ActionLogic
Introduction
This is a business logic abstraction gem that provides structure to the organization and composition of business logic in a Ruby or Rails application. ActionLogic
is inspired by gems like ActiveInteraction, DecentExposure, Interactor, Light-Service, Mutations, Surrounded, Trailblazer and Wisper.
Why another business logic abstraction gem? ActionLogic
provides teams of various experience levels with a minimal yet powerful set of abstractions that promote easy to write and easy to understand code. By using ActionLogic
, teams can more quickly and easily write business logic that honors the SOLID principles, is easy to test and easy to reason about, and provides a flexible foundation from which teams can model and define their application's business domains by focusing on reusable units of work that can be composed and validated with one another.
Contents
- Backstory
- Overview
ActionContext
ActionTask
ActionUseCase
ActionCoordinator
- Succeeding an
ActionContext
- Failing an
ActionContext
- Halting an
ActionContext
- Custom
ActionContext
Status - Error Handling
- Attribute Validations
- Type Validations
- Custom Type Validations
- Presence Validations
- Custom Presence Validations
- Before Validations
- After Validations
- Around Validations
- Benchmarking
- Installation
- Contributing
Backstory
Consider a traditional e-commerce Rails application. Users can shop online and add items to their shopping cart until they are ready to check out. The happy path scenario might go something like this: the user submits their order form, an orders controller action records the order in the database, submits the order total to a payment processor, waits for a response from the payment processor, and upon a success response from the payment processor sends an order confirmation email to the user, the order is sent internally to the warehouse for fulfillment which requires creating various records in the database, and finally the server responds to the initial POST request with a rendered html page including a message indicating the order was successfully processed. In this work flow there are at least 7 distinct steps or tasks that must be satisfied in order for the application's business logic to be considered correct according to specifications.
Although this flow works well for most users, there are other users whose credit card information might be expired or users who might attempt to check out when your application's payment processor service is down. Additional edge case scenarios start to pop up in error logs as exception emails fill up your inbox. What happens when that user that is notorious for having 100 tabs open forgets to complete the checkout process and submits a two week old order form that includes an item that your e-commerce store no longer stocks? What happens if an item is sold out? The edge cases and exception emails pile up, and as each one comes in you add more and more logic to that controller action.
What once was a simple controller action designed with only the happy path of a successful checkout in mind has now become 100 lines long with 5 to 10 levels of nested if statements. The voice of Uncle Bob starts ringing in your ears and you know there must be a better way. You think on it for awhile and consider not only the technical challenges of refactoring this code, but you'd also like to make this code reusable and modular. You want this code to be easy to test and easy to maintain. You want to honor the SOLID principles by writing classes that are singularly focused and easy to extend. You reason these new classes should only have to change if the business logic they execute changes. You see that there are relationships between the entities and you see the possibility of abstractions that allow entities of similar types to interact nicely with each other. You begin thinking about interfaces and the Liskov Substitution Principle, and eventually your mind turns towards domains and data modeling. Where does it end you wonder?
But you remember your team. It's a team of people all wanting to do their best, and represent a variety of backgrounds and experiences. Each person has varying degress of familiarity with different types of abstractions and approaches, and you wonder what abstractions might be as easy to work with for a new developer as they are for an experienced developer? You consider DSL's you've used in the past and wonder what is that ideal balance between magic and straightforward OOP design?
As more and more questions pile up in the empty space of your preferred text editor, you receive another exception email for a new problem with the order flow. The questions about how to refactor this code transform into asking questions about how can you edit the existing code to add the new fix? Add a new nested if statement? You do what you can given the constraints you're faced with, and add another 5 lines and another nested if statement. You realize there is not enough time to make this refactor happen, and you've got to push the fix out as soon as possible. Yet, as you merge your feature branch in master and deploy a hotfix, you think surely there must be a better way.
ActionLogic
was born from many hours thinking about these questions and considering how it might be possible to achieve a generic set of abstractions to help guide
business logic that would promote the SOLID principles and be easy for new and experienced developers to understand and extend. It's not a perfect abstraction (as nothing is),
but can help simplify your application's business logic by encouraging you to consider the smallest units of work required for your business logic while offering features
like type and presence validation that help reduce or eliminate boiler plate, defensive code (nil checks anyone?). However, as with all general purpose libraries, your mileage
will vary.
