Resulting

Resulting is a gem to help with result handling and coordinating validations and saving of (primarily) ActiveRecord objects.

Installation

Add this line to your application's Gemfile:

gem 'resulting'

And then execute:

$ bundle install

Or install it yourself as:

$ gem install resulting

Usage

There is a common pattern in a rails controller doing something like this:

  • Controller action calls a service (or two or three)
  • That service returns an object (or two or three)
  • If you want to see if it was successful, that service may have its own result object, or you can check if the object was persisted.
  • Maybe the service does that and just returns and object for serialization (more likely since we should have skinny controllers). But then you have the same problem.
  • Result objects are hard to manage outside a framework explicitly and ruthlessly designed to use them.

There are of course some very complicated situations, but many situations can be solved using Resulting.

While Resulting can be used in any way you see fit, the way I use it is described below.

Custom Results

First we create a result object specific to our controller action. The reason for this is we don't need a result object that is so flexible it is basically an openstruct, but we want something with a slightly nicer API than a hash.

Additionally, this adds clarity, with a tested class, to the result object. So anyone looking for what this result contains knows to just look for a result named after the controller action.

class CreateUserAndWidgetResult
  include Resulting::Resultable

  def user
    value[:user]
  end

  def widget
    value[:widget]
  end
end

Controller

Now in our controller, since we know the result looks like, we can simply grab the relevant objects out of it and set them to instance variables for a view, or put them into JSON with JBuilder or whatever.

We can also just check to see if our result was successful, the simplest API for a result object.

class NewController
  def create
    result = UserAndWidgetCreateService.call(params)
    @user = result.user
    @widget = result.widget

    if result.success?
      redirect_to :show
    else
      render :new
    end
  end
end

Result Object and Handlers

Finally, in our service, we initialize the result object we first defined with our objects that we need to validate, save, and do whatever on.

Then we call the shortcut .validate_and_save method which will validate all objects, this ensures that all objects will have errors even if the first one fails validation.

If they are all valid, it will call .save (not save!) on each of them, inside an ActiveRecord::Base.transaction. If all .save calls return true. Then we will return a succesful result. If any of them return false, we will bail out early and raise an ActiveRecord::Rollback error inside of the transaction.

class UserAndWidgetCreateService
  def call(params)
    new_result = CreateResult.success({
      user: User.build(params[:user])
      widget: Widget.build(params[:widget])
      role: Role.build(user: user, widget: widget, role: :admin)
    })

    Resulting.validate_and_save(result)
  end
end

Details

  1. Resulting::Runner
    1. .run_all
    2. .run_until_failure
    3. With blocks
    4. Options (:failure_case, :wrapper)
    5. With Rails
  2. Resulting::Handler
  3. Resulting::Result
    1. Constructors (.new, .success, and .failure
    2. .wrap
    3. Methods (#value, #success?, and #failure?
    4. Values
  4. Resulting::Helpers

Resulting::Runner

The Resulting::Runner's will take a result object, and if that result is failing, return immediately. That way you can safely pass the results to any method that takes them without worrying about acting on a failed result. (It's almost like a monad, but definitely not a monad).

.run_all(result, method:)

This will call the given method on every object in result.values. It will keep track whether or not all calls to method on each object were true.

If all calls to method were true, it will return a successful result.

.run_until_failure(result, method:)

This will call the given method on every object in result.values UNTIL it sees a failure. At that point, it will bail out and stop calling the method.

With Blocks

Both of these methods take an optional block:

  • In run_all, the block will be run no matter what. The return value of the block will be &&'d with the current success value of calling method on all the values. That new success value will determine whether the call was successful. ruby Resulting::Runner.run_all(result, method: :validate) do # Validate other things # return true end
  • In run_until_falure, the block will be run no matter what. The return value of the block will be &&'d with the current success value of calling method on all the values. That new success value will determine whether the call was successful. ruby Resulting::Runner.run_until_failure(result, method: :validate) do # Save other things # return true end

*NOTE: The return value of the block is what is used to determine success.* Be mindful of the return value.

