# `ImagePipe.ProcessingPool`
[🔗](https://github.com/hlindset/image_pipe/blob/v0.2.0/image_pipe/lib/image_pipe/processing_pool.ex#L1)

Limits how many images are processed at once, with a queue for requests that
wait for a turn and a deadline for each image.

    children = [
      {ImagePipe.ProcessingPool, name: MyApp.Pool, max_concurrency: 8, max_queue: 16},
      {ImagePipe, name: MyApp.Images, processing_pool: MyApp.Pool, sources: [...]},
      MyAppWeb.Endpoint
    ]

Every configuration that names the pool with `:processing_pool` shares its
capacity, across mounts and `ImagePipe.run/4` calls on the same node.
Workers finishing after a timeout or cancellation keep their slots until
their operations and cleanup finish.
[Limiting concurrent processing](processing-controls.md) covers choosing the
limits and what counts toward them.

## Errors

`ImagePipe.run/4` returns these errors, and a request gets a `503` for each:

  * `{:processing, :overloaded}` - the queue is full.
  * `{:processing, :queue_timeout}` - the request waited longer than
    `:queue_timeout`.
  * `{:processing, :timeout}` - the image took longer than
    `:processing_timeout`.
  * `{:processing, :unavailable}` - the pool isn't running.

## Options

* `:name` (`t:atom/0`) - Name that `ImagePipe.config/1`'s `:processing_pool` refers to.

* `:max_concurrency` (`t:pos_integer/0`) - Required. Most images processed at once.

* `:max_queue` (`t:non_neg_integer/0`) - Most requests waiting for a turn. A request that finds the queue full fails with `{:processing, :overloaded}`. With `0`, requests never wait. The default value is `64`.

* `:queue_timeout` (`t:pos_integer/0`) - Longest wait for a turn, in milliseconds. A request that waits longer fails with `{:processing, :queue_timeout}`. The default value is `10000`.

* `:processing_timeout` (`t:pos_integer/0`) - Request deadline once processing starts, in milliseconds, from reading the original to sending the last byte. A request that takes longer fails with `{:processing, :timeout}`. Its worker keeps the slot until the current operation and cleanup finish. The default value is `30000`.

# `child_spec`

Returns a specification to start this module under a supervisor.

See `Supervisor`.

# `start_link`

Starts a pool with validated admission and timeout options.

# `stats`

Returns the number of workers holding processing slots and of requests waiting, as
`%{active: 3, queued: 0}`.

Active workers include those finishing after a timeout or cancellation.

---

*Consult [api-reference.md](api-reference.md) for complete listing*