Overview
There are three levels of abstraction provided by ActionLogic
:
ActionTask
(a concrete unit of work)ActionUseCase
(organizes two or moreActionTasks
)ActionCoordinator
(coordinates two or moreActionUseCases
)
Each level of abstraction operates with a shared, mutable data structure referred to as a context
and is an instance of ActionContext
. This shared context
is threaded
through each ActionTask
, ActionUseCase
and / or ActionCoordinator
until all work is completed. The resulting context
is returned to the original caller
(typically in a Rails application this will be a controller action). In the problem described above we might have an ActionUseCase
for organizing the checkout order flow,
and each of the distinct steps would be represented by a separate ActionTask
. However, overtime it may make more sense to split apart the singular ActionUseCase
for the order
flow into smaller ActionUseCases
that are isolated by their domain (users, payment processor, inventory / warehouse, email, etc.). Considering that we limit our ActionUseCases
to
single domains, then the ActionCoordinator
abstraction would allow us to coordinate communication between the ActionUseCases
and their ActionTasks
to fulfill the necessary
work required when a user submits a checkout order form.
The diagram below illustrates how the ActionTask
, ActionUseCase
and ActionCoordinator
abstractions work together, and the role of ActionContext
as the primary, single input:
ActionContext
The glue that binds the three layers of abstraction provided in ActionLogic
is ActionContext
. Anytime an ActionTask
, ActionUseCase
or ActionCoordinator
is invoked
an instance of ActionContext
is created and passed as an input parameter to the receiving execution context. Because each of the three abstractions works in the same way
with ActionContext
, it is intended to be a relatively simple "learn once understand everywhere" abstraction.
Instances of ActionContext
are always referred to within the body of call
methods defined in any ActionTask
, ActionUseCase
or ActionCoordinator
as context
. An
instance of ActionContext
is a thin wrapper around Ruby's standard library OpenStruct
. This allows
instances of ActionContext
to be maximally flexible. Arbitrary attributes can be defined on a context
and their values can be of any type.
In addition to allowing arbitrary attributes and values to be defined on a context
, instances of ActionContext
also conform to a set of simple rules:
- Every
context
instance is instantiated with a defaultstatus
of:success
- A
context
responds tosuccess?
which returns true if thestatus
is:success
- A
context
responds tofail!
which sets thestatus
to:failure
- A
context
responds tofail?
which returns true if thestatus
is:failure
- A
context
rseponds tohalt!
which sets thestatus
to:halted
- A
context
responds tohalted?
which returns true if thestatus
is:halted
Enough with the words, let's look at some code! The following shows an instance of ActionContext
and its various abilities:
context = ActionLogic::ActionContext.new
context # => #<ActionLogic::ActionContext status=:success>
# default status is `:success`:
context.status # => :success
# defining a new attribute called `name` with the value `"Example"`:
context.name = "Example"
# retrieving the value of the `name` attribute:
context.name # => "Example"
# you can set attributes to anything, including Procs:
context.lambda_example = -> { "here" }
context.lambda_example # => #<Proc:0x007f8d6b0a9ba0@-:11 (lambda)>
context.lambda_example.call # => "here"
# contexts can be failed:
context.fail!
context.status # => :failure
context.failure? # => true
# contexts can also be halted:
context.halt!
context.status # => :halted
context.halted? # => true
Now that we have seen what ActionContext
can do, let's take a look at the lowest level of absraction in ActionLogic
that consumes instances of ActionContext
, the ActionTask
abstraction.
ActionTask
At the core of every ActionLogic
work flow is an ActionTask
. These classes are the lowest level of abstraction in ActionLogic
and are where concrete work is performed. All ActionTasks
conform to the same structure and incorporate all features of ActionLogic
including validations and error handling.
To implement an ActionTask
class you must define a call
method. You can also specify any before, after or around validations or an error handler. The following code example demonstrates an ActionTask
class that includes before and after validations, and also demonstrates how an ActionTask
is invoked :
class ActionTaskExample
include ActionLogic::ActionTask
validates_before :expected_attribute1 => { :type => String },
:expected_attribute2 => { :type => Fixnum, :presence => true }
validates_after :example_attribute1 => { :type => String, :presence => ->(example_attribute1) { !example_attribute1.empty? } }
def call
# adds `example_attribute1` to the shared `context` with the value "Example value"
context.example_attribute1 = "New value from context attributes: #{context.expected_attribute1} #{context.expected_attribute2}"
end
end
# ActionTasks are invoked by calling an `execute` static method directly on the class with an optional hash of key value pairs:
result = ActionTaskExample.execute(:expected_attribute1 => "example", :expected_attribute2 => 123)
# The result object is the shared context object (an instance of ActionContext):
result # => #<ActionLogic::ActionContext expected_attribute1="example", expected_attribute2=123, status=:success, example_attribute1="New value from context attributes: example 123">
The ActionTaskExample
is invoked using the static method execute
which takes an optional hash of attributes that is converted into an ActionContext
.
Assuming the before validations are satisfied, the call
method is invoked. In the body of the call
method the ActionTask
can access the shared ActionContext
instance via a context
object. This shared context
object allows for getting and setting attributes as needed. When the call
method returns, the context
is validated against any defined after validations, and the context
is then returned to the caller.
