Upgrade to Pro
— share decks privately, control downloads, hide ads and more …
Speaker Deck
Features
Speaker Deck
PRO
Sign in
Sign up for free
Search
Search
高度なコード
Search
Aaron Patterson
September 22, 2014
Technology
13
1.3k
高度なコード
Aaron Patterson
September 22, 2014
Tweet
Share
More Decks by Aaron Patterson
See All by Aaron Patterson
RubyKaigi 2025: Class New, A New Approach
tenderlove
0
88
RubyKaigi Dev Meeting 2025
tenderlove
1
3.5k
Speeding up Instance Variables in Ruby 3.3
tenderlove
2
450
[Feature #20425] Speeding up delegate methods
tenderlove
3
290
RailsConf 2023
tenderlove
30
1.2k
Don't @ Me! Faster Instance Variables with Object Shapes
tenderlove
1
470
RailsConf 2022 Keynote
tenderlove
2
580
Some Assembly Required
tenderlove
1
590
HexDevs 2021
tenderlove
1
480
Other Decks in Technology
See All in Technology
業務自動化プラットフォーム Google Agentspace に入門してみる #devio2025
maroon1st
0
170
Oracle Cloud Infrastructure:2025年9月度サービス・アップデート
oracle4engineer
PRO
0
310
GopherCon Tour 概略
logica0419
2
160
stupid jj tricks
indirect
0
7.5k
WebアプリケーションのUI構築で気を付けてるポイント
tomokusaba
0
170
生成AIで「お客様の声」を ストーリーに変える 新潮流「Generative ETL」
ishikawa_satoru
1
230
BtoBプロダクト開発の深層
16bitidol
0
130
Azure SynapseからAzure Databricksへ 移行してわかった新時代のコスト問題!?
databricksjapan
0
110
それでも私はContextに値を詰めたい | Go Conference 2025 / go conference 2025 fill context
budougumi0617
4
910
Geospatialの世界最前線を探る [2025年版]
dayjournal
2
440
20250929_QaaS_vol20
mura_shin
0
110
Tomorrow graphlib, Let us use everybody
hayaosuzuki
0
140
Featured
See All Featured
Reflections from 52 weeks, 52 projects
jeffersonlam
352
21k
個人開発の失敗を避けるイケてる考え方 / tips for indie hackers
panda_program
114
20k
Mobile First: as difficult as doing things right
swwweet
224
9.9k
The Web Performance Landscape in 2024 [PerfNow 2024]
tammyeverts
9
840
Improving Core Web Vitals using Speculation Rules API
sergeychernyshev
18
1.2k
Practical Tips for Bootstrapping Information Extraction Pipelines
honnibal
PRO
23
1.5k
Being A Developer After 40
akosma
90
590k
ReactJS: Keep Simple. Everything can be a component!
pedronauck
667
120k
[RailsConf 2023 Opening Keynote] The Magic of Rails
eileencodes
30
9.7k
The Illustrated Children's Guide to Kubernetes
chrisshort
48
51k
Agile that works and the tools we love
rasmusluckow
330
21k
Building an army of robots
kneath
306
46k
Transcript
( ͡ ° ͜ ʖ ͡ °) .oO(Hi)
ຖ߃ྫ$ ͷ$ ຊޠೳྗࢼݧ
Thanks!!!
Aaron Patterson
Ϩου ϋτ
None
None
Ruby Core Rails Core
ί ϛ ο τ ͠ ͯ! ϙ Π ϯ τ
Β ͓ ͏ ʂ
Revert Commits Count Too!
More mistakes == more points!!!!
