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
Fixing a performance issue with the mustache gem
Search
Dominic Baggott
October 12, 2015
Programming
0
500
Fixing a performance issue with the mustache gem
@alicebartlett and I gave this talk at LRUG, on Monday 12th October 2015.
Dominic Baggott
October 12, 2015
Tweet
Share
Other Decks in Programming
See All in Programming
Team topologies and the microservice architecture: a synergistic relationship
cer
PRO
0
1k
関数型まつりレポート for JuliaTokai #22
antimon2
0
150
Go1.25からのGOMAXPROCS
kuro_kurorrr
1
800
Webの外へ飛び出せ NativePHPが切り拓くPHPの未来
takuyakatsusa
2
320
童醫院敏捷轉型的實踐經驗
cclai999
0
180
Beyond Portability: Live Migration for Evolving WebAssembly Workloads
chikuwait
0
390
Create a website using Spatial Web
akkeylab
0
300
GoのGenericsによるslice操作との付き合い方
syumai
3
680
なぜ適用するか、移行して理解するClean Architecture 〜構造を超えて設計を継承する〜 / Why Apply, Migrate and Understand Clean Architecture - Inherit Design Beyond Structure
seike460
PRO
1
660
Effect の双対、Coeffect
yukikurage
5
1.4k
#kanrk08 / 公開版 PicoRubyとマイコンでの自作トレーニング計測装置を用いたワークアウトの理想と現実
bash0c7
1
290
[初登壇@jAZUG]アプリ開発者が気になるGoogleCloud/Azure+wasm/wasi
asaringo
0
130
Featured
See All Featured
Building Flexible Design Systems
yeseniaperezcruz
328
39k
Why You Should Never Use an ORM
jnunemaker
PRO
56
9.4k
Connecting the Dots Between Site Speed, User Experience & Your Business [WebExpo 2025]
tammyeverts
5
210
How GitHub (no longer) Works
holman
314
140k
Fantastic passwords and where to find them - at NoRuKo
philnash
51
3.3k
The Power of CSS Pseudo Elements
geoffreycrofte
77
5.8k
Dealing with People You Can't Stand - Big Design 2015
cassininazir
367
26k
Sharpening the Axe: The Primacy of Toolmaking
bcantrill
44
2.4k
Unsuck your backbone
ammeep
671
58k
Optimizing for Happiness
mojombo
379
70k
Designing for humans not robots
tammielis
253
25k
Bootstrapping a Software Product
garrettdimon
PRO
307
110k
Transcript
fixing a performance issue with the mustache gem (a bit
less boring than it sounds)
@alicebartlett & @evilstreak
We used to work at GDS, on gov.uk
In this talk: - Adding a feature - Finding a
bug - working around the bug - fixing the bug
GOV.uk is the best place to find government services and
information
These are Manuals They are multi- page long form documents
New feature: search within a manual
This is search within a manual A list of results
from a Manual with a nested list of results from the rest of GOV.UK
From this:
To this: From this:
None
Search pages use mustache (specifically: the ruby implementation of mustache
via the mustache gem)
From this:
To this: From this:
To this: then from this:
This is very uninteresting.
Using partials in this way is in the docs: from:
https://mustache.github.io/mustache.5.html#Partials
full PR is here: https://github.com/alphagov/frontend/pull/784
Time to deploy!
This should have been straightforward
None
-
-
We deploy to staging
We proceed to production
Slowly, shit hits the fan.
