An open API service indexing awesome lists of open source software.

https://github.com/drexed/lite-command

Ruby Command based framework (aka service objects)
https://github.com/drexed/lite-command

command-pattern ruby service-object

Last synced: 8 months ago
JSON representation

Ruby Command based framework (aka service objects)

Awesome Lists containing this project

README

          

# Lite::Command

[![Gem Version](https://badge.fury.io/rb/lite-command.svg)](http://badge.fury.io/rb/lite-command)

Lite::Command provides an API for building simple and complex command based service objects.

## Installation

Add this line to your application's Gemfile:

```ruby
gem 'lite-command'
```

And then execute:

$ bundle

Or install it yourself as:

$ gem install lite-command

## Table of Contents

* [Configuration](#configuration)
* [Usage](#usage)
* [Execution](#execution)
* [Dynamic Faults](#dynamic-faults)
* [Raising Faults](#raising-faults)
* [Context](#context)
* [Attributes](#attributes)
* [Validations](#validations)
* [States](#states)
* [Statuses](#statuses)
* [Hooks](#hooks)
* [Lifecycle Hooks](#lifecycle-hooks)
* [Status Hooks](#status-hooks)
* [State Hooks](#state-hooks)
* [Children](#children)
* [Throwing Faults](#throwing-faults)
* [Sequences](#sequences)
* [Results](#results)
* [Examples](#examples)
* [Disable Instance Calls](#disable-instance-calls)
* [Generator](#generator)

## Configuration

`rails g lite:command:install` will generate the following file in your application root:
`config/initalizers/lite_command.rb`

```ruby
Lite::Command.configure do |config|
config.raise_dynamic_faults = true
end
```

## Usage

Defining a command is as simple as inheriting the base class and adding a `call` method
to a command object (required).

```ruby
class DecryptSecretMessage < Lite::Command::Base

def call
if invalid_magic_numbers?
invalid!("Invalid crypto message")
else
context.decrypted_message = SecretMessage.decrypt(context.encrypted_message)
end
end

private

def invalid_magic_numbers?
# Some logic...
end

end
```

> [!TIP]
> You should treat all command as emphemeral objects, so you should think about making
> all of your domain logic private and leaving the default command API is exposed.

## Execution

Executing a command can be done as an instance or class call. It returns the command instance
in a frozen state. These will never call will never raise an execption, but will be kept track
of in its internal state.

```ruby
DecryptSecretMessage.call(...)
# - or -
DecryptSecretMessage.new(...).call

# On success, fault or exception:
#=>
```

> [!TIP]
> Class calls is the prefered format due to its readability. Read the [Disable Instance Calls](#disable-instance-calls)
> section on how to prevent instance style calls.

Commands can be called with a `!` bang method to raise a `Lite::Command::Fault` or the
original `StandardError` based exceptions.

```ruby
DecryptSecretMessage.call!(...)
# - or -
DecryptSecretMessage.new(...).call!

# On success:
#=>

# On fault:
#=> raises Lite::Command::Fault

# On exception:
#=> raises StandardError
```

### Raising Faults

Sometimes its suitable to raise the offending soft call command fault later
in a call stack. Use the `raise!` method to reraise the fault or original
error (if they differ). `original: false` is the default.

```ruby
cmd = DecryptSecretMessage.call(...)
Apm.track_stat("DecryptSecretMessage.called")
# other stuff...

# On success:
cmd.raise! #=> nil

# On fault:
cmd.raise!(original: false) #=> raises Lite::Command::Fault
cmd.raise!(original: true) #=> raises Lite::Command::Fault

# On exception:
cmd.raise!(original: false) #=> raises Lite::Command::Error
cmd.raise!(original: true) #=> raises StandardError

# Access the exception objects directly
cmd.original_exception #=>
cmd.command_exception #=>
```

### Dynamic Faults

Dynamic faults are custom faults named after your command. This is especially
helpful for catching + running custom logic or filtering out specific
exceptions from your APM service.

```ruby
class DecryptSecretMessage < Lite::Command::Base

def call
fail!("Some failure")
end

private

# Disable raising dynamic faults on a per command basis.
# The `raise_dynamic_faults` configuration option must be
# enabled for this method to have any affect.
def raise_dynamic_faults?
false
end

end

DecryptSecretMessage.call!(...)
#=> raises DecryptSecretMessage::Failure
```

## Context

Accessing the call arguments can be done through its internal context.
It can be used as internal storage to be accessed by it self and any
of its children commands.

