Dealing with Now and why I'm almost done with C# and Java

May 03, 2011

It's a common scenario to have to use the current time within code. A good example is timestamping an audit trail. As part of your code to create an Audit object, you might do something like:

var audit = new Audit
	UserName = user.Name,
	Dated = DateTime.Now,

The problem with the above code is that it's hard to test. Personally, I think testing that the proper time is used for your audit record is important. Depending on what field you're in, your lawyers might think so too. However, if we wrote a very specific test, we might be in for a surprise. Can you spot the problem?

public void ItSetsTheAuditTimeToRightNow()
	var audit = CreateAuditItem(new User{Name = "Leto"});

Sometimes such a test will pass, sometimes it won't. Why? Because between setting DateTime.Now in the code and then verifying it in the test, the time might have changed.

What solution do we use? Well, when all you have is a hammer...That's right, I've seen this problem "solved" by injecting an abstract dependency into objects. Now we can mock and fully control the creation of time.

And you know what? That's just stupid. Getting the current time shouldn't need abstraction nor decoupling. I want to use DateTime.Now in my code and I want it to be testable. When a language or framework or whatever makes you jump through hoops to get the current time then something is broken.

In Ruby, you use Time.now. Why? Because Time in Ruby is an object. In C# or Java it'd be a class. But in Ruby, as well as most (all?) dynamic languages, it's an actual object. So we can just change it to do/be/return whatever we want whenever we want (like at runtime).

I don't think C++ developers who've spent 10+ years dealing with memory management have any concept of how much energy/productivity they spend on memory management. I don't think C#/Java developers have any concept of how much energy/productivity they spend on decoupling. Worse, I think many C#/Java developers don't understand that DI, interface-based programming, and fear of statics, is result of language limitations. Developers just think it's how you code - regardless of the language.

Not too long ago that was me...IoC, mocking, interfacing out the wazoo. I didn't realize that how I coded was really just a by-product of how my language required me to code. I think this was one of those fundamental eye openers for me. I've had a handful in my career and I always treasure them when they happen.

I do want to point out that C#'s lambdas largely solve this simple case. I don't need anyone to post a comment including Func<DateTime>. It isn't perfect, but it's certainly more than close enough for me. However, move to something just a little more complicated, and you're back at writing infrastructure and architecture rather than adding business value.