GitLab utilities

We have developed a number of utilities to help ease development:

`MergeHash`

Refer to merge_hash.rb:

Deep merges an array of hashes:

Gitlab::Utils::MergeHash.merge(
  [{ hello: ["world"] },
   { hello: "Everyone" },
   { hello: { greetings: ['Bonjour', 'Hello', 'Hallo', 'Dzien dobry'] } },
    "Goodbye", "Hallo"]
)

Gives:

[
  {
    hello:
      [
        "world",
        "Everyone",
        { greetings: ['Bonjour', 'Hello', 'Hallo', 'Dzien dobry'] }
      ]
  },
  "Goodbye"
]

Extracts all keys and values from a hash into an array:

Gitlab::Utils::MergeHash.crush(
  { hello: "world", this: { crushes: ["an entire", "hash"] } }
)

Gives:

[:hello, "world", :this, :crushes, "an entire", "hash"]

`Override`

Refer to override.rb:

This utility can help you check if one method would override another or not. It is the same concept as Java's @Override annotation or Scala's override keyword. However, we only run this check when ENV['STATIC_VERIFICATION'] is set to avoid production runtime overhead. This is useful for checking:
- If you have typos in overriding methods.
- If you renamed the overridden methods, which make the original override methods irrelevant.
  
  Here's a simple example:
```
class Base
  def execute
  end
end

class Derived < Base
  extend ::Gitlab::Utils::Override

  override :execute # Override check happens here
  def execute
  end
end
```
  This also works on modules:
```
module Extension
  extend ::Gitlab::Utils::Override

  override :execute # Modules do not check this immediately
  def execute
  end
end

class Derived < Base
  prepend Extension # Override check happens here, not in the module
end
```
  Note that the check only happens when either:
  - The overriding method is defined in a class, or:
  - The overriding method is defined in a module, and it's prepended to a class or a module.
  Because only a class or prepended module can actually override a method. Including or extending a module into another cannot override anything.

Interactions with `ActiveSupport::Concern`, `prepend`, and `class_methods`

When you use ActiveSupport::Concern that includes class methods, you do not get expected results because ActiveSupport::Concern doesn't work like a regular Ruby module.

Since we already have Prependable as a patch for ActiveSupport::Concern to enable prepend, it has consequences with how it would interact with override and class_methods. As a workaround, extend ClassMethods into the defining Prependable module.

This allows us to use override to verify class_methods used in the context mentioned above. This workaround only applies when we run the verification, not when running the application itself.

Here are example code blocks that demonstrate the effect of this workaround: following codes:

module Base
  extend ActiveSupport::Concern

  class_methods do
    def f
    end
  end
end

module Derived
  include Base
end

# Without the workaround
Base.f    # => NoMethodError
Derived.f # => nil

# With the workaround
Base.f    # => nil
Derived.f # => nil

`StrongMemoize`

Refer to strong_memoize.rb:

Memoize the value even if it is nil or false.

We often do @value ||= compute. However, this doesn't work well if compute might eventually give nil and you don't want to compute again. Instead you could use defined? to check if the value is set or not. It's tedious to write such pattern, and StrongMemoize would help you use such pattern.

Instead of writing patterns like this:
```
class Find
  def result
    return @result if defined?(@result)

    @result = search
  end
end
```
You could write it like:
```
class Find
  include Gitlab::Utils::StrongMemoize

  def result
    search
  end
  strong_memoize_attr :result

  def enabled?
    Feature.enabled?(:some_feature)
  end
  strong_memoize_attr :enabled?
end
```
Using strong_memoize_attr on methods with parameters is not supported. It does not work when combined with override and might memoize wrong results.

Use strong_memoize_with instead.
```
# bad
def expensive_method(arg)
  # ...
end
strong_memoize_attr :expensive_method

# good
def expensive_method(arg)
  strong_memoize_with(:expensive_method, arg)
    # ...
  end
end
```
There's also strong_memoize_with to help memoize methods that take arguments. This should be used for methods that have a low number of possible values as arguments or with consistent repeating arguments in a loop.
```
class Find
  include Gitlab::Utils::StrongMemoize

  def result(basic: true)
    strong_memoize_with(:result, basic) do
      search(basic)
    end
  end
end
```

Clear memoization

class Find
  include Gitlab::Utils::StrongMemoize
end

Find.new.clear_memoization(:result)

`RequestCache`

Refer to request_cache.rb.

This module provides a simple way to cache values in RequestStore, and the cache key would be based on the class name, method name, optionally customized instance level values, optionally customized method level values, and optional method arguments.

A simple example that only uses the instance level customised values is:

class UserAccess
  extend Gitlab::Cache::RequestCache

  request_cache_key do
    [user&.id, project&.id]
  end

  request_cache def can_push_to_branch?(ref)
    # ...
  end
end

This way, the result of can_push_to_branch? would be cached in RequestStore.store based on the cache key. If RequestStore is not currently active, then it would be stored in a hash, and saved in an instance variable so the cache logic would be the same.

We can also set different strategies for different methods:

class Commit
  extend Gitlab::Cache::RequestCache

  def author
    User.find_by_any_email(author_email)
  end
  request_cache(:author) { author_email }
end

`ReactiveCaching`

Read the documentation on ReactiveCaching.

GitLab utilities

MergeHash

Override

Interactions with ActiveSupport::Concern, prepend, and class_methods

StrongMemoize

RequestCache