Slides Ruby Conf 2014 on using simple techniques to create slimer, clearer models, controllers, and views in Ruby on Rails.
blog: http://www.railsonmaui.com
Code samples: https://github.com/justin808/fat-code-refactoring-techniques
2. Concerns, Decorators, Presenters, Service
Objects, Helpers, Help me Decide!
RailsConf 2014
Chicago
April 22, 2014
!
Justin Gordon
@railsonmaui
Rails Consultant
www.railsonmaui.com
2
11. DHH Quote
JG: "This is starting to boil down to utilize the framework
capabilities and move beyond only when necessary.”
DHH: "Which is really just an extension of KISS (Keep It
Simple, Stupid). When you use the
framework code for what it’s
intended, you’re not cutting
against the grain. You don’t need to write as
much code. It’s clearer to everyone because it’s the same
approach everyone else is taking."
11
13. Refactoring Examples in Pull Requests
• https://github.com/justin808/fat-code-refactoring-techniques/
pulls
• Based on Michael Hartl’s “Rails Tutorial” MicroBlog example
application
13
14. Objectives
Patterns &
Techniques
14
DRY
Methods
< 5 Lines
Classes
< 100 lines
One Instance
Variable in View
Easy to
Test
Concerns
Draper
Decorators
Validation
Classes
Presenters
Split-up
Controllers
Clarity
Easy to
Change
Guidelines
Move Logic
to Models
Easy to
Find
15. • Huge model file with even larger spec file.
• Break up the model/spec using Rails concerns.Try to break it
up by domain, but any logical split will help.
15
Scenario
16. Scenario
• You’ve got duplicated code in two models, different database
tables.
• Tease out a concern that applies to both models. Since your
models extend ActiveRecord::Base, using regular
inheritance is problematic. Instead, use a concern.
16
18. Rails Concerns
18
Big Model
some-domain class macros
some-domain instance methods
some-domain class methods
other class macros
other instance methods
other class methods
Domain Concern
some-domain
class macros
some-domain
instance methods
some-domain
class methods
19. Concerns: How
• Discover set of related code for a problem domain
• Create a module with extends ActiveSupport::Concern
• Move code into the Concern
• Break out tests into corresponding test file for the Concern
19
20. DHH on Domain vs. Technical Refactoring
"I’ve not yet found a case where the scope of the current file/
class couldn’t be brought under control by using a domain-driven
extraction approach."
"In a sea of 60 methods, there will always be
domain-based groupings, rather than technical
groupings. Never seen that not be the case."
20
21. Concerns: Example
• Break out Emailable Concern out of User model
• Captures domain logic of lower case emails on user model
• Benefits: Smaller model, smaller spec
21
22. Objectives
Patterns &
Techniques
22
DRY
Methods
< 5 Lines
Classes
< 100 lines
One Instance
Variable in View
Easy to
Test
Concerns
Draper
Decorators
Validation
Classes
Presenters
Split-up
Controllers
Clarity
Easy to
Change
Guidelines
Move Logic
to Models
Easy to
Find
23. Scenario
• Model file creating detailed validation messages with HTML
tags and URL links.
• Move the message creation code into a Draper Decorator for
the model.These decorators work great for model based
presentation code.
23
27. Draper Decorators: Why?
• Removing presentation code from your model or model-
concerns
• Consolidating some helper, view, controller methods by models
• Presentation code relating to one model, but multiple
controllers/views
• Consolidation of flash messages related to a given model
27
28. Draper Decorators: Why
• Decorators are the ideal place to:
• format complex data for user display
• define commonly-used representations of an object, like a
name method that combines first_name and last_name
attributes
• mark up attributes with a little semantic HTML, like turning a
url field into a hyperlink
28
30. Example
Several views have code that format the micropost.created_at:
!
Posted <%= time_ago_in_words(micropost.created_at) %> ago.
30
31. Scenario
• You have duplicated rendering code in several files.
• Remedy:
1. If rendering code, use a partial.
2. If ruby code, use either a view helper or create a static
method on a utility class. View helpers have access other
helpers. Utility classes require extra work to call view
context methods.
31
32. Objectives
Patterns &
Techniques
32
DRY
Methods
< 5 Lines
Classes
< 100 lines
One Instance
Variable in View
Easy to
Test
Concerns
Draper
Decorators
Validation
Classes
Presenters
Split-up
Controllers
Clarity
Easy to
Change
Guidelines
Move Logic
to Models
Easy to
Find
33. Scenario
• You are setting too many instance variables in the controller
action.You also have local variables being assigned in the view.
• Presenter pattern: Create a PORO that wraps up the values
and logic going from the controller to the view.
33
34. Scenario
• Fragment caching in your view, but some extra queries still run
• Use the Presenter pattern, with memoization in the instance
methods.
