC pauses, idle time, and more."
   spec.homepage = "https://github.com/jhawthorn/vernier"
   spec.license = "MIT"
 
   unless ENV["IGNORE_REQUIRED_RUBY_VERSION"]
-    spec.required_ruby_version = ">= 3.2.1"
+    spec.required_ruby_version = ">= 3.3.0"
   end
 
   spec.metadata["homepage_uri"] = spec.homepage

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: vernier
 version: !ruby/object:Gem::Version
-  version: 1.10.1
+  version: 1.12.0
 platform: ruby
 authors:
 - John Hawthorn
@@ -51,7 +51,7 @@ dependencies:
     - - ">="
       - !ruby/object:Gem::Version
         version: '0'
-description: Next-generation Ruby 3.2.1+ sampling profiler. Tracks multiple threads,
+description: Next-generation Ruby 3.3+ sampling profiler. Tracks multiple threads,
   GVL activity, GC pauses, idle time, and more.
 email:
 - john@hawthorn.email
@@ -106,6 +106,7 @@ files:
 - lib/vernier/output/firefox.rb
 - lib/vernier/output/markdown.rb
 - lib/vernier/output/top.rb
+- lib/vernier/output_helpers.rb
 - lib/vernier/parsed_profile.rb
 - lib/vernier/result.rb
 - lib/vernier/stack_table.rb
@@ -127,14 +128,14 @@ required_ruby_version: !ruby/object:Gem::Requirement
   requirements:
   - - ">="
     - !ruby/object:Gem::Version
-      version: 3.2.1
+      version: 3.3.0
 required_rubygems_version: !ruby/object:Gem::Requirement
   requirements:
   - - ">="
     - !ruby/object:Gem::Version
       version: '0'
 requirements: []
-rubygems_version: 4.0.6
+rubygems_version: 4.0.20
 specification_version: 4
 summary: A next generation CRuby profiler
 test_files: []