go-aqueue

Go Asynchronous Queue is a better alternative to go channels. This package requires go1.8 or newer.

Background

The Message Oriented Programing (or MOP) is not a novelty and is a concept known for decades. This concept was implemented in many different languages either natively or through various libraries and it is indispensable tool for developing higly scalable and extremelly reliable applications and services.

It is nice to see an attempt from Go team to embed it into the language as native feature. However, there is a number of shortcomings in the current implementation of go channels and this package is an attemt to provide a better version.

At the core of MOP is an asynchronows queue and Go channel is a version of it. The asyncronous queue used in MOP may resemble some known concurrency patterns (pub/sub, pipeline, etc.) but it has different requirements:

• It is intended to communicae between exactly two parties (i.e. fair queuing is out of scope)
• It should not (arguably) provide any buffering. This requirement is based on personal experience and I'm yet to see a use case where buffering provides any advantages.

Comparison with go channels

Here is a list of advantages and disadvantages of this package in comparison with go channels:

Pros:

• AQueue can be gracefully closed from any thread and both reader and writer will exit with an error instead of panic.
• AQueue provides blocking as well as non-blocking methods for wrining to the queue and reading from it.
• Some AQueue methods return cancel() function that can be used for implementing timeouts (i.e. in conjunction with context.WithDeadline() or context.WithTimeout()) or gracefully closing the queue on context.Done().

Cons:

• Type safety can only be assured at run time, mostly because of lack of generics in golang.
• Performance of AQueue is 2-5 times slower than that of native channels. The difference in performance depends on hardware and there are benchmarks included with the package that will show the actual difference on your particular hardware.

ToDo

• Provide more code examples
• Improve documentation

About the Author

The author of this package has extensive experience implementing MOP concept from highly efficient multimedia processing on low powered embedded devices to parallel execution of thousands of cloud based applications performing financial modeling.

License

See LICENSE.