2. This talk
FLASK NYC 2013
E x p e r i e n c e
a n d L e s s o n s
F l a s k
@
H a p p i f y
A n 1 8 - m o n t h - o l d
a c t i v e F l a s k
c o d e b a s e
H o w I g o t t o
F l a s k
Flask has become a
VIBRANT, ACTIVE toolkit
for HTTP servers.
But there aren’t a lot of
large examples from which
to learn. LET’S CHANGE
THAT.
5. what is a “microframework?”
•just the essentials, nothing more
•extensibility through plugins
•readable source
•minimal magic
•easy to build a toy/demo
•lack of patterns for larger projects
6. FLASK NYC 2013
PROJECT LINES OF CODE COMPLEXITY
Backbone.js < 1000 LOW
Flask (not Werkzeug) ~2000 (?) LOW
Redis 100,000 MEDIUM
Django 170,000 MEDIUM
MongoDB 500,000 HIGH
PostgreSQL > 1,000,000 HIGH!
complexity spectrum
* STATS FROM OHLOH (http://www.ohloh.net/) and http://www.andreas-dewes.de/code_is_beautiful/
7. ‣ java/spring/.NET
‣ a FEW rails-based startups (sinatra mixed in)
‣ scala services (scalatra, play)
‣ Python a natural fit for happify (data analysis, machine learning, nlp)
‣ flask selected as toolkit
my journey to flask
8. A Sampling of NYC Python
A Startup Junkie's Journey to Python via Java, Ruby and Scala
Andy Parsons
@andyparsons
My Startups and Their Stacks What Are We Building? And in general...
Happiness Speed Magic Readability
BUSINESS WEB API SERVICES DATABASES TEAM SIZE
Pro Photo ASP.NET C#, Perl SQL Server 12
Hyperlocal
Content
(Geolocation
)
RoR Sinatra Scala
PostgreSQL,
MongoDB
10
Ereading and
Book
Recommende
r
RoR Scalatra Scala/Java MongoDB 8
Life Gaming Flask
Python/
Gevent
Python
PostgreSQL,
MongoDB
3
Building Happify: Comparing Language Options
SPEED CONCURRENCY TOOLING
PACKAGING /
DEPENDENCIES
DEPLOYMENT TESTING
FRAMEWORKS
/ LIBS
RUBY
SCALA
PYTHON
Fast enough Achieved through Process.fork
Healthy, lots of
choice
Gems Capistrano Mature
Growing collection
of libs, but weak in
NLP
Fast! Real concurrency,Akka
Still early. IDE’s
weak.
Complex. JARs, SBT, Ivy,
Maven.
? Maturing
Early for native
scala libraries, rely
on Java interop
Fast enough
Achieved through
multiprocessing
Healthy, lots of
choice
Eggs Fabric Mature
Massive collection
of libraries
READABILITY
HAPPINESS /
PRODUCTIVITY
COMMUNITY /
ACTIVITY
PROPENSITY FOR
MAGIC
MATURITY /
STABILITY
HIRING
RUBY
SCALA
PYTHON
OK. Emphasis
metaprogramming can
present challenges.
High Huge and active
Metaprogramming is sometimes
abused. RoR is too magical and
many imitated its philosophy
Stable, but much
catchup remaining for
gems to support 1.9
Tough. Medium
learning curve,
hard to find
seasoned devs
Concise, expressive but
NOT simple.
Highly variable! Small and active
Too many ways to accomplish
things. Scala “implicits”
encourage magic
Language spec
changing, breakages
still occurring in dot
releases
Very difficult.
Steep learning
curve
Highly readable High Big and active Explicit better than implicit
Most mature here.
But the 2.x/3.x break
is bad.
OK. Shallow
learning curve.
FLAME BAIT
BASICS
Scala Ruby Python
23
18
8
How do they “Stack” Up?
Happify is an angel-funded consumer
destination “life game.”
HTML5, Mobile, and Facebook front ends.
Currently in stealth mode.
Score is based on:
Green = 2
Yellow = 1
Red = 0
Python
36%
Ruby
60%
Scala
4%
Github Project Relative %
Gratuitous Perf Charts!
* from: http://shootout.alioth.debian.org
pycon 2012
poster
9. what is happify?
‣ science of happiness
‣ positive psych trainer
‣ consumer subscription business
‣ launched in beta after 9 months
‣ ~100K users
‣ imminent public launch
FLASK NYC 2013
10. happify stack abbreviated biography
•FEbruary 2012: james dennis @j2labs for coffee
•march 2012: started coding poc (jinja templates only)
•june 2012: alpha poc launched
•september 2012: first real design/ux work
•october 2012: backbone/coffeescript introduced
•december 2013: redis added, closed beta launched
•march 2013: ES added, open beta launch
•june 2013: commerce, new feature set, scale up to 100K users
•august 2013: ”final” design, refactoring for public launch
•now: frontend and backend performance, press
14. ‣ api for mobile
‣ api for backbone
‣ backbone app bootstrapping
‣ template rendering
‣ administration ui
‣ aggregates data from postgres, redis, elasticsearch
‣ serializes context-specific object graphs
the role of flask - data switch
15. ‣ project layout
‣ config
‣ routing setup
‣ db connection management
‣ serialization: shallow and deep
‣ FAB automation
code
16. ‣ better config (multiple environments, config heirarchy)
‣ authentication
‣ assets pipeline
‣ model validation and serialization layer (waffling)
batteries i wish were included
batteries i could do without
‣ trailing slash redirects (but easy to control)
17. python 3! well, almost kinda
•flask/werkzeug support completed
•many extensions not yet
•fabric, boto, sentry ...
•too bad- unicode is a mess without it
•Handy: http://python3wos.appspot.com/
18. ‣ flask is an excellent choice for apps like ours
‣ know what you are getting and what you aren’t, make no assumptions
‣ prepare to refactor relentlessly
‣ don’t worry that there’s no single way to do it!
conclusions