The diagram below is a visual representation of how an ActionTask
is evaluted when its execute
method is invoked from a caller:
Although this example is for the ActionTask
abstraction, ActionUseCase
and ActionCoordinator
follow the same pattern. The difference is that ActionUseCase
is designed to organize multiple ActionTasks
, and ActionCoordinator
is designed to organize many ActionUseCases
.
ActionUseCase
As business logic grows in complexity the number of steps or tasks required to fulfill that business logic tends to increase. Managing this complexity is a problem every team must face.
Abstractions can help teams of varying experience levels work together and promote code that remains modular and simple to understand and extend. ActionUseCase
represents a layer of
abstraction that organizes multiple ActionTasks
and executes each ActionTask
in the order they are defined. Each task receives the same shared context
so tasks can be composed together.
To implement an ActionUseCase
class you must define a call
method and a tasks
method. You also can specify any before, after or around validations or an error handler.
The following is an example showcasing how an ActionUseCase
class organizes the execution of multiple ActionTasks
and defines before and after validations on the shared context
:
class ActionUseCaseExample
include ActionLogic::ActionUseCase
validates_before :expected_attribute1 => { :type => String },
:expected_attribute2 => { :type => Fixnum, :presence => true }
validates_after :example_task1 => { :type => TrueClass, :presence => true },
:example_task2 => { :type => TrueClass, :presence => true },
:example_task3 => { :type => TrueClass, :presence => true },
:example_usecase1 => { :type => TrueClass, :presence => true }
# The `call` method is invoked prior to invoking any of the ActionTasks defined by the `tasks` method.
# The purpose of the `call` method allows us to prepare the shared `context` prior to invoking the ActionTasks.
def call
context # => #<ActionLogic::ActionContext expected_attribute1="example", expected_attribute2=123, status=:success>
context.example_usecase1 = true
end
def tasks
[ActionTaskExample1,
ActionTaskExample2,
ActionTaskExample3]
end
end
class ActionTaskExample1
include ActionLogic::ActionTask
validates_after :example_task1 => { :type => TrueClass, :presence => true }
def call
context # => #<ActionLogic::ActionContext expected_attribute1="example", expected_attribute2=123, status=:success, example_usecase1=true>
context.example_task1 = true
end
end
class ActionTaskExample2
include ActionLogic::ActionTask
validates_after :example_task2 => { :type => TrueClass, :presence => true }
def call
context # => #<ActionLogic::ActionContext expected_attribute1="example", expected_attribute2=123, status=:success, example_usecase1=true, example_task1=true>
context.example_task2 = true
end
end
class ActionTaskExample3
include ActionLogic::ActionTask
validates_after :example_task3 => { :type => TrueClass, :presence => true }
def call
context # => #<ActionLogic::ActionContext expected_attribute1="example", expected_attribute2=123, status=:success, example_usecase1=true, example_task1=true, example_task2=true>
context.example_task3 = true
end
end
# To invoke the ActionUseCaseExample, we call its execute method with the required attributes:
result = ActionUseCaseExample.execute(:expected_attribute1 => "example", :expected_attribute2 => 123)
result # => #<ActionLogic::ActionContext expected_attribute1="example", expected_attribute2=123, status=:success, example_usecase1=true, example_task1=true, example_task2=true, example_task3=true>
By following the value of the shared context
from the ActionUseCaseExample
to each of the ActionTask
classes, it is possible to see how the shared context
is mutated to accomodate the various attributes and their values each execution context adds to the context
. It also reveals the order in which the ActionTasks
are evaluated, and indicates that the call
method of the ActionUseCaseExample
is invoked prior to any of the ActionTasks
defined in the tasks
method.
To help visualize the flow of execution when an ActionUseCase
is invoked, this diagram aims to illustrate the relationship between ActionUseCase
and ActionTasks
and the order in which operations are performed:
ActionCoordinator
Sometimes the behavior we wish our Ruby or Rails application to provide requires us to coordinate work between various domains of our application's business logic.
The ActionCoordinator
abstraction is intended to help coordinate multiple ActionUseCases
by allowing you to define a plan of which ActionUseCases
to invoke
depending on the outcome of each ActionUseCase
execution. The ActionCoordinator
abstraction is the highest level of abstraction in ActionLogic
.
