Monadic
helps dealing with exceptional situations, it comes from the sphere of functional programming and bringing the goodies I have come to love in Scala to my ruby projects (hence I will be using more Scala like constructs than Haskell).
My motivation to create this gem was that I often work with nested Hashes and need to reach deeply inside of them so my code is sprinkled with things like some_hash.fetch(:one, {}).fetch(:two, {}).fetch(:three, "unknown").
We have the following monadics (monads, functors, applicatives and variations):
- Option (Maybe in Haskell) - Scala like with a rubyesque flavour
- Either - more Haskell like
- Validation
What's the point of using monads in ruby? To me it started with having a safe way to deal with nil objects and other exceptions. Thus you contain the erroneous behaviour within a monad - an indivisible, impenetrable unit.
Monad purists might complain that there is no unit method to get the zero monad, I didn't include them, as I didn't find this idiomatic to the ruby language. I prefer to focus on the pragmatic uses of monads. If you want to learn moar about monads, see the references section at the bottom.
A monad is most effectively described as a computation that eventually returns a value. -- Wolfgang De Meuter
Usage
Option
Is an optional type, which helps to handle error conditions gracefully. The one thing to remember about option is: 'What goes into the Option, stays in the Option'.
Option(User.find(123)).name._ # ._ is a shortcut for .fetch
# if you prefer the alias Maybe instead of option
Maybe(User.find(123)).name._
# confidently diving into nested hashes
Maybe({})[:a][:b][:c] == None
Maybe({})[:a][:b][:c].fetch('unknown') == None
Maybe(a: 1)[:a]._ == 1
Basic usage examples:
# handling nil (None serves as NullObject)
obj = nil
Option(obj).a.b.c == None
# None stays None
Option(nil)._ == "None"
"#{Option(nil)}" == "None"
Option(nil)._("unknown") == "unknown"
Option(nil).empty? == true
Option(nil).truly? == false
# Some stays Some, unless you unbox it
Option('FOO').downcase == Some('foo')
Option('FOO').downcase.fetch == "foo"
Option('FOO').downcase._ == "foo"
Option('foo').empty? == false
Option('foo').truly? == true
Map, select:
Option(123).map { |value| User.find(value) } == Option(someUser) # if user found
Option(0).map { |value| User.find(value) } == None # if user not found
Option([1,2]).map { |value| value.to_s } == Option(["1", "2"]) # for all Enumerables
Option('foo').select { |value| value.start_with?('f') } == Some('foo')
Option('bar').select { |value| value.start_with?('f') } == None
Treat it like an array:
Option(123).to_a == [123]
Option([123, 456]).to_a == [123, 456]
Option(nil) == []
Falsey values (kind-of) examples:
user = Option(User.find(123))
user.name._
user.fetch('You are not logged in') { |user| "You are logged in as #{user.name}" }.should == 'You are logged in as foo'
if user != nil
"You are logged in as foo"
else
"You are not logged in"
user.subscribed? # always true
user.subscribed?.truly? # true if subscribed is true
user.subscribed?.fetch(false) # same as above
user.subscribed?.or(false) # same as above
Remember! an Option is never false (in Ruby terms), if you want to know if it is false, call #empty?
of #truly?
#truly?
will return true or false, always.
Slug example
# instead of
def slug(title)
if title
title.strip.downcase.tr_s('^[a-z0-9]', '-')
end
end
# or
def slug(title)
title && title.strip.downcase.tr_s('^[a-z0-9]', '-')
end
# do it with a default
def slug(title)
Option(title).strip.downcase.tr_s('^[a-z0-9]', '-')._('unknown-title')
end
Either
Its main purpose here to handle errors gracefully, by chaining multiple calls in a functional way and stop evaluating them as soon as the first fails.
Assume you need several calls to construct some object in order to be useful, after each you need to check for success. Also you want to catch exceptions and not let them bubble upwards.
What is specific to this implementation is that exceptions are caught within the execution blocks. This way I have all error conditions wrapped in one place.
Success
represents a successfull execution of an operation (Right in Scala, Haskell).
Failure
represents a failure to execute an operation (Left in Scala, Haskell).
The Either()
wrapper will treat all falsey values nil
, false
or empty?
as a Failure
and all others as Success
. If that does not suit you, use Success
or Failure
only.
result = parse_and_validate_params(params). # must return a Success or Failure inside
bind ->(user_id) { User.find(user_id) }. # if #find returns null it will become a Failure
bind ->(user) { (user); user }. # if authorized? raises an Exception, it will be a Failure
bind ->(user) { UserDecorator(user) }
if result.success?
@user = result.fetch # result.fetch or result._ contains the
render 'page'
else
@error = result.fetch
render 'error_page'
end
You can use alternate syntaxes to achieve the same goal:
# block and Haskell like >= operator
Either(operation).
>= { successful_method }.
>= { failful_operation }
# start with a Success, for instance a parameter
Success('pzol').
bind ->(previous) { good }.
bind -> { bad }
Either.chain do
bind -> { good } # >= is not supported for Either.chain, only bind
bind -> { better } # better returns Success(some_int)
bind ->(previous_result) { previous_result + 1 }
end
either = Either(something)
either += truth? Success('truth, only the truth') : Failure('lies, damn lies')
Exceptions are wrapped into a Failure:
Either(true).
bind -> { fail 'get me out of here' } # return a Failure(RuntimeError)
Another example:
Success(params).
bind ->(params) { Either(params.fetch(:path)) } # fails if params does not contain :path
bind ->(path) { load_stuff(params) } #
Storing intermediate results in instance variables is possible, although it is not very elegant:
result = Either.chain do
bind { @map = { one: 1, two: 2 } }
bind { @map.fetch(:one) }
bind { |p| Success(p + 100) }
end
result == Success(101)
Validation
The Validation applicative functor, takes a list of checks within a block. Each check must return either Success of Failure.
If Successful, it will return Success, if not a Failure monad, containing a list of failures.
Within the Failure() provide the reason why the check failed.
Example:
def validate(person)
check_age = ->(age_expr) {
age = age_expr.to_i
case
when age <= 0; Failure('Age must be > 0')
when age > 130; Failure('Age must be < 130')
else Success(age)
end
}
check_sobriety = ->(sobriety) {
case sobriety
when :sober, :tipsy; Success(sobriety)
when :drunk ; Failure('No drunks allowed')
else Failure("Sobriety state '#{sobriety}' is not allowed")
end
}
check_gender = ->(gender) {
gender == :male || gender == :female ? Success(gender) : Failure("Invalid gender #{gender}")
}
Validation() do
check { check_age.(person.age); }
check { check_sobriety.(person.sobriety) }
check { check_gender.(person.gender) }
end
end
References
- Wikipedia Monad
- Learn You a Haskell - for a few monads more
- Monad equivalent in Ruby
- Option Type
- NullObject and Falsiness by @avdi
- andand
- ick
- Monads in Ruby
- The Maybe Monad in Ruby
- Monads in Ruby with nice syntax
- Maybe in Ruby
- Monads on the Cheap
- Rumonade - another excellent (more scala-like) monad implementation
- Monads for functional programming
- Monads as a theoretical foundation for AOP
- What is an applicative functor?
Installation
Add this line to your application's Gemfile:
gem 'monadic'
And then execute:
$ bundle
Or install it yourself as:
$ gem install monadic
Contributing
- Fork it
- Create your feature branch (
git checkout -b my-new-feature
) - Commit your changes (
git commit -am 'Added some feature'
) - Push to the branch (
git push origin my-new-feature
) - Create new Pull Request