logidze 0.8.1 → 1.0.0.rc1

data/bench/performance/setup.rb

@@ -1,308 +0,0 @@
-begin
-  require 'bundler/inline'
-rescue LoadError => e
-  $stderr.puts 'Bundler version 1.10 or later is required. Please update your Bundler'
-  raise e
-end
-
-gemfile(true) do
-  source 'https://rubygems.org'
-  gem 'activerecord', '~>4.2'
-  gem 'pg'
-  gem 'paper_trail', '~>4.2', require: false
-  gem 'pry-byebug'
-  gem 'faker'
-  gem 'benchmark-ips'
-  gem 'memory_profiler'
-end
-
-DB_NAME = ENV['DB_NAME'] || 'logidze_query_bench'
-
-begin
-  system("createdb #{DB_NAME}")
-rescue
-  $stdout.puts "DB already exists"
-end
-
-$LOAD_PATH.unshift File.expand_path('../../../lib', __FILE__)
-
-require 'active_record'
-require 'logger'
-require 'logidze'
-
-ActiveRecord::Base.send :include, Logidze::HasLogidze
-
-ActiveRecord::Base.establish_connection(adapter: 'postgresql', database: DB_NAME)
-
-at_exit do
-  ActiveRecord::Base.connection.disconnect!
-end
-
-require 'paper_trail'
-
-module LogidzeBench
-  module_function
-  def setup_db
-    ActiveRecord::Schema.define do
-      # PaperTrail setup
-      create_table :versions, force: true do |t|
-        t.string :item_type, null: false
-        t.integer :item_id, null: false
-        t.string :event, null: false
-        t.string :whodunnit
-        t.text :object
-        t.jsonb :object_changes
-
-        t.datetime :created_at
-      end
-
-      add_index :versions, [:item_type, :item_id]
-
-      # Logidze setup
-      enable_extension :hstore
-
-      execute <<~SQL
-        DO $$
-          BEGIN
-          EXECUTE 'ALTER DATABASE ' || current_database() || ' SET logidze.disabled TO off';
-          END;
-        $$
-        LANGUAGE plpgsql;
-      SQL
-
-      execute <<~SQL
-        CREATE OR REPLACE FUNCTION logidze_logger() RETURNS TRIGGER AS $body$
-          DECLARE
-            changes jsonb;
-            new_v integer;
-            ts bigint;
-            size integer;
-            history_limit integer;
-            current_version integer;
-            merged jsonb;
-            iterator integer;
-            item record;
-          BEGIN
-            ts := (extract(epoch from now()) * 1000)::bigint;
-
-            IF TG_OP = 'INSERT' THEN
-              changes := to_jsonb(NEW.*) - 'log_data';
-              new_v := 1;
-
-              NEW.log_data := json_build_object(
-                'v',
-                1,
-                'h',
-                jsonb_build_array(
-                  jsonb_build_object(
-                    'ts',
-                    ts,
-                    'v',
-                    new_v,
-                    'c',
-                    changes
-                  )
-                )
-              );
-            ELSIF TG_OP = 'UPDATE' THEN
-              history_limit := TG_ARGV[0];
-              current_version := (NEW.log_data->>'v')::int;
-
-              IF NEW = OLD THEN
-                RETURN NEW;
-              END IF;
-
-              IF current_version < (NEW.log_data#>>'{h,-1,v}')::int THEN
-                iterator := 0;
-                FOR item in SELECT * FROM jsonb_array_elements(NEW.log_data->'h')
-                LOOP
-                  IF (item.value->>'v')::int > current_version THEN
-                    NEW.log_data := jsonb_set(
-                      NEW.log_data,
-                      '{h}',
-                      (NEW.log_data->'h') - iterator
-                    );
-                  END IF;
-                  iterator := iterator + 1;
-                END LOOP;
-              END IF;
-
-              changes := hstore_to_jsonb_loose(
-                hstore(NEW.*) - hstore(OLD.*)
-              ) - 'log_data';
-
-              new_v := (NEW.log_data#>>'{h,-1,v}')::int + 1;
-
-              size := jsonb_array_length(NEW.log_data->'h');
-
-              NEW.log_data := jsonb_set(
-                NEW.log_data,
-                ARRAY['h', size::text],
-                jsonb_build_object(