To implement an ActionCoordinator
class, you must define a call
method in addition to a plan
method. The purpose of the plan
method is to define a state
transition map that links together the various ActionUseCase
classes the ActionCoordinator
is organizing, as well as allowing you to define error or halt
scenarios based on the result of each ActionUseCase
. The following code example demonstrates a simple ActionCoordinator
:
class ActionCoordinatorExample
include ActionLogic::ActionCoordinator
def call
context.required_attribute1 = "required attribute 1"
context.required_attribute2 = "required attribute 2"
end
def plan
{
ActionUseCaseExample1 => { :success => ActionUseCaseExample2,
:failure => ActionUseCaseFailureExample },
ActionUseCaseExample2 => { :success => nil,
:failure => ActionUseCaseFailureExample },
ActionUseCaseFailureExample => { :success => nil }
}
end
end
class ActionUseCaseExample1
include ActionLogic::ActionUseCase
validates_before :required_attribute1 => { :type => String }
def call
context # => #<ActionLogic::ActionContext status=:success, required_attribute1="required attribute 1", required_attribute2="required attribute 2">
context.example_usecase1 = true
end
# Normally `tasks` would define multiple tasks, but in this example, I've used one ActionTask to keep the overall code example smaller
def tasks
[ActionTaskExample1]
end
end
class ActionUseCaseExample2
include ActionLogic::ActionUseCase
validates_before :required_attribute2 => { :type => String }
def call
context # => #<ActionLogic::ActionContext status=:success, required_attribute1="required attribute 1", required_attribute2="required attribute 2", example_usecase1=true, example_task1=true>
context.example_usecase2 = true
end
# Normally `tasks` would define multiple tasks, but in this example, I've used one ActionTask to keep the overall code example smaller
def tasks
[ActionTaskExample2]
end
end
# In this example, we are not calling ActionUseCaseFailureExample, but is used to illustrate the purpose of the `plan` of our ActionCoordinator
# in the event of a failure in one of the consumed `ActionUseCases`
class ActionUseCaseFailureExample
include ActionLogic::ActionUseCase
def call
end
def tasks
[ActionTaskLogFailure,
ActionTaskEmailFailure]
end
end
class ActionTaskExample1
include ActionLogic::ActionTask
validates_after :example_task1 => { :type => TrueClass, :presence => true }
def call
context # => #<ActionLogic::ActionContext status=:success, required_attribute1="required attribute 1", required_attribute2="required attribute 2", example_usecase1=true>
context.example_task1 = true
end
end
class ActionTaskExample2
include ActionLogic::ActionTask
validates_after :example_task2 => { :type => TrueClass, :presence => true }
def call
context # => #<ActionLogic::ActionContext status=:success, required_attribute1="required attribute 1", required_attribute2="required attribute 2", example_usecase1=true, example_task1=true, example_usecase2=true>
context.example_task2 = true
end
end
result = ActionCoordinatorExample.execute
result # => #<ActionLogic::ActionContext status=:success, required_attribute1="required attribute 1", required_attribute2="required attribute 2", example_usecase1=true, example_task1=true, example_usecase2=true, example_task2=true>
Succeeding an ActionContext
By default, the value of the status
attribute of instances of ActionContext
is :success
. Normally this is useful information for the caller of an ActionTask
,
ActionUseCase
or ActionCoordinator
because it informs the caller that the various execution context(s) were successful. In other words, a :success
status
indicates that none of the execution contexts had a failure or halted execution.
Failing an ActionContext
Using context.fail!
does two important things: it immediately stops the execution of any proceeding business logic (prevents any additional ActionTasks
from executing)
and also sets the status of the context
as :failure
. This status is most applicable to the caller or an ActionCoordinator
that might have a plan specifically for a :failure
status of a resulting ActionUseCase
.
The following is a simple example to show how a context
is failed within a call
method:
class ActionTaskExample
include ActionLogic::ActionTask
def call
if failure_condition?
context.fail!
end
end
def failure_condition?
true
end
end
result = ActionTaskExample.execute
result # => #<ActionLogic::ActionContext status=:failure, message="">
When failing a context
it is possible to also specify a message:
class ActionTaskExample
include ActionLogic::ActionTask
def call
if failure_condition?
context.fail! "Something was invalid"
end
end
def failure_condition?
true
end
end
result = ActionTaskExample.execute
result # => #<ActionLogic::ActionContext status=:failure, message="Something was invalid">
result.message # => "Something was invalid"
From the above example we see how it is possible to fail!
a context
while also specifying a clarifying message about the failure condition. Later, we retrieve
that failure message via the message
attribute defined on the returned context
.
Halting an ActionContext
Like, failing a context, Using context.halt!
does two important things: it immediately halts the execution of any proceeding business logic (prevents any additional ActionTasks
from executing) and also sets the status of the context
as :halted
. The caller may use that information to define branching logic or an ActionCoordinator
may use that
information as part of its plan
.
However, unlike failing a context
, halting is designed to indicate that no more processing is required, but otherwise execution was successful.