ΦΨϫ͞Μ
ΰϒͪΌΜ
νϡʔνϡʔ
None
Ϗοάσʔλ
ϝλϧʹ$ ۙ͘ͳΔͨΊʹ
Node.JS
None
Speeding up Rails 4.2
ߟ͑ࣄ
Rack
͏ऴΘΓ
Rack ָ͡Όͳ͍
Rack 2.0 ൃද͠ͳ͍* * Rack 2.0 ൃද͠·͢ɻ
the_metal http://github.com/tenderlove/the_metal
def call(env) end άϩʔόϧม R ack 1.x
ετϦʔϛϯά class Stream def each loop { yield "hello world\n"
} end end $ def call(env) [200, {}, Stream.new] end R ack 1.x
def call(request, response) end IO IO the_m etal
request IOͱ ϦΩΣετͷใ! Λ͍࣋ͬͯ·͢ɻ
responseIOͱ Ϩεϙϯεͷใ! Λ͍࣋ͬͯ·͢ɻ
શ࣮ྫ TheMetal.create_server(->(req, res) { res.write_head 200, 'Content-Type' => 'text/plain' res.write
"Hello World\n" res.finish }).listen 9292, '0.0.0.0'
Great for HTTP/1.1 Extendable to HTTP/ 2.0
RackΛͬͯ ָΛ͠·͠ΐ͏
ߟ͑ࣄ͚ͩɻ
ಡΜͰԼ͍͞ʂ http://github.com/tenderlove/the_metal
Speeding up Rails 4.2
Adequate$ Record
Performance$ Tools
benchmark/ips
benchmark/ips require 'benchmark/ips' require 'set' $ list = ('a'..'zzzz').to_a set
= Set.new list $ Benchmark.ips do |x| x.report("set access") { set.include? "foo" } $ x.report("ary access") { list.include? "foo" } end G EM
ग़ྗ Calculating ------------------------------------- set access 68622 i/100ms ary access 395
i/100ms ------------------------------------------------- set access 3047175.3 (±12.7%) i/s - 14959596 in 5.018692s ary access 3899.2 (±7.1%) i/s - 19750 in 5.096118s ߹ ܭ IPS
Set Include: 3047175.3 / sec
Array Include: 3899.2 / sec
IPS: Higher Is Better
Blackbox Testing
Ωϟογϡͷςετ cache1 = Cache1.new cache2 = Cache2.new $ cache1["x"] =
Object.new cache2["x"] = Object.new $ Benchmark.ips do |x| x.report("cache1") { cache1["x"] } x.report("cache2") { cache2["x"] } end
࣮ߦͷάϥϑ 10,000 ܁Γฦ࣌ؒ͢ (seconds) 0.01 0.1 1 10 100 ΩϟογϡɹαΠζ
10 elements 100 elements 1000 elements 100000 elements Ωϟογϡ 1 Ωϟογϡ 2
Cacheͷ࣮ class Cache1 def initialize @cache = {} end def
[] k; @cache[k]; end def []= k,v; @cache[k] = v; end end $ class Cache2 def initialize @cache = [] end def [] k; x, = @cache.assoc(k); x; end def []= k,v; @cache << [k, v]; end end ఆ ઢܗ
۩ମతͳྫ Routes
Route දͷେ͖͞ link_to ʹؔ͋Δʁ
route ΛೖΕΔ class MyTest routes = ActionDispatch::Routing::RouteSet.new routes.draw { resources(:articles)
N.times do |num| resources num.to_s.to_sym end } end 10, 100, 1000
Sec /100k calls 9.5 9.7 9.9 10.1 10.3 1 2
3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 link_to
URL ͷ͞ ࣮ߦʹ ؔ͋Γ·͔͢ʁ
͞ΛมԽ class MyTest routes = ActionDispatch::Routing::RouteSet.new link = N.times.map(&:to_s).join '/'
$ routes.draw { get "/#{link}/:id", :as => :article, :controller => :articles, :action => :show } end 10, 100, 1000
ඵ / 10ສճͷݺͼग़͠ 9 10.5 12 13.5 15 1 2
3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 link_to
Object Allocations
GC.stat()
߹ܭͷׂΓͯ GC.stat(:total_allocated_object)
۩ମతͳྫɿ Views
Request Benchmark task :allocated_objects do app = Ko1TestApp::Application.instance app.app do_test_task(app)
env = rackenv "/books/new" do_test_task(app, env.dup) before = GC.stat :total_allocated_object TEST_CNT.times { do_test_task(app, env.dup) } after = GC.stat :total_allocated_object puts (after - before) / TEST_CNT end "/books/new"
Request Benchmark task :allocated_objects do app = Ko1TestApp::Application.instance app.app do_test_task(app)
env = rackenv "/books/new" do_test_task(app, env.dup) before = GC.stat :total_allocated_object TEST_CNT.times { do_test_task(app, env.dup) } after = GC.stat :total_allocated_object puts (after - before) / TEST_CNT end
Request Benchmark task :allocated_objects do app = Ko1TestApp::Application.instance app.app do_test_task(app)
env = rackenv "/books/new" do_test_task(app, env.dup) before = GC.stat :total_allocated_object TEST_CNT.times { do_test_task(app, env.dup) } after = GC.stat :total_allocated_object puts (after - before) / TEST_CNT end
ςετͷ݁Ռ Request ͷΦϒδΣΫτஔ 2000 2150 2300 2450 2600 4-0-stable 4-1-stable
master 2000
Ͳ͏ͬͯ$ ӕΛ͔ͭ͘
ςετͷ݁Ռ Request ͷΦϒδΣΫτஔ 0 650 1300 1950 2600 4-0-stable 4-1-stable
master
4-0-stable ͔Β ~19%ݮΓ·ͨ͠
4-1-stable ͔Β ~14%ݮΓ·ͨ͠
allocation_tracer https://github.com/ko1/allocation_tracer
