nested_select_profiling

nested_select_profiling.rb

# This is a simple gist for performance and RAM usage testing for a nested_select gem
# 
# Objects relations: course -> topics -> lessons ( all relations are has_many )
# 
# ------------------------------------------------------------------------- 
# Schemas: 
# 
#   create_table "lessons", force: :cascade do |t|
#     t.string "title", null: false, comment: "Тема урока"
#     t.string "description", null: false, comment: "Описание урока"
#     t.integer "topic_id", null: false
#     t.datetime "created_at", precision: 6, null: false
#     t.datetime "updated_at", precision: 6, null: false
#     t.integer "partition", default: 1, null: false
#     t.integer "position", default: 0
#     t.integer "parent_id", default: 0
#     t.integer "education_year"
#     t.text "text"
#     t.boolean "actived", default: true
#     t.boolean "is_public", default: false
#     t.boolean "online", default: false
#     t.datetime "deleted_at"
#     t.integer "kind"
#     t.string "recording_link"
#     t.jsonb "body"
#     t.datetime "content_updated_at"
#     t.index ["content_updated_at"], name: "index_lessons_on_content_updated_at", where: "(content_updated_at IS NOT NULL)"
#     t.index ["topic_id"], name: "index_lessons_on_topic_id"
#   end
#  
#  create_table "topics", force: :cascade do |t|
#     t.string "title", null: false, comment: "Тема модуля"
#     t.string "description", null: false, comment: "Описание модуля"
#     t.integer "course_id", null: false
#     t.datetime "created_at", precision: 6, null: false
#     t.datetime "updated_at", precision: 6, null: false
#     t.integer "partition", default: 1, null: false
#     t.integer "position", default: 0
#     t.boolean "free"
#     t.datetime "deleted_at"
#     t.integer "kind"
#     t.integer "lessons_count", default: 0, null: false
#     t.index "to_tsvector('russian'::regconfig, (title)::text)", name: "gin_index_topics_on_title", using: :gin
#     t.index ["course_id"], name: "index_topics_on_course_id"
#   end
#
# -------------------------------------------------------------------------
# 
# Rem: Outside of this method I have preselected a set of course ids, and providing them via root_ids attribute
#
require 'nested_select'

def compare_nested_select( root_ids, root_collection_size, silence_ar_logger_for_memory_profiling: true)
  puts "\n------- CPU comparison, for root_collection_size: #{root_collection_size} ----"
  ActiveRecord::Base.logger.silence do
    Benchmark.bm do |benchmark|
      benchmark.report("nested_select") do
        50.times do
          Topic.where(course_id: root_ids.sample(root_collection_size)).includes(:lessons)
               .select(:id, :position, :title, :course_id, lessons: [:id, :title, :topic_id, :position]).load
        end
      end

      benchmark.report("simple includes") do
        50.times do
          Topic.where(course_id: root_ids.sample(root_collection_size)).includes(:lessons)
               .select(:id, :position, :title, :course_id,).load
        end
      end
    end
  end
  
  puts "\n----------------- Memory comparison, for root_collection_size: #{root_collection_size} ---------"
  ids_sample = root_ids.sample(root_collection_size) # it should be the same id for both tests, since we comparing memory consumption

  ActiveRecord::Base.logger.level = :error if silence_ar_logger_for_memory_profiling
  ns = MemoryProfiler.report do
    Topic.where(course_id: ids_sample)
         .includes(:lessons)
         .select(:id, :position, :title, :course_id, lessons: [:id, :title, :topic_id, :position]).load
  end

  direct_include = MemoryProfiler.report do
    Topic.where(course_id: ids_sample).select(:id, :position, :title, :course_id,).includes(:lessons).load
  end
  
  ActiveRecord::Base.logger.level = :debug if silence_ar_logger_for_memory_profiling
  
  puts "------ Nested Select memory consumption for root_collection_size: #{root_collection_size} ------"
  ns.pretty_print(detailed_report: false, scale_bytes: true)
  
  puts "------ Full preloading memory consumption for root_collection_size: #{root_collection_size} ----"
  direct_include.pretty_print(detailed_report: false, scale_bytes: true)
  
  puts "RAM ratio improvements x#{direct_include.total_retained_memsize.to_f / ns.total_retained_memsize} on retain objects"
  puts "RAM ratio improvements x#{direct_include.total_allocated_memsize.to_f / ns.total_allocated_memsize} on total_allocated objects"
end