How To Avoid Using Globals For Information Passing

If you’ve run into the classic noob C programmer’s problem of “Why won’t my function update my variable?” then you might have resorted to using a global variable instead of passing the variable into the function as a parameter. Global variables allow ‘write access’ to number from both the caller and the callee, like this,

but global variables are considered bad for lots of reasons. Not least of which is that your function can no longer double anything other than your specific number.

Instead you need to pass number into doubleIt as a pointer parameter.

How Many Threes Are There In The Universe?

The obvious answer is “quite a lot”. There’s the three on the front of my house, there are a couple more in my phone number and there’s one on the number plate of the car across the street from me right now. There are also several copies of my house number three, stored in the databases of various shopping sites who have been kind enough to deliver goodies to my home address over the years. Ditto for my phone number threes; and the DVLA, police ANPR, and several insurance companies presumably have copies of the three on my neighbour’s car number plate. Now in some ways these threes are very different from each other – my neighbour’s car has little to do with my phone. And neither of these have very much at all to do with that group of three wildebeest over yonder on the African savannah. And yet, in some ways, some of them are very similar – the pair of threes in my phone number, for example, share quite a lot of context and, other than their position in the string of digits, you’d be hard pressed to distinguish one from the other. You’ll notice though that I’ve not used the digit ‘3’ anywhere in the discussion so far. And yet that’s what’s encoded in Amazon’s database, even though my house is adorned with the word ‘three’ on a little piece of green Lakeland slate. If I happened to live in certain parts of Canada or in France, I might have the word ‘trois’ in place of the word ‘three’. So, not only are there many different threes, there are also many different representations of the same three. What they all have in common though is some notion of ‘threeness’. All threes are clearly distinct from all fours, fives and 3.14159s. That’s what it means to be three. We can take this platonic idea of three and call it the master three. If we only have one master three then it makes sense that all the other threes we have, the real-world threes rather than the platonic, idealised, ‘perfect’ three, can really just refer to the master three to get their threeness characteristic. In fact they probably should do that so that we’re not duplicating information needlessly, and potentially having conflicting ideas of threeness. Threeness is threeness and it only needs to be defined in one place. All the other bits and pieces of information related to the real-world threes are specific to to those threes rather than having anything to do with the general idea of threeness. The encoding format of the bit pattern on my phone’s SIM card, the font and colour of the numberplate, the origin of the stone used to make my house number. None of these things help us understand the threeness of these threes. This idea of a separation of concerns is at the heart of referential semantics. But, apart from the very real benefit of letting me express clearly what I mean, what does it allow me to do in practice? Well, as we’ve seen, I can refer to the threeness attribute of the master three from my own threes, maintaining a strong and secure threeness without risk of corrupting it and crucially, from a practical standpoint, without incurring any extra storage or processing costs.

Referential Semantics

This is going to be a bit abstract but it’s honestly really worth trying to get your head around if you want a good solid foundation in pointers and referential semantics in general.

Pretty much everybody starts their programming career by doing some Hello World stuff and then moving on to some simple variable manipulation like this:

If you run this code you’ll see that b is equal to 2 but a is still equal to 1. This is known as value semantics or sometimes copy semantics. When we do int b = a we’re making b store a copy of the value stored in a. Subsequent changes to b will have no effect on a.

In reference semantics we’d do this instead:

This would result in both a and *b having a value of 2. Here, a is set to 2 indirectly and implicitly through the reference provided by b.

You can read about pointer operators if you don’t understand all those * and & thingies but what’s important here is the idea of references.

In value semantics we’re dealing with objects directly but in reference semantics, a reference provides a link to another thing. The referrer refers to the referent. C pointers are one example of a reference but there are other ways to implement reference semantics – in other words reference is a generic term and pointer is a specific example of a reference.

Referencing in turn is a specific case of indirection, the process of gaining access to one thing via another. The opposite of referencing is dereferencing. There’s more info in my article on C’s pointer syntax.

That’s the top and bottom of references really but if you think about them, they do open up your mind to some fairly deep observations about the world. For example here’s a question about numbers.

Finally, here are a couple of subtleties. See if you can answer them yourself before reading my take on them.

What’s the difference between a variable and a pointer? Both are ‘indirections’. The defining difference is that a pointer is a variable whose value is an address which further indirects to a value whereas a variable indirects to a value in a single step.

What’s the difference between an address and a pointer? Not very much really. Both can be used in the same places e.g.

int value;
int i = *(&value);

is equivalent to

int value;
int i;
int *ptr = &value;
i = *ptr;