ྫ ObjectSpace::AllocationTracer.trace do 1000.times { ["foo", {}] } end $
ObjectSpace::AllocationTracer.allocated_count_table
Output {:T_NONE=>0, :T_OBJECT=>0, :T_CLASS=>0, :T_MODULE=>0, :T_FLOAT=>0, :T_STRING=>1000, :T_REGEXP=>0, :T_ARRAY=>1000, :T_HASH=>1000,
:T_ZOMBIE=>0}
Speeding up Helpers
Profile Request / Response
TOTAL (pct) SAMPLES (pct) FRAME 813 (9.5%) 813 (9.5%) ActiveSupport::SafeBuffer#initialize
699 (8.1%) 350 (4.1%) block in ActiveRecord::Read#read_attribute 486 (5.7%) 298 (3.5%) ActionController::UrlFor#url_options 670 (7.8%) 274 (3.2%) ActionDispatch::Journey::Format#evaluate 773 (9.0%) 253 (2.9%) ActionDispatch#parameterize_args 1172 (13.6%) 220 (2.6%) ActiveRecord::Persistence#instantiate 213 (2.5%) 213 (2.5%) block in SQLite3::Statement#each 208 (2.4%) 208 (2.4%) ActiveSupport::SafeBuffer#html_safe? 204 (2.4%) 204 (2.4%) ActionDispatch::UrlFor#routes_generation? 245 (2.9%) 191 (2.2%) block (2 levels) in Class#class_attribute
ActiveSupport::SafeBuffer #initialize
Ͳ͜Ͱ࡞ͬͨͷʁ
Tag Options def tag_option(key, value, escape) if value.is_a?(Array) value =
escape ? safe_join(value, " ") : value.join(" ") else value = escape ? ERB::Util.h(value) : value end %(#{key}="#{value}") end
HTML Sanitization in Rails
Ordinary String >> x = "foo" => "foo" >> x.class
=> String >> x.html_safe? => false
SafeBuffer >> x = "foo" => "foo" >> y =
x.html_safe => "foo" >> y.class => ActiveSupport::SafeBuffer >> y.html_safe? => true
ERB::Utils.h
ERB::Utils.h def html_escape(s) s = s.to_s if s.html_safe? s else
s.gsub(HTML_ESCAPE_REGEXP, HTML_ESCAPE).html_safe end end
Creates 2 Objects.
Tag Options def tag_option(key, value, escape) if value.is_a?(Array) value =
escape ? safe_join(value, " ") : value.join(" ") else value = escape ? ERB::Util.h(value) : value end %(#{key}="#{value}") end
String SafeBuffer String
Extract Method def unwrapped_html_escape(s) # :nodoc: s = s.to_s if
s.html_safe? s else s.gsub(HTML_ESCAPE_REGEXP, HTML_ESCAPE) end end $ def html_escape(s) unwrapped_html_escape(s).html_safe end
Update Callers def tag_option(key, value, escape) if value.is_a?(Array) value =
escape ? safe_join(value, " ") : value.join(" ") else value = escape ? ERB::Util.unwrapped_html_escape(value) : value end %(#{key}="#{value}") end
String String
~200 Allocations Per Request for /books/new
YMMV (Your Mileage (kilometers?) May Vary)
Speeding up Output
ERB Template <% books.each do |book| %> <tr><td> <%= book.name
%> </tr></td> <% end %>
Compiled Template @output_buffer = output_buffer || ActionView::OutputBuffer.new;@output_buffer.safe_append='<h1>Listing books</h1> $ <table>
<thead> <tr> <th>Name</th> <th colspan="3"></th> </tr> </thead> $ <tbody> '.freeze; @books.each do |book| @output_buffer.safe_append=' <tr> <td>'.freeze;@output_buffer.append=( book.name );@output_buffer.safe_append='</td> <td>'.freeze;@output_buffer.append=( link_to 'Show', book );@output_buffer.safe_append='</ td> </tr> '.freeze; end @output_buffer.safe_append=' </tbody> </table> $ <br>
Compiled Template @output_buffer = ActionView::OutputBuffer.new @output_buffer.safe_append=' <tr> <td>'.freeze @output_buffer.append=( book.name
) HTML Literal
safe_append= class OutputBuffer def safe_append=(value) return self if value.nil? super(value.to_s)
end end W hy?
Don’t accept nil
safe_append= class OutputBuffer def safe_append=(value) super(value.to_s) end end
safe_append= class OutputBuffer def safe_append=(value) super(value) end end
safe_append= class OutputBuffer def safe_append=(value) super end end
safe_append= class OutputBuffer end ߴͳίʔυʂ
Results
Allocations Per Request 0 275 550 825 1100 T_STRING T_ARRAY
T_HASH T_NODE T_DATA OTHER 4-0-stable 4-1-stable master
~19% reduction since 4-0-stable
~14% reduction since 4-1-stable
·ͱΊɻ
Eliminate Objects
Ұ൪͍ίʔυ ଘࡏ͠ͳ͍ίʔυ
Limit Types
Fewer Types = Less Code
Less Code = Faster Code
Measure, measure measure measure
MEASURE!
Rails 4.2 will be the fastest ever!
͋Γ͕ͱ͏ʂ