SimpleAgent is a simplification/abstraction layer around the base Elixir Agent module.

Often times, Agents are used to store a simple value, such as an atom or an integer. This is used as a flag or a count which multiple processes can access/update. In these cases, the full Agent module is used with boilerplate closure code that is repetative, adds noise to the code, and can be eliminated. For example, to create an agent, update the value, then retrieve that value, you would run:

{:ok, agent} = Agent.start_link(fn -> nil end)
Agent.update(agent, fn _ -> :final_value end)
final_value = Agent.get(agent, fn val -> val end)

SimpleAgent boils these calls down to a more readable:

agent = SimpleAgent.start!
SimpleAgent.update! agent, :final_value
completed = SimpleAgent.get! agent

For Integer manipulation, SimpleAgent takes this code:

{:ok, agent} = Agent.start_link(fn -> 0 end)
IO.puts Agent.get_and_update(fn val -> {val + 1, val + 1} end)
IO.puts Agent.get_and_update(fn val -> {val - 1, val - 1} end)
IO.puts Agent.get_and_update(fn val -> {val + 1, val + 1} end)
IO.puts Agent.get_and_update(fn val -> {val + 1, val + 1} end)

and boils it down to the more readable:

agent = SimpleAgent.start! 0
IO.puts SimpleAgent.increment! agent
IO.puts SimpleAgent.decrement! agent
IO.puts SimpleAgent.increment! agent
IO.puts SimpleAgent.increment! agent

SimpleAgent is very useful in testing. For example:

test "foo calls bar 3 times" do
  bar_call_agent = SimpleAgent.start! 0
  :meck.new(Bar)
  :meck.expect(Bar, :bar, fn -> SimpleAgent.increment!(bar_call_agent) end)
  Foo.foo()
  assert SimpleAgent.get?(bar_call_agent) == 3
end

Why only simple types?

When a complex state such as a map or a dict is in use, the correct way to manipulate the complex state is in the Agent server via a closure. This prevents the entire state from being copied from the Agent Server to the Client (see the Agent docs for more information on this). For states with these complex types, you should use the full Agent module. SimpleAgent is for those cases where the "entire state" is a single simple Integer, String, or Atom (including nil, true, and false).

Features:

• Simple types and updates reduce chances of errors, so all calls raise exceptions instead of requiring boilerplate pattern matching, and start!/2 is available instead of start_link
• No closures are required.
  • get!/1 uses &(&1)
  • update!/2 takes the value instead of a function and uses fn _ -> value end
• nil support
  • start!/2 defaults the initial value to nil when not specified
  • nil?/1 checks for the nil state
  • clear/1 sets the nil state
• increment!/2 and decrement!/2 allow for simple manipulation of integer states.