SYNQueue
is a subclass of NSOperationQueue
so you get:
- Serial or concurrent queues
- Task priority
- Multiple queues
- Dependencies
- KVC/KVO
- Thread safety
But it goes beyond NSOperationQueue
and NSOperation
to offer:
- Task persistence (via protocol)
- Queue specific logging (via protocol)
- Retries (exponential back-off)
With a good queuing solution you can provide a much better user experience in areas such as:
- Web requests
- Saving/creating content (images, video, audio)
- Uploading data
When we started building SYNQueue
, persistence was the most important feature for us since we hadn't seen a good generic implementation of it anywhere. With that in mind, we designed each SYNQueueTask
(:NSOperation
) to hold just metadata about the task rather than code itself.
The actual code to perform the task gets passed to the queue in the form of a taskHandler
closure. Each SYNQueueTask
must have a taskType
key which corresponds to a specific taskHandler
.
For a thorough example see the demo project in the top level of the repository.
let queue = SYNQueue(queueName: "myQueue", maxConcurrency: 2, maxRetries: 3,
logProvider: ConsoleLogger(), serializationProvider: NSUserDefaultsSerializer(),
completionBlock: { [weak self] in self?.taskComplete($0, $1) })
The
logProvider
andserializationProvider
must conform to theSYNQueueLogProvider
andSYNQueueSerializationProvider
protocols respectively.See NSUserDefaultsSerializer.swift and ConsoleLogger.swift for example implementations.
The
completionBlock
is the block to run when a task in the queue completes (success or failure).
let t1 = SYNQueueTask(queue: queue, taskID: "1234", taskType: "uploadPhoto", dependencyStrs: [], data: [:])
The
queue
is the queue you will add the task to.taskID
is a unique ID for the task.taskType
is the generic type of task to perform. EachtaskType
will have its owntaskHandler
.data
is any data your task will need to perform its job.
let t2 = SYNQueueTask(queue: queue, taskID: "5678", taskType: "submitForm", dependencyStrs: [], data: [:])
t2.addDependency(t1)
queue.addOperation(t2)
queue.addOperation(t1)
Notice that even though we add task
t2
to the queue first, it will not execute until its dependency,t1
has finished executing.`
You may have realized that you are free to serialize tasks however you like through the SYNQueueSerializationProvider
protocol. The one caveat is that all tasks must be idempotent. That is, even if called multiple times, the outcome of the task should be the same. For example: x = 1
is idempotent, x++
is not.
Hopefully this makes sense given that a serialized task may get interrupted before it finishes, and when we deserialize this task we will run it again. We only remove the serialized task after it has completed (success or failure).