The following is a simple example to show how a context
is halted within a call
method:
class ActionTaskExample
include ActionLogic::ActionTask
def call
if halt_condition?
context.halt!
end
end
def halt_condition?
true
end
end
result = ActionTaskExample.execute
result # => #<ActionLogic::ActionContext status=:halted, message="">
When failing a context
it is possible to also specify a message:
class ActionTaskExample
include ActionLogic::ActionTask
def call
if halt_condition?
context.halt! "Something required a halt"
end
end
def halt_condition?
true
end
end
result = ActionTaskExample.execute
result # => #<ActionLogic::ActionContext status=:halted, message="Something required a halt">
result.message # => "Something required a halt"
From the above example we see how it is possible to halt!
a context
while also specifying a clarifying message about the halt condition. Later, we retrieve
that halt message via the message
attribute defined on the returned context
.
Custom ActionContext
Status
It is worthwhile to point out that you should not feel limited to only using the three provided statuses of :success
, :failure
or :halted
. It is easy to implement your
own system of statuses if you prefer. For example, consider a system that is used to defining various status codes or disposition codes to indicate the result of some business
logic. Instances of ActionContext
can be leveraged to indicate these disposition codes by using the status
attribute, or by defining custom attributes. You are encouraged
to expirement and play with the flexibility provided to you by ActionContext
in determining what is optimal for your given code contexts and your team.
class RailsControllerExample < ApplicationController
def create
case create_use_case.status
when :disposition_1 then ActionUseCaseSuccess1.execute(create_use_case)
when :disposition_2 then ActionUseCaseSuccess2.execute(create_use_case)
when :disposition_9 then ActionUseCaseFailure.execute(create_use_case)
else
ActionUseCaseDefault.execute(create_use_case)
end
end
private
def create_use_case
@create_use_case ||= ActionUseCaseExample.execute(params)
end
end
Although this contrived example would be ideal for an ActionCoordinator
(because the result of ActionUseCaseExample
drives the execution of the next ActionUseCase
), this
example serves to show that status
can be used with custom disposition codes to drive branching behavior.
Error Handling
During execution of an ActionTask
, ActionUseCase
or ActionCoordinator
you may wish to define custom behavior for handling errors. Within any of these classes
you can define an error
method that receives as its input the error exception. Invoking an error
method does not make any assumptions about the status
of the
underlying context
. Execution of the ActionTask
, ActionUseCase
or ActionCoordinator
also stops after the error
method returns, and execution of the work
flow continues as normal unless the context
is failed or halted.
The following example is a simple illustration of how an error
method is invoked for an ActionTask
:
class ActionTaskExample
include ActionLogic::ActionTask
def call
context.before_raise = true
raise "Something broke"
context.after_raise = true
end
def error(e)
context.error = "the error is passed in as an input parameter: #{e.class}"
end
end
result = ActionTaskExample.execute
# the status of the context is not mutated
result.status # => :success
result.error # => "the error is passed in as an input parameter: RuntimeError"
result.before_raise # => true
result.after_raise # => nil
It is important to note that defining an error
method is not required. If at any point in the execution of an ActionTask
, ActionUseCase
or ActionCoordinator
an uncaught exception is thrown and an error
method is not defined, the exception is raised to the caller.
Attribute Validations
The most simple and basic type of validation offered by ActionLogic
is attribute validation. To require that an attribute be defined on an instance of ActionContext
, you
need only specify the name of the attribute and an empty hash with one of the three validation types (before, after or around):
class ActionTaskExample
include ActionLogic::ActionTask
validates_before :required_attribute1 => {}
def call
end
end
result = ActionTaskExample.execute(:required_attribute1 => true)
result.status # => :success
result.required_attribute1 # => true
However, in the above example, if we were to invoke the ActionTaskExample
without the required_attribute1
parameter, the before validation would fail and raise
an ActionLogic::MissingAttributeError
and also detail which attribute is missing:
class ActionTaskExample
include ActionLogic::ActionTask
validates_before :required_attribute1 => {}
def call
end
end
ActionTaskExample.execute # ~> context: ActionTaskExample message: [:required_attribute1] (ActionLogic::MissingAttributeError)
Attribute validations are defined in the same way regardless of the timing of the validation (before, after or around). Please refer to the relevant sections for examples of their usage.