-                  'ts',
-                  ts,
-                  'v',
-                  new_v,
-                  'c',
-                  changes
-                ),
-                true
-              );
-
-              NEW.log_data := jsonb_set(
-                NEW.log_data,
-                '{v}',
-                to_jsonb(new_v)
-              );
-
-              IF history_limit IS NOT NULL AND history_limit = size THEN
-                merged := jsonb_build_object(
-                  'ts',
-                  NEW.log_data#>'{h,1,ts}',
-                  'v',
-                  NEW.log_data#>'{h,1,v}',
-                  'c',
-                  (NEW.log_data#>'{h,0,c}') || (NEW.log_data#>'{h,1,c}')
-                );
-
-                NEW.log_data := jsonb_set(
-                  NEW.log_data,
-                  '{h}',
-                  jsonb_set(
-                    NEW.log_data->'h',
-                    '{1}',
-                    merged 
-                  ) - 0
-                );
-              END IF;
-            END IF;
-
-            return NEW;
-          END;
-          $body$
-          LANGUAGE plpgsql;
-      SQL
-
-      create_table :users, force: true do |t|
-        t.string :email
-        t.integer :position
-        t.string :name
-        t.text :bio
-        t.integer :age
-        t.timestamps
-      end
-
-      create_table :logidze_users, force: true do |t|
-        t.string :email
-        t.integer :position
-        t.string :name
-        t.text :bio
-        t.integer :age
-        t.jsonb :log_data, default: '{}', null: false
-        t.timestamps
-      end
-
-      execute <<~SQL
-        CREATE TRIGGER logidze_on_logidze_users
-        BEFORE UPDATE OR INSERT ON logidze_users FOR EACH ROW
-        WHEN (current_setting('logidze.disabled') <> 'on')
-        EXECUTE PROCEDURE logidze_logger();
-      SQL
-    end
-  end
-
-  module_function
-  def populate(n = 1_000)
-    n.times do
-      params = fake_params
-      User.create!(params)
-      LogidzeUser.create!(params)
-    end
-  end
-
-  module_function
-  def cleanup
-    LogidzeUser.delete_all
-    User.delete_all
-    PaperTrail::Version.delete_all
-  end
-
-  module_function
-  def generate_versions(num = 1)
-    num.times do 
-      User.find_each do |u|
-        u.update!(fake_params(sample: true))
-      end
-
-      LogidzeUser.find_each do |u|
-        u.update!(fake_params(sample: true))
-      end
-
-      # make at least 1 second between versions
-      sleep 1
-    end
-    Time.now
-  end
-
-  module_function
-  def fake_params(sample: false)
-    params = {
-      email: Faker::Internet.email,
-      position: Faker::Number.number(3),
-      name: Faker::Name.name,
-      age: Faker::Number.number(2),
-      bio: Faker::Lorem.paragraph
-    }
-
-    return params.slice(%i(email position name age bio).sample) if sample
-    params
-  end
-end
-
-module ARandom
-  def random(num = 1)
-    rel = order('random()')
-    num == 1 ? rel.first : rel.limit(num)
-  end
-end
-
-class User < ActiveRecord::Base
-  extend ARandom
-  has_paper_trail
-
-  def self.diff_from(ts)
-    includes(:versions).map { |u| { 'id' => u.id, 'changes' => u.diff_from(ts) } }
-  end
-
-  def self.diff_from_joined(ts)
-    eager_load(:versions).map { |u| { 'id' => u.id, 'changes' => u.diff_from(ts) } }
-  end
-
-  def diff_from(ts)
-    changes = {}
-    versions.each do |v|
-      next if v.created_at < ts
-      merge_changeset(changes, v.changeset)
-    end
-    changes
-  end
-
-  private
-
-  def merge_changeset(acc, data)
-    data.each do |k,v|
-      unless acc.key?(k)
-        acc[k] = { 'old' => v[0] }
-      end
-      acc[k]['new'] = v[1]
-    end
-  end
-end
-
-class LogidzeUser < ActiveRecord::Base
-  extend ARandom
-  has_logidze
-end
-
-# Run migration only if neccessary
-LogidzeBench.setup_db if ENV['FORCE'].present? || !ActiveRecord::Base.connection.tables.include?('logidze_users')