> [!NOTE]
> Attributes that do **NOT** exist on the context will return `nil`.

```ruby
class DecryptSecretMessage < Lite::Command::Base

def call
# `ctx` is an alias to `context`
context.decrypted_message = SecretMessage.decrypt(ctx.encrypted_message)
end

end

cmd = DecryptSecretMessage.call(encrypted_message: "a22j3nkenjk2ne2")
cmd.context.decrypted_message #=> "Hello World"
cmd.ctx.fake_message #=> nil
```

### Attributes

Delegate methods for a cleaner command setup by declaring `required` and
`optional` arguments. `required` only verifies that argument was pass to the
context or can be called via defined method or another delegated method.
Is an `:if` or `:unless` callable option on a `required` delegation evaluates
to false, it will be delegated as an `optional` attribute.

```ruby
class DecryptSecretMessage < Lite::Command::Base

required :user, :encrypted_message
required :secret_key, from: :user
required :algo, :algo_detector, if: :signed_in?
optional :version

def call
context.decrypted_message = SecretMessage.decrypt(
encrypted_message,
decryption_key: ENV["DECRYPT_KEY"],
algo: algo,
version: version || 2
)
end

private

def algo_detector
@algo_detector ||= AlgoDetector.new(encrypted_message)
end

def signed_in?
ctx.user.signed_in?
end

end

# With valid options:
cmd = DecryptSecretMessage.call(user: user, encrypted_message: "ll23k2j3kcms", version: 9)
cmd.status #=> "success"
cmd.context.decrypted_message #=> "Hola Mundo"

# With invalid options:
cmd = DecryptSecretMessage.call
cmd.status #=> "invalid"
cmd.reason #=> "Encrypted message is a required argument. User is an undefined argument..."
cmd.metadata #=> {
#=> user: ["is a required argument", "is an undefined argument"],
#=> encrypted_message: ["is a required argument"]
#=> }
```

### Validations

The full power of active model valdations is available to validate
any and all delegated arguments.

```ruby
class DecryptSecretMessage < Lite::Command::Base

required :encrypted_message
optional :version

validates :encrypted_message, length: 10..999
validates :version, inclusion: { in: %w[v1 v3 v8], allow_blank: true }
validate :validate_decrypt_magic_numbers

def call
context.decrypted_message = SecretMessage.decrypt(encrypted_message)
end

private

def validate_decrypt_magic_numbers
return if encrypted_message.starts_with?("~x01~")

errors.add(:encrypted_message, :invalid, message: "has invalid magic numbers")
end

end

# With valid options:
cmd = DecryptSecretMessage.call(encrypted_message: "ll23k2j3kcms", version: "v1")
cmd.status #=> "success"
cmd.context.decrypted_message #=> "Hola Mundo"

# With invalid options:
cmd = DecryptSecretMessage.call(encrypted_message: "idk", version: "v23")
cmd.status #=> "invalid"
cmd.reason #=> "Encrypted message is too short (minimum is 10 character). Encrypted message has invalid magic numbers. Version is not included in list."
cmd.metadata #=> {
#=> user: ["is not included in list"],
#=> encrypted_message: ["is too short (minimum is 10 character)", "has invalid magic numbers"]
#=> }
```

## States
`state` represents the condition of all the code command should execute.

| Status | Description |
| ------------- | ----------- |
| `pending` | Command objects that have been initialized. |
| `executing` | Command objects that are actively executing code. |
| `complete` | Command objects that executed to completion without fault/exception. |
| `interrupted` | Command objects that could **NOT** be executed to completion due to a fault/exception. |

