Question

I would like to improve testability of some legacy code. To achieve this, I am introducing interfaces for existing classes (and have existing classes implement those) and factories that create an instance of a test object or an object of the original class, depending on some configuration setting.

I can foresee some internal feedback along the lines of ‘but this will affect performance’, but I would like to be able to test some code (for a service layer in this case), without having to deploy all underlying layers and setup a database server.

Do you have any experience where the indirection introduced did affect performance noticeably? What would be a constructive way to respond to the aforementioned feedback?

Thanks,

Martijn

Answer 1

According to a book from Don Box (Essential .NET volume 1), the effect of calling a function on an interface as compared with calling it on a class is one machine instruction, as there is one more indirection. This would mean that on a 2 GHz processor, it would be 0.0000000005 seconds slower to call an interface method than a class method.

This refers to the .net release 1 implementation (I have an old edition of this book). I am not sure if that changed a lot for newer versions, or how it applies to Mono, but I would definitely not assume dramatic effects.

As you see, on modern computers, the effect should be neglectable except in the rare case that you have to hunt for every nanosecond.