Skip to main content

An instantiated object should be "ok"

I've been QA'ing quite a bit of work recently and one common theme I've noticed across both Java and C# projects I have been looking at is that we occasionally open ourselves up unessacarily to Exceptions by the way objects are being created.

My general rule of thumb (which I have seen mentioned in a Pluralsight video recently but also always re-iterate in various Robust Software talks I have done) is that you shouldn't be able to create an object and then call a method or access a property that then throws an exception. At worst, it should return null (I'm not going to moan about that now).

I've created an example below. We have two Dojos, one is good and one is bad. The bad dojo looks very familiar though. It's a little class written in the style that seems often encouraged. In fact, many classes start life as something like this. Then as years go on, you and other colleagues add more features to the class and it's instantiation becomes a second class citizen. One reason for this I guess is that changing how an old class is instantiated can often involve some significant refactoring.

Below is a series of classes that describes people that can attend a Dojo. Either a Sensei or a Student. They're not really pertinent to the problem.

The Bad Dojo looks like a nice little class initially doesn't it. It's small, succint and extremely simple. It's tempting to leave it like this as it has no methods on it.

The issue is that we want to use the list of students in the summary, when we use this class we start to run into problems.... can you spot what will go wrong?

Because BadDojo.Summary uses Students without checking to see if it's initialized, you run the chance of getting a null reference exception. In the example above Students definitely will be null.
There's lots of ways around this, one is to always check to see if reference properties are null in the first place. This can often be prudent. My initial preference is to always make sure the class is constructed in a safe way in the first place through forcing the consuming class to instantiate the object with the values it needs. You will often need to do both.

The Good Dojo below allows itself to be created either with or without students but always insists in a Sensei. If no students are supplied it initializes it's own internal list of students to ensure it's in a useable state. To me this is the best way to operate as you can then worry less about someone in the future forgetting to check if Students is null first. if(this.Students != null ....)
Now we can use the Summary property safely!

Pfft all obvious you fool!

Well yes this seems obvious now doens't it. Check your codebase, are you definitely adhering to this! In a language like F# we are less susceptible to these issues but we work with what we have.

Another side effect of being strict about how our object is created is that we have made it Immutable. I think I've already posted about this. But I will probably create another post on this soon based on these classes.


Popular posts from this blog

NESTA - Next Gen.

via Following on from an article on the BBC about Raspberry Pi, this next gen report has some interesting findings. The scariest stat which I picked out from the BBC website was "out of the 28,767 teachers who were awarded Qualified Teacher Status... in 2010, only three qualified in computing or computing science as their primary qualification" Having worked as a computer science teacher for a year in a school that was a specialist in Computing I can concur that the uptake in Comp Sci was woeful. 2 Students for A2... The other teachers backgrounds in Computer Science was also fairly woeful (most knowing a bit about Office but still a paltry amount even about that). I couldn't speak for my counterpart that I was covering however. I suspect they were fairly up on things. All in all what kills me is that Computer science is not a secondary level subject. Areas are often covered, a little in IT, a little in DT subjects (if kids choose Systems and Contr

Accessing the UI Thread with Tasks in F#

I have a Windows Forms program written in F# that can deploy a code base to n number of sites at once (you select the sites you would like to deploy to and it goes off and completes a number of tasks (backing up current sites, various unpacking and moving of files etc... ). Once you start it, it begins it's merry journey and begins to update the UI with what has happened. At the moment this method of updating the UI is not pretty because the threads I am doing the work on can't update the UI so I perform some fiendery to make that happen (don't ask). I knew there was a better way using some newer .NET features but I just hadn't got round to having a fiddle yet. I have now found that if you use the built in Task class but break your code up in a nicer way and then chain the tasks together you can then pass the correct context into the task that you want to talk to the UI. Here's a little script to give you a feel for it. You can press the "start" butt