Type Validations
In addition to attribute validations, ActionLogic
also allows you to validate against the type of the value of the attribute you expect to be defined in an instance
of ActionContext
. To understand the default types ActionLogic
validates against, please see the following example:
class ActionTaskExample
include ActionLogic::ActionTask
validates_after :integer_test => { :type => Fixnum },
:float_test => { :type => Float },
:string_test => { :type => String },
:truthy_test => { :type => TrueClass },
:falsey_test => { :type => FalseClass },
:hash_test => { :type => Hash },
:array_test => { :type => Array },
:symbol_test => { :type => Symbol },
:nil_test => { :type => NilClass }
def call
context.integer_test = 123
context.float_test = 1.0
context.string_test = "test"
context.truthy_test = true
context.falsey_test = false
context.hash_test = {}
context.array_test = []
context.symbol_test = :symbol
context.nil_test = nil
end
end
result = ActionTaskExample.execute
result # => #<ActionLogic::ActionContext status=:success,
# integer_test=123,
# float_test=1.0,
# string_test="test",
# truthy_test=true,
# falsey_test=false,
# hash_test={},
# array_test=[],
# symbol_test=:symbol,
# nil_test=nil>
It's important to point out that Ruby's true
and false
are not Boolean
but TrueClass
and FalseClass
respectively. Additionally, nil
's type is NilClass
in Ruby.
Also potentially surprising to some is that Ruby's integer type is of class Fixnum
, but floats are of class Float
.
As we saw with attribute validations, if an attribute's value does not conform to the type expected, ActionLogic
will raise an ActionLogic::AttributeTypeError
with a detailed description about which attribute's value failed the validation:
class ActionTaskExample
include ActionLogic::ActionTask
validates_after :integer_test => { :type => Fixnum }
def call
context.integer_test = 1.0
end
end
ActionTaskExample.execute # ~> context: ActionTaskExample message: Attribute: integer_test with value: 1.0 was expected to be of type Fixnum but is Float (ActionLogic::AttributeTypeError)
In addition to the above default types it is possible to also validate against user defined types.
Custom Type Validations
If you would like to validate the type of attributes on a given context
with your application's classes, ActionLogic
is happy to provide that functionality.
Let's consider the following example:
class ExampleClass
end
class ActionTaskExample
include ActionLogic::ActionTask
validates_after :example_attribute => { :type => ExampleClass }
def call
context.example_attribute = ExampleClass.new
end
end
result = ActionTaskExample.execute
result # => #<ActionLogic::ActionContext status=:success, example_attribute=#<ExampleClass:0x007f95d1922bd8>>
In the above example, a custom class ExampleClass
is defined. In order to type validate against this class, the required format for the name of the class is simply
the class constant ExampleClass
.
If a custom type validation fails, ActionLogic
provides the same ActionLogic::AttributeTypeError
with a detailed explanation about what attribute is in violation
of the type validation:
class ExampleClass
end
class OtherClass
end
class ActionTaskExample
include ActionLogic::ActionTask
validates_after :example_attribute => { :type => ExampleClass }
def call
context.example_attribute = OtherClass.new
end
end
ActionTaskExample.execute # ~> context: ActionTaskExample message: Attribute: example_attribute with value: #<OtherClass:0x007fb5ca04edb8> was expected to be of type ExampleClass but is OtherClass (ActionLogic::AttributeTypeError)
Attribute and type validations are very helpful, but in some situations this is not enough. Additionally, ActionLogic
provides presence validation so you can also verify that
a given attribute on a context not only has the correct type, but also has a value that is considered present
.
Presence Validations
ActionLogic
also allows for presence validation for any attribute on an instance of ActionContext
. Like other validations, presence validations can be specified in before, after or
around validations.
By default, presence validations simply check to determine if an attribute's value is not nil
or is not false
. To define a presence validation, you need only specify :presence => true
for the attribute you wish to validate against:
class ActionTaskExample
include ActionLogic::ActionTask
validates_before :example_attribute => { :presence => true }
def call
end
end
result = ActionTaskExample.execute(:example_attribute => 123)
result # => #<ActionLogic::ActionContext example_attribute=123, status=:success>
However, if a presence validation fails, ActionLogic
will raise an ActionLogic::PresenceError
with a detailed description about the attribute failing the presence validation
and why:
class ActionTaskExample
include ActionLogic::ActionTask
validates_before :example_attribute => { :presence => true }
def call
end
end
ActionTaskExample.execute(:example_attribute => nil) # ~> context: ActionTaskExample message: Attribute: example_attribute is missing value in context but presence validation was specified (ActionLogic::PresenceError)
Custom Presence Validations
Sometimes when wanting to validate presence of an attribute with an aggregate type (like Array
or Hash
), we may want to validate that such a type is not empty. If
you wish to validate presence for a type that requires inspecting the value of the attribute, ActionLogic
allows you the ability to define a custom Proc
to validate
an attribute's value against.
class ActionTaskExample
include ActionLogic::ActionTask
validates_before :example_attribute => { :presence => ->(attribute) { attribute.any? } }
def call
end
end
result = ActionTaskExample.execute(:example_attribute => ["element1", "element2", "element3"])
result # => #<ActionLogic::ActionContext example_attribute=["element1", "element2", "element3"], status=:success>
In the example above, we define a lambda that accepts as input the value of the attribute on the context
. In this case, we are interested in verifying that
example_attribute
is not an empty Array
or an empty Hash
. This passes our before validation because ActionTaskExample
is invoked with an example_attribute
whose value is an array consisting of three elements.