data/bench/performance/update_bench.rb

@@ -1,36 +0,0 @@
-require 'benchmark/ips'
-require './setup'
-
-params = {
-  age: Faker::Number.number(2),
-  email: Faker::Internet.email
-}
-
-params2 = {
-  email: Faker::Internet.email,
-  position: Faker::Number.number(3),
-  name: Faker::Name.name,
-  age: Faker::Number.number(2),
-  bio: Faker::Lorem.paragraph
-}
-
-LogidzeBench.cleanup
-LogidzeBench.populate
-
-Benchmark.ips do |x|
-  x.report('PT UPDATE #1') do
-    User.random.update!(params)
-  end
-
-  x.report('Logidze UPDATE #1') do
-    LogidzeUser.random.update!(params)
-  end
-
-  x.report('PT UPDATE #2') do
-    User.random.update!(params2)
-  end
-
-  x.report('Logidze UPDATE #2') do
-    LogidzeUser.random.update!(params2)
-  end
-end

data/bench/triggers/Makefile

@@ -1,56 +0,0 @@
-# Database name
-ifndef DB
-	DB = logidze_bench
-endif
-
-# Transactions count
-ifndef T
-	T = 10000
-endif
-
-all: plain hstore jsonb jsonb2 keys keys2
-
-setup:
-	createdb $(DB) -w
-	psql -q -d $(DB) -c 'CREATE EXTENSION IF NOT EXISTS hstore;'
-
-plain:
-	$(info )
-	$(info ======  [START] Update without triggers ======)
-	pgbench -i -q $(DB)
-	pgbench -f bench.sql -t $(T) -r $(DB)
-
-hstore:
-	$(info )
-	$(info ======  [START] Update with hstore-based triggers ======)
-	pgbench -i -q $(DB)
-	psql -q -d $(DB) -f hstore_trigger_setup.sql 
-	pgbench -f bench.sql -t $(T) -r $(DB)
-
-jsonb:
-	$(info )
-	$(info ======  [START] Update with jsonb-minus triggers ======)
-	pgbench -i -q $(DB)
-	psql -q -d $(DB) -f jsonb_minus_setup.sql 
-	pgbench -f bench.sql -t $(T) -r $(DB)
-
-jsonb2:
-	$(info )
-	$(info ======  [START] Update with jsonb-minus triggers ======)
-	pgbench -i -q $(DB)
-	psql -q -d $(DB) -f jsonb_minus_2_setup.sql 
-	pgbench -f bench.sql -t $(T) -r $(DB)
-
-keys:
-	$(info )
-	$(info ======  [START] Update with loop thru keys triggers (v1) ======)
-	pgbench -i -q $(DB)
-	psql -q -d $(DB) -f keys_trigger_setup.sql
-	pgbench -f bench.sql -t $(T) -r $(DB) 
-
-keys2:
-	$(info )
-	$(info ======  [START] Update with loop thru keys triggers (v2) ======)
-	pgbench -i -q $(DB)
-	psql -q -d $(DB) -f keys2_trigger_setup.sql
-	pgbench -f bench.sql -t $(T) -r $(DB)  

data/bench/triggers/Readme.md

@@ -1,58 +0,0 @@
-# Triggers benchmarks
-
-This benchmark uses standard _pg\_bench_ table `pgbench_accounts`.
-We consider several approaches for calculating records diff: one uses _hstore_ extension, two uses jsonb functions and two others iterate through record fields.
-
-# Usage
-
-Create database:
-
-```sh
-make setup
-```
-
-You can provide database name by `DB` variable (defaults to "logidze_bench").
-
-Run all benchmarks:
-
-```sh
-make
-```
-
-or separate benchmark:
-
-```sh
-make hstore
-
-make jsonb
-
-make jsonb2
-
-make keys
-
-make keys2
-
-# Raw update, no triggers
-make plain
-```
-
-You can specify the number of transactions by `T` variable (defaults to 10000):
-
-```sh
-make T=1000000
-```
-
-# Results
-
-The benchmark shows that hstore variant is the most efficient (running on MacPro 2013, 2.4 GHz Core i5, 4GB, SSD, 1 million transactions per test):
-
-|Mode    | TPS  | Statement latency (ms) |
-|--------|------|------------------------|
-| plain  | 3628 | 0.113                  |
-| hstore | 3015 | 0.168                  |
-| jsonb  | 1647 | 0.363                  |
-| jsonb2 | 1674 | 0.354                  |
-| keys   | 2355 | 0.219                  |
-| keys2  | 2542 | 0.210                  |
-
-_Logidze_ uses hstore variant.