This library is a translation in Elixir of the famous poolboy : https://github.com/devinus/poolboy ... I just don't like to mix Erlang and Elixir code (Elixir is clearer than its father) and this library is a great one.
Version 0.1.0
- ttl to deallocate overflow workers ( see PR of poolboy )
- changing pool size dynamically
Version 0.2.0
- Possibility to broadcast messages to all workers.
We first define quickly a worker. Notice handle_info callback
iex(2)> defmodule PoolgirlTestWorker do
...(2)> @moduledoc false
...(2)>
...(2)> use GenServer
...(2)> @behaviour PoolgirlWorker
...(2)>
...(2)> def start_link(_args) do
...(2)> GenServer.start_link(__MODULE__, [], [])
...(2)> end
...(2)>
...(2)> def init(_opts) do
...(2)> {:ok, :undefined}
...(2)> end
...(2)>
...(2)> def handle_call(_msg, _from, state) do
...(2)> {:reply, :ok, state}
...(2)> end
...(2)> def handle_cast(_msg, state) do
...(2)> {:noreply, state}
...(2)> end
...(2)> def handle_info({:broadcast, msg}, state) do
...(2)> {:noreply, state}
...(2)> end
...(2)> end
We start the pool with a dynamic configuration but we could have put it in our .exs config file.
iex(3)> Poolgirl.start_link([{:name, {:local, :poolgirl_test}},
{:worker_module, YourModule}, {:size, 10}, {:max_overflow, 5},
{:broadcast_to_workers, :true}])
{:ok, #PID<0.498.0>}
We can check out a worker to do some work
iex(4)> pid = Poolgirl.checkout(:poolgirl_test)
\#PID<0.509.0>
We have to check in it when we have finished to use it
iex(5)> Poolgirl.checkin(:poolgirl_test, pid)
:ok
We can send a message to the entire pool, even to workers that are working when the message is broadcast. You just have to implement the callback in your worker. The message is like {:broadcast, msg}. msg is your message and :broadcast atom is added. So when the following call is done, the message that reaches the worker is {:broadcast, {:test}}
iex(6)> Poolgirl.async_broadcast_to_pool(:poolgirl_test, {:test})
:ok
You can have a report of the pool
iex(7)> Poolgirl.status(:poolgirl_test)
{:ready, {:available_workers, 10}, {:overflow_workers, 0}, {:checked_out_workers, 0}}
You can reduce / increase the pool size. When you increase the pool, if some process are waiting for a worker, they obtain a worker immediately. If some workers are in an overflow state, they become "regular" workers (up to the correct size of course). If you decrease the pool size, the pool is not reduce if all workers are busy working. The number of workers will be reduced when the workers are checked in. You can set an overflow_ttl. This means that the worker is not destroyed until the ttl has expired. This option permits to save time.
iex(8)> Poolgirl.change_size(:poolgirl_test, 5)
:ok
This function gives the configuration. All these parameters can be set in the configuration.
iex(9)> Poolgirl.give_conf(:poolgirl_test)
%{broadcast_to_workers: true, max_overflow: 5, overflow_ttl: 0, size: 5,
strategy: :lifo}
In your config file, you can put something like :
config :poolgirl, [ worker_module: YourModule, size: 10, max_overflow: 5 ] and then in your supervisor code
config = Application.get_env(:my_app, :poolgirl)
children = [
Poolgirl.child_spec(:pool_id, config, [ < args and options> ])
]
Bugs are mine and for the rest, all credits are for Devinus and his fellow maintainers of the project. Support them !
Poolgirl, like her boyfriend, is available in the public domain (see UNLICENSE). Poolgirl is also optionally available under the ISC license (see LICENSE), meant especially for jurisdictions that do not recognize public domain works.