-
We rollback and investigate on staging
Breaking things isn’t bad if you have a good plan
to recover
Profiling
We found out the problem was this CPU spike
# Profile the code RubyProf.start ... [code to profile] ...
result = RubyProf.stop # Print a flat profile to text printer = RubyProf::FlatPrinter.new(result) printer.print(STDOUT) Measuring code with ruby-prof is easy
Profile of our code %self total self wait child calls
name 5.41 0.271 0.086 0.000 0.185 675 Mustache::Parser#scan_tags 3.90 0.062 0.062 0.000 0.000 36 BasicObject#instance_eval 3.34 1.031 0.053 0.000 0.978 1 ActionView::CompiledTemplates#_app 2.98 0.047 0.047 0.000 0.000 6 <Class::Dir>#[] 2.52 0.042 0.040 0.001 0.001 74 <Module::ActiveSupport::Multibyte> 2.36 0.040 0.037 0.000 0.002 1095 Mustache::Parser#regexp 2.08 0.051 0.033 0.000 0.018 46 <Module::Marshal>#load 1.99 0.032 0.032 0.000 0.000 6 Kernel#require 1.98 0.032 0.032 0.000 0.000 2181 String#to_s 1.67 0.506 0.027 0.000 0.480 1007 *Array#each 1.55 0.050 0.025 0.000 0.025 972 Psych::ScalarScanner#tokenize 1.48 0.024 0.024 0.000 0.000 168 <Class::File>#exist? 1.35 0.068 0.021 0.000 0.046 663 Mustache::Parser#position 1.31 0.066 0.021 0.000 0.045 675 Mustache::Parser#scan_text
Profile of our code %self total self wait child calls
name 5.41 0.271 0.086 0.000 0.185 675 Mustache::Parser#scan_tags 3.90 0.062 0.062 0.000 0.000 36 BasicObject#instance_eval 3.34 1.031 0.053 0.000 0.978 1 ActionView::CompiledTemplates#_app 2.98 0.047 0.047 0.000 0.000 6 <Class::Dir>#[] 2.52 0.042 0.040 0.001 0.001 74 <Module::ActiveSupport::Multibyte> 2.36 0.040 0.037 0.000 0.002 1095 Mustache::Parser#regexp 2.08 0.051 0.033 0.000 0.018 46 <Module::Marshal>#load 1.99 0.032 0.032 0.000 0.000 6 Kernel#require 1.98 0.032 0.032 0.000 0.000 2181 String#to_s 1.67 0.506 0.027 0.000 0.480 1007 *Array#each 1.55 0.050 0.025 0.000 0.025 972 Psych::ScalarScanner#tokenize 1.48 0.024 0.024 0.000 0.000 168 <Class::File>#exist? 1.35 0.068 0.021 0.000 0.046 663 Mustache::Parser#position 1.31 0.066 0.021 0.000 0.045 675 Mustache::Parser#scan_text Method name Time Calls
Profile of our code %self total self wait child calls
name 5.41 0.271 0.086 0.000 0.185 675 Mustache::Parser#scan_tags 3.90 0.062 0.062 0.000 0.000 36 BasicObject#instance_eval 3.34 1.031 0.053 0.000 0.978 1 ActionView::CompiledTemplates#_app 2.98 0.047 0.047 0.000 0.000 6 <Class::Dir>#[] 2.52 0.042 0.040 0.001 0.001 74 <Module::ActiveSupport::Multibyte> 2.36 0.040 0.037 0.000 0.002 1095 Mustache::Parser#regexp 2.08 0.051 0.033 0.000 0.018 46 <Module::Marshal>#load 1.99 0.032 0.032 0.000 0.000 6 Kernel#require 1.98 0.032 0.032 0.000 0.000 2181 String#to_s 1.67 0.506 0.027 0.000 0.480 1007 *Array#each 1.55 0.050 0.025 0.000 0.025 972 Psych::ScalarScanner#tokenize 1.48 0.024 0.024 0.000 0.000 168 <Class::File>#exist? 1.35 0.068 0.021 0.000 0.046 663 Mustache::Parser#position 1.31 0.066 0.021 0.000 0.045 675 Mustache::Parser#scan_text
Before and after self calls name self calls name 0.051
2 BasicObject#instance_eva 0.086 675 Mustache::Parser#scan_ta 0.042 6 <Class::Dir>#[] 0.062 36 BasicObject#instance_eva 0.014 110 Mustache::Parser#scan_ta 0.053 1 ActionView::CompiledTemp 0.014 337 *Array#each 0.047 6 <Class::Dir>#[] 0.004 145 Sprockets::Asset#depende 0.040 74 <Module::ActiveSupport:: 0.004 149 *Array#map 0.037 1095 Mustache::Parser#regexp 0.003 110 Mustache::Parser#scan_te 0.033 46 <Module::Marshal>#load 0.003 109 Mustache::Parser#positio 0.032 6 Kernel#require 0.003 110 Mustache::Parser#scan_un 0.032 2181 String#to_s 0.002 2 <Class::IO>#binread 0.027 1007 *Array#each 0.002 186 <Class::File>#exist? 0.025 972 Psych::ScalarScanner#tok 0.002 774 Kernel#is_a? 0.024 168 <Class::File>#exist? 0.002 186 Hike::Trail#stat 0.021 663 Mustache::Parser#positio 0.002 232 String#split 0.021 675 Mustache::Parser#scan_te
compiling templates is slow and partials are compiled before each
use
Recursion
Our results page has a list within a list
This is broken This is OK
- Iteration - Repetition - Recursion
- Iteration - Repetition - Recursion
Repetition
Repetition
This is fast but ghastly
- Iteration - Repetition ?? - Recursion
Recursion is a bit of a dick move
Recursion: Use responsibly
_results.mustache: {{#results}} {{#is_multiple_results}} <ol class= >{{>_results}}</ol> {{/is_multiple_results}} {{^is_multiple_results}} <li> result
code goes here </li> {{/is_multiple_results}} {{/results}}
_results.mustache: {{#results}} {{#is_multiple_results}} <ol class= >{{>_results}}</ol> /*Call self*/ {{/is_multiple_results}} {{^is_multiple_results}}
<li> result code goes here </li> {{/is_multiple_results}} {{/results}}
The problem with this is it’s non-obvious
- Iteration - Repetition ?? - Recursion ??
None
Reader, we chose the recursion.
Fixing the gem
Fixing the gem was the right thing to do
Fixing the gem prevents trouble in the future
Looking for prior art
The fix
We added a benchmark instead of a test
Writing the pull request https://github.com/mustache/mustache/pull/205
Explain the problem
Explain the change
Be proactive about objections
Keep nudging for progress
Over half our effort went on getting the change accepted
OK thanks bye
Links
Questions?