> [!CAUTION]
> States are automatically transitioned and should **NEVER** be altered manually.

```ruby
cmd = DecryptSecretMessage.call
cmd.state #=> "complete"

cmd.pending? #=> false
cmd.executing? #=> false
cmd.complete? #=> true
cmd.interrupted? #=> false

# `complete` or `interrupted`
cmd.executed?
```

## Statuses

`status` represents the state of the domain logic executed via the `call` method.
A status of `success` is returned even if the command has **NOT** been executed.

| Status | Description |
| --------- | ----------- |
| `success` | Call execution completed without fault/exception. |
| `noop` | **Fault** to skip completion of call execution early for an unsatisfied condition where proceeding is pointless. |
| `invalid` | **Fault** to stop call execution due to missing, bad, or corrupt data. |
| `failure` | **Fault** to stop call execution due to an unsatisfied condition where it blocks proceeding any further. |
| `error` | **Fault** to stop call execution due to a thrown `StandardError` based exception. |

> [!IMPORTANT]
> Each **fault** status has a setter method ending in `!` that invokes a matching fault procedure.
> Metadata may also be passed to enrich your fault response.

```ruby
class DecryptSecretMessage < Lite::Command::Base

def call
if context.encrypted_message.empty?
noop!("No message to decrypt")
elsif context.encrypted_message.start_with?("== womp")
invalid!("Invalid message start value", metadata: { i18n: "gb.invalid_start_value" })
elsif context.encrypted_message.algo?(OldAlgo)
failure!("Unsafe encryption algo detected")
else
context.decrypted_message = SecretMessage.decrypt(ctx.encrypted_message)
end
rescue CryptoError => e
Apm.report_error(e)
error!("Failed decryption due to: #{e}", original_exception: e)
end

end

cmd = DecryptSecretMessage.call(encrypted_message: "2jk3hjeh2hj2jh")
cmd.status #=> "invalid"
cmd.reason #=> "Invalid message start value"
cmd.metadata #=> { i18n: "gb.invalid_start_value" }

cmd.success? #=> false
cmd.noop? #=> false
cmd.invalid? #=> true
cmd.invalid?("Other reason") #=> false
cmd.failure? #=> false
cmd.error? #=> false

# `success` or `noop`
cmd.ok? #=> false
cmd.ok?("Other reason") #=> false

# NOT `success`
cmd.fault? #=> true
cmd.fault?("Other reason") #=> false

# `invalid` or `failure` or `error`
cmd.bad? #=> true
cmd.bad?("Other reason") #=> false
```

## Hooks

Use hooks to run arbituary code at transition points and on finalized internals.
All hooks are ran in the order they are defined. Hooks types can be defined
multiple times. Hooks are ran in the following order:

```ruby
1. after_initialize
2. before_execution
3. before_validation
4. after_validation
5. on_executing
6. on_[success, noop, invalid, failure, error]
7. on_[complete, interrupted]
8. after_execution
```

### Lifecycle Hooks

Define before and after callbacks to call around execution.

```ruby
class DecryptSecretMessage < Lite::Command::Base

after_initialize :some_method
before_validation :some_method
after_validation :some_method
before_execution :some_method
after_execution :some_method

def call
# ...
end

end
```

### Status Hooks

Define one or more callbacks that are called after execution for
specific statuses.

```ruby
class DecryptSecretMessage < Lite::Command::Base

on_success :some_method
on_noop :some_method
on_invalid :some_method
on_failure :some_method
on_error :some_method

def call
# ...
end

end
```

### State Hooks

Define one or more callbacks that are called during transitions between states.

```ruby
class DecryptSecretMessage < Lite::Command::Base

on_pending :some_method
on_executing :some_method
on_complete :some_method
on_interrupted :some_method

def call
# ...
end

end
```

## Children

When building complex commands, its best that you pass the parents context to the
child command (unless neccessary) so that it gains automated indexing and the
parents `cmd_id`.

```ruby
class DecryptSecretMessage < Lite::Command::Base

def call
context.merge!(decryption_key: ENV["DECRYPT_KEY"])
ValidateSecretMessage.call(context)
end

end
```