• @foobar ||= calculate_foobar
34
35. Presenters
35
Presenter Object
Wrapping Data
Needed by View
Smaller Controller
Action Creating Only
the Presenter Instance
Big Controller
Action Setting
Many Instance
Variables
View with ONE
Instance Variable
View with MANY
Instance Variables
before
after
36. Scenario
• Problem:A controller file is huge with many actions and many more
private methods.
• Solution:
1. Split up the controller into multiple files by having your routing file
map to different controllers.
2. Put any common functionality in a controller concern, similar to
how you would do it for a model.An alternative is having an
inheritance hierarchy of controllers. Mix-ins are more flexible.
36
37. Scenario
• Problem:
• Your Presenter class needs to access the view context, but it’s PORO.
• Solution:
1. Use this include in your PORO: “include Draper::ViewHelpers”.
2. Pass the controller instance into the constructor of the Presenter (include
required helpers in controller), or set the view context in the view file.
3. Pass the view context into the methods that need it on the Presenter.
37
38. Objectives
Patterns &
Techniques
38
DRY
Methods
< 5 Lines
Classes
< 100 lines
One Instance
Variable in View
Easy to
Test
Concerns
Draper
Decorators
Validation
Classes
Presenters
Split-up
Controllers
Clarity
Easy to
Change
Guidelines
Move Logic
to Models
Easy to
Find
39. 39
If a minor posts profane words:
!
1. The post shall not be valid.
2. A counter will track how many times the
minor tried to use profanity.
3. The minor's parents shall be notified.
4. A special flash alert will alert the minor to
profanity usage.
Business Case
40. –David Heinemeier Hansson
“I've yet to see a compelling "make action a
service object" example in the wild. Maybe
they exist somewhere, though. Then again,
maybe unicorns are real too.”
40
https://gist.github.com/dhh/10022098
Service Objects?
41. Service Objects Example
41
Big Micropost
Create Action
on Controller
MicropostCreationService
ControllerResponse
Flash, Flash-now, status code
Tiny Micropost
Create Action on
Controller
https://github.com/justin808/fat-code-refactoring-techniques/pull/6
before
after
42. A Bit Humbling…
DHH: "Sorry to keep shooting the patterns down, but this is
exactly what I mean when I say that most code does not need
patterns, it just needs to be rewritten better."
JG: "I think it's a pattern either way.The pattern you presented is
to use validators rather than a separate object."
DHH: Right, which Rails already has built in, and the code is
easier to follow with less work.
42
43. Single Purpose Controller
• Controller with only one action
• https://github.com/justin808/fat-code-refactoring-techniques/
pull/7
43
Big Micropost
Create Action on
Controller
Micropost Controller
Just for Create
Rest of the
Micropost Controller
44. DHH on Controllers
“It’s [controller] intended to process the incoming request, fetch
the model, and direct the user to a view or another action. If
you’re yanking logic of that nature out of the controller, you’re
making an anemic controller. Shoving this into a
service object is imo the lazy approach that
doesn’t deliver any benefits in terms of
simpler code. It imo is the sweep-it-under-the-rug approach.
44
45. DHH on the work of a Controller
"I’ve yet to see compelling controller code that couldn’t be
slimmed down by simply writing it better, spinning off another
controller, or moving domain logic to the model. Here’s another
example of a code ping pong I did off a convoluted action in
RedMine: https://gist.github.com/dhh/10023987”
45
47. Scenario
• Excessive model logic in complicated controller method.
• Either:
• Move model logic out of controller and into the models,
utilizing Rails features such as validation.
• Create a non-AR based model to handle an interaction
between two models (aka “Service Object”)
47
48. POR (Plain Old Rails)
• Use Rails Models,Validation, and Controller for their proper
jobs
• KISS (Keep It Simple Stupid)
• Don’t Invent Patterns That Don’t Need to be Invented
• Know the why of the Rails way
• Know the Rails way before deviating
48
49. Refactoring Steps
• Move validation code and checks out of controller to model
• Move creation of flash message to decorator
• Move validation code to validation class
49
50. References
• Rails Guides: http://guides.rubyonrails.org/
• Patterns to Refactor Fat ActiveRecord Models: http://
blog.codeclimate.com/blog/2012/10/17/7-ways-to-decompose-
fat-activerecord-models/
• DHH’s Example of 2 Controllers with Concerns: https://
gist.github.com/dhh/10022098
50
51. Thanks!
Special thanks to those that helped review my code samples to this talk: @dhh,
@jeg2, @gylaz, @jodosha, @dreamr, @thatrubylove, @therealadam,
@robzolkos, Thoughtbot’s Learn program forum and Ruby Rogues Parley Forum
51
Rails on Maui HQ, aka Sugar Ranch Maui
52. Thanks!
• More details at my blog:
http://www.railsonmaui.com
• Feel free to contact me
regarding your projects
• justin@railsonmaui.com
• http://airpair.me/railsonmaui
52