Options (:failure_case, :wrapper)

failure_case is an optional argument. It should be a lambda that describes what to do at the end if a failure is encountered. By default it's just a lambda that returns false.

For example, when validating, if all :validate calls have returned false, we just want to return false. However, if we are saving, and one of the saves returns false, we actually want to do raise ActiveRecord::Rollback.

Odds are you will either return false or raise some error, but any lambda will do.

wrapper is something that will wrap the whole result handling process. The common example here would be to wrap all saves in an ActiveRecord::Base.transaction block to ensure we can rollback safely.

With Rails (.validate, .save, and .validate_and_save)

Most of the time this is used within rails, and as described there are some things you will commonly want to do.

Resulting.validate(param)

Is equivalent to:

Resulting::Runner.run_all(param, method: :validate)

Resulting.save(param)

Is equivalent to:

Resulting::Runner.run_until_failure(
  param,
  method: :save,
  failure_case: -> { raise ActiveRecord::Rollback },
  wrapper: -> { ActiveRecord::Base.method(:transaction) },
)

Both of these still take blocks.

Finally, Resulting.validate_and_save will just call one after the other. This one does not take a block, so it assumes you just want to validating everything and then save it.

Resulting::Result

This is a generic result class that implements Resulting::Resultable.

Constructors: (.new, .success, .failure)

  • .new(success, value) stores the value and sets success to the first parameter
  • .success(value) stores the value and sets success to true
  • .failure(value) stores the value and sets success as false

.wrap

Resulting::Result.wrap is worth calling out on its own. Result.wrap(value) will do the following:

  • If value is a result (i.e. implements Resulting::Resultable) it returns the value.
  • If value is anything else, it will return Result.success(value).
$ foo = Object.new
$ result = Resulting::Result.wrap(foo)
$ result
=> #<Resulting::Result:0x00007f91dd072238 @success=true, @value=#<Object:0x00007f91db929950>>
$ Resulting::Result.wrap(result)
=> #<Resulting::Result:0x00007f91dd072238 @success=true, @value=#<Object:0x00007f91db929950>>

You can use wrap to ensure you have a result object if you need it.

Methods: #value, #success?, and #failure?

A result has helper methods, #success? and #failure? which just check whether success is truthy, and the obj is stored as the value.

success = obj.validate # => true
result = Resulting::Result.new(success, obj)

result.success? # => true
result.value # => obj

#values

values returns the value collapsed into an array. This variable is iterated over by the two runner methods.

*NOTE: Resulting assumes any methods calls on the value mutate the value itself and that it is passed by reference.*

  • If value is a Hash, values is value.values.flatten
  • If value is anything else, values is Array(value).flatten
    • (This will wrap objects in an array, and leave arrays alone.)

When building your own result you can override this to provide different behavior. You could use this maintain access to an object but not call a method on it, or to add data you want acted on from a side effect.

class MyResult
  def user
    values[:user]
  end

  def hashed_password
    values[:password] # Omit from values, so it's not acted on
  end

  def values
    [user, user.side_effect_record]
  end
end

In this case, we have a password (or any object in memory we don't/can't persist). It will be on the result object so we can do something with it, but by omitting it from #values, we don't have to worry about it being acted on.

In contrast, let's say in our services we create some record as a side effect which couldn't be created at the time we created the result (this is pretty contrived, but go with it), then we can add that to the values as something to be validated, saved, or whatever when the runners process the result.

Resulting::Helpers

If you include Resulting::Helpers in a given class or module, you get the some nifty helper shortcuts.

Success(value) # Equal to Resulting::Result.success(value)
Failure(value) # Equal to Resulting::Result.failure(value)

Development

After checking out the repo, run bin/setup to install dependencies. Then, run rake spec to run the tests. You can also run bin/console for an interactive prompt that will allow you to experiment.

To install this gem onto your local machine, run bundle exec rake install. To release a new version, update the version number in version.rb, and then run bundle exec rake release, which will create a git tag for the version, push git commits and tags, and push the .gem file to rubygems.org.

Contributing

Bug reports and pull requests are welcome on GitHub at https://github.com/dewyze/resulting.