### Throwing Faults

Throwing faults allows you to bubble up child faults up to the parent. Use it to create
branches within your logic and create clean tracing of your command results. You can use
`throw!` as a catch-all or any of the bang status method `failure!`. Any `reason` and
`metadata` will be bubbled up from the original fault.

```ruby
class DecryptSecretMessage < Lite::Command::Base

def call
context.merge!(decryption_key: ENV["DECRYPT_KEY"])
cmd = ValidateSecretMessage.call(context)

if cmd.invalid?("Invalid magic numbers")
failure!(cmd) # Manually throw a specific fault
elsif command.fault?
throw!(cmd) # Automatically throws a matching fault
else
context.decrypted_message = SecretMessage.decrypt(ctx.encrypted_message)
end
end

end
```

## Sequences

A sequence is a command that calls commands in a linear fashion.
This is useful for composing multiple steps into one call.

> [!NOTE]
> Sequences only stop processing on `invalid`, `failure`, and `error`
> faults. This is due to the the idea the `noop` performs no work,
> so its no different than just passing the context forward. To change
> this behavior, just override the `ok?` method with you logic, eg: just `success`

> [!WARNING]
> Do **NOT** define a call method in this class. The sequence logic is
> automatically defined by the sequence class.

```ruby
class ProcessCheckout < Lite::Command::Sequence

required :user

step FinalizeInvoice
step ChargeCard, if: :card_available?
step SendConfirmationEmail, SendConfirmationText
step NotifyWarehouse, unless: proc { ctx.invoice.fullfilled_by_amazon? }

# Do NOT define a call method.

private

def card_available?
user.has_card?
end

end

seq = ProcessCheckout.call(...)
#
```

## Results

During any point in the lifecyle of a command, `to_hash` can be called to dump out
the current values. The `index` value is auto-incremented and the `cmd_id` is static
when its passed to child commands. This helps with debugging and logging.

```ruby
command = DecryptSecretMessage.call(...)
command.to_hash #=> {
#=> index: 1,
#=> cmd_id: "018c2b95-b764-7615-a924-cc5b910ed1e5",
#=> command: "FailureCommand",
#=> outcome: "failure",
#=> state: "interrupted",
#=> status: "failure",
#=> reason: "Command stopped due to some failure",
#=> metadata: {
#=> errors: { name: ["is too short"] },
#=> i18n_key: "command.failure"
#=> },
#=> caused_by: 3,
#=> caused_exception: "[ChildCommand::Failure] something is wrong from within",
#=> thrown_by: 2,
#=> thrown_exception: "[FailureCommand::Failure] something is wrong from within",
#=> runtime: 0.0123
#=> }
```

## Examples

### Disable Instance Calls

```ruby
class DecryptSecretMessage < Lite::Command::Base

private_class_method :new

def call
# ...
end

end

DecryptSecretMessage.new(...).call
#=> raise NoMethodError
```

## Generator

`rails g command NAME` will generate the following file:

```erb
app/commands/[NAME]_command.rb
```

If a `ApplicationCommand` file in the `app/commands` directory is available, the
generator will create file that inherit from `ApplicationCommand` if not it will
fallback to `Lite::Command::Base`.

## 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](https://rubygems.org).

## Contributing

Bug reports and pull requests are welcome on GitHub at https://github.com/[USERNAME]/lite-command. This project is intended to be a safe, welcoming space for collaboration, and contributors are expected to adhere to the [Contributor Covenant](http://contributor-covenant.org) code of conduct.

## License

The gem is available as open source under the terms of the [MIT License](https://opensource.org/licenses/MIT).

## Code of Conduct

Everyone interacting in the Lite::Command project’s codebases, issue trackers, chat rooms and mailing lists is expected to follow the [code of conduct](https://github.com/[USERNAME]/lite-command/blob/master/CODE_OF_CONDUCT.md).