However, if a custom presence validation fails, ActionLogic
will raise an ActionLogic::PresenceError
with a detailed description about the attribute failing
the custom presence validation:
class ActionTaskExample
include ActionLogic::ActionTask
validates_before :example_attribute => { :presence => ->(attribute) { attribute.any? } }
def call
end
end
ActionTaskExample.execute(:example_attribute => []) # ~> context: ActionTaskExample message: Attribute: example_attribute is missing value in context but custom presence validation was specified (ActionLogic::PresenceError)
In the above example, we have failed to pass the presence validation for example_attribute
because the value of example_attribute
is an empty array. When
the custom presence validation lambda is called, the lambda returns false
and the ActionLogic::PresenceError
is thrown, with an error message indicating
the attribute that failed the presence validation while also indicating that a custom presence validation was specified.
Before Validations
If you combine Rails ActionController's before_filter
and ActiveModel's validates
then you have approximately what ActionLogic
defines as validates_before
.
Before validations can be defined in any of the ActionLogic
abstractions (ActionTask
, ActionUseCase
and ActionCoordinator
). In each abstraction a validates_before
operation is performed before invoking the call
method.
Before validations allow you to specify a required attribute and optionally its type and / or presence. The following example illustrates how to specify a before validation on a single attribute:
class ActionTaskExample
include ActionLogic::ActionTask
validates_before :example_attribute => { :type => Array, :presence => ->(attribute) { attribute.any? } }
def call
end
end
result = ActionTaskExample.execute(:example_attribute => [1, 2, 3])
result # => #<ActionLogic::ActionContext example_attribute=[1, 2, 3], status=:success>
The following example illustrates how to specify a before validation for multiple attributes:
class ActionTaskExample
include ActionLogic::ActionTask
validates_before :example_attribute => { :type => Array, :presence => ->(attribute) { attribute.any? } },
:example_attribute2 => { :type => Fixnum }
def call
end
end
result = ActionTaskExample.execute(:example_attribute => [1, 2, 3], :example_attribute2 => 1)
result # => #<ActionLogic::ActionContext example_attribute=[1, 2, 3], example_attribute2=1, status=:success>
After Validations
If you combine Rails ActionController's after_filter
and ActiveModel's validates
then you have approximately what ActionLogic
defines as validates_after
.
After validations can be defined in any of the ActionLogic
abstractions (ActionTask
, ActionUseCase
and ActionCoordinator
). In each abstraction a validates_after
operation is performed after invoking the call
method.
After validations allow you to specify a required attribute and optionally its type and / or presence. The following example illustrates how to specify an after validation on a single attribute:
class ActionTaskExample
include ActionLogic::ActionTask
validates_after :example_attribute => { :type => Array, :presence => ->(attribute) { attribute.any? } }
def call
context.example_attribute = [1, 2, 3]
end
end
result = ActionTaskExample.execute
result # => #<ActionLogic::ActionContext example_attribute=[1, 2, 3], status=:success>
The following example illustrates how to specify an after validation for multiple attributes:
class ActionTaskExample
include ActionLogic::ActionTask
validates_after :example_attribute => { :type => Array, :presence => ->(attribute) { attribute.any? } },
:example_attribute2 => { :type => Fixnum }
def call
context.example_attribute = [1, 2, 3]
context.example_attribute2 = 1
end
end
result = ActionTaskExample.execute
result # => #<ActionLogic::ActionContext example_attribute=[1, 2, 3], example_attribute2=1, status=:success>
Around Validations
If you combine Rails ActionController's around_filter
and ActiveModel's validates
then you have approximately what ActionLogic
defines as validates_around
.
Around validations can be defined in any of the ActionLogic
abstractions (ActionTask
, ActionUseCase
and ActionCoordinator
). In each abstraction a validates_around
operation is performed before and after invoking the call
method.
