middlewarer

How it works #

Middlewarer is a Go generator which takes in the name of a Go interface (let’s call it Foo) and generates a middleware framework for it.

It then creates a new file and defines within it a struct implementing Foo which will act as the middleware.

Users can then call the generated wrapper function WrapFoo, which accepts as parameters an implementation of Foo to wrap and middleware functions to call. These middleware functions can be nil, meaning that no boilerplate code is required for methods the user doesn’t want to attach middleware to. This wrapper function then returns another implementation of Foo which can now be used in place of the original implementation passed to the wrapper.

The passed middleware functions can then easily operate on the method arguments (e.g. sanitize them, change their format etc), add logging, metrics and more without the programmer having to ever touch their original implementation of Foo.

Why I created it #

I created middlewarer as a solution to a problem I encountered with dinkel. In dinkel, I had an implementation of the DB interface for every system I would test and a Strategy implementation for every fuzzing strategy.

However, I noticed that I would have to refactor my code drastically when I wanted to add a Prometheus exporter to these implementations, which would provide metrics on query generation latency, found bugs, invalid query rate, etc. I couldn’t add a generic Prometheus exporter to them, as there isn’t a clean way of generalizing the metrics of the different systems and strategies.

Additionally, not all interface methods provide insightful metrics, so it was important for me that I don’t have to worry about having to create unnecessary boilerplate code for every interface method that I didn’t want to attach middleware software to.

As a solution, I came up with the idea of writing my first Go generator, which allows easy creation of an efficient middleware framework for any interface which doesn’t require the programmer to define middleware functions for every interface method. The solution worked flawlessly for dinkel and I use it there for logging and metrics collection.