Around validations allow you to specify a required attribute and optionally its type and / or presence. The following example illustrates how to specify an around validation on a single attribute:
class ActionTaskExample
include ActionLogic::ActionTask
validates_around :example_attribute => { :type => Array, :presence => ->(attribute) { attribute.any? } }
def call
end
end
result = ActionTaskExample.execute(:example_attribute => [1, 2, 3])
result # => #<ActionLogic::ActionContext example_attribute=[1, 2, 3], status=:success>
The following example illustrates how to specify an around validation for multiple attributes:
class ActionTaskExample
include ActionLogic::ActionTask
validates_around :example_attribute => { :type => Array, :presence => ->(attribute) { attribute.any? } },
:example_attribute2 => { :type => Fixnum }
def call
end
end
result = ActionTaskExample.execute(:example_attribute => [1, 2, 3], :example_attribute2 => 1)
result # => #<ActionLogic::ActionContext example_attribute=[1, 2, 3], example_attribute2=1, status=:success>
Benchmarking
At some point you may want to benchmark and profile the performance of your code. ActionLogic
allows for benchmarking that
range from simple defaults to highly customizable options depending on your use case and needs.
Enable Benchmarking
Because benchmarking negatively impacts performance, we must explicitly tell ActionLogic
that we want to benchmark (otherwise
it defaults to ignore benchmarking). To do this, we configure ActionLogic
using the configure
method. With the provided
config
object, we explicitly enable benchmarking by setting config.benchmark = true
:
ActionLogic.configure do |config|
config.benchmark = true
end
Benchmark Logging
Additionally, ActionLogic
writes a benchmark log to $stdout
by default, or you can override this default configuration
by specifying a log file. To do this, you configure ActionLogic
to use a File
object for logging benchmark results via the
ActionLogic.configure
method:
ActionLogic.configure do |config|
config.benchmark = true
config.benchmark_log = File.open("benchmark.log", "w")
end
Benchmark Log Formatting
By default, ActionLogic
formats benchmark logs in the following format:
context:ValidateAroundPresenceTestUseCase user_time:0.000000 system_time:0.000000 total_time:0.000000 real_time:0.000135
...
The default format is intended to be machine readable for easy parsing and is not intended to be used for human reading.
However, if you wish to change the format of the log output, ActionLogic
allows you to override the default formatter by
allowing you to provide your own formatter:
ActionLogic.configure do |config|
config.benchmark = true
config.benchmark_log = File.open("benchmark.log", "w")
config.benchmark_formatter = YourCustomFormatter
end
Where YourCustomFormatter
subclasses ActionLogic::ActionBenchmark::DefaultFormatter
:
class CustomFormatter < ActionLogic::ActionBenchmark::DefaultFormatter
def log_coordinator(benchmark_result, execution_context_name)
benchmark_log.puts("The ActionCoordinator #{execution_context_name} took #{benchmark_result.real} to complete.")
end
def log_use_case(benchmark_result, execution_context_name)
benchmark_log.puts("The ActionUseCase #{execution_context_name} took #{benchmark_result.real} to complete.")
end
def log_task(benchmark_result, execution_context_name)
benchmark_log.puts("The ActionTask #{execution_context_name} took #{benchmark_result.real} to complete.")
end
end
From the example above, you can see that a custom formatter is required to define three methods: log_coordinator
, log_use_case
and log_task
. The log_t cqcoordinator
method is called when a ActionCoordinator
context is benchmarked. The use_case
and task
methods are invoked when ActionUseCase
and ActionTask
contexts are benchmarked, respectively.
Each of the three log methods receives two input parameters: benchmark_result
and execution_context_name
where benchmark_result
is a Ruby
standard library Benchmark
result object, and execution_context_name
is the class name of the ActionLogic
context.
Once configured, you can verify that the formatter outputs to the specified log file by executing your ActionLogic
contexts
and verifying that the log file is written to with the correct format:
The ActionUseCase TestUseCase2 took 0.00011722202179953456 to complete.
The ActionTask TestTask3 took 4.570698365569115e-05 to complete.
...
Custom Benchmark Handling
By default, ActionLogic
benchmarks execution contexts using Ruby's Benchmark
module. If you are content with a Benchmark
result object, then
you do not need to specify a custom benchmark handler. However, if you wish to have maximum control, or you require something different than Ruby's
Benchmark
module, you can define a custom handler like so:
class CustomHandler
def call
# custom logic
yield
# custom logic
end
end
Your custom handler is free to define any custom logic, but you must yield during the body of the call
method. This is what triggers the execution
context and will allow your custom handler to measure the length of execution. If you do not yield, the relevant ActionCoordinator
, ActionUseCase
or ActionTask
will not be executed and will result in no execution to benchmark.
Additionally, you must register your custom handler with ActionLogic
using ActionLogic.configure
:
ActionLogic.configure do |config|
config.benchmark = true
config.benchmark_log = File.open("benchmark.log", "w")
config.benchmark_handler = CustomHandler.new
end
Installation
Add ActionLogic
to your project's Gemfile:
gem 'action_logic'
Don't forget to bundle:
$ bundle
Contributing
Interested in contributing to ActionLogic
? If so that is awesome! <3
Please see the contributing doc for details.