Programming (explanations for the non-programmers, mostly)

Okay, so, thinking about pointers, and C.

The thing here is in part that C has an odd relationship between declarations and things. The famous claim is "declaration reflects use". So, the idea is this. Imagine that ip is a pointer to an int, and currently valid (it points at something). Then *ip is "the thing pointed to by ip". How do you declare a pointer to int? "int *ip;" And people often find that confusing, because the * (in that context) means "the contents of the thing", so it goes from a pointer to a not-a-pointer. But when you put it on a declaration, it converts the thing into a pointer. And this seems to be Very Confusing.

What "declaration reflects use" means is this: "int ???;" declares that ??? is an int. In "int *ip;" the ??? is "*ip". So "*ip" is an int, which means that ip must be a pointer-to-int.

So, okay, back to the general question of "pointers". Okay, so, the computer's memory holds a bunch of things, and in practice they are at some level all numbers. And you can view the computer's memory as having an inherent order to it. If you have 8GB (~8 billion bytes) of storage, there is such a thing as "the first byte", "the second byte", and so on.

And that means that one of the things a number could mean to the programmer is "a location in memory where a thing is being stored".

What this allows you to do is make things that use a pointer to find the data they are going to operate on, so that running the same code with two different pointers does the same thing to two different things.

One analogy people sometimes use is that the "address" (the "which location in memory" thing) is reasonably analagous to a street address. It could be something like "304 Elm Street" or "317 Plum Street". So say you have a window-washing business. You have a list of addresses, and then you go to each address and wash the windows there. You don't have to have a separate set of instructions for each set of windows; you have a general set of instructions for "wash the windows", and then you just apply it to each address in turn.

So the most common things to do with pointers are to "dereference" them, which is to say, interact with the thing they point to, or to "take the address of" a thing, which is to say, obtain a value which can be used to point to that thing. And you can also do arithmetic on them, answering questions like "okay, but what's the address of the object after this one?"

And a huge number of problems, including about 99% of all the crashes you will ever see, involve pointers which are wrong somehow. Either the pointer is just plain garbage (does not actually describe a location in memory, say), or it's wrong in some other way, like pointing to a different thing.

 

Yeah, it's confusing as hell. I honestly still can't reliably do complicated things like "pointer to function returning array of pointers" without typedefs for the intermediate stuff.

BTW, that is also why int* ip, jp; doesn't do what the novice programmer probably expects; it ends up being parsed as if it were int *ip; int jp; and jp is just a regular int.

Same thing with arrays, mostly; int a[5]; means "a is an array of 5 things, such that a[x] is an int". Although a[5] isn't a valid object, because C is 0-indexed and the members are 0 through 4.

 

Lua is also 1-indexed. Oddly, it seems to give me less trouble in lua specifically than in some other languages.

 

Yup. And the real answer is, dereferencing a pointer would give you an int. Confusing until that clicks. Then still confusing but less so.

 

I learned with int *ip and am used to that because the * is part of the declaration of the specific variable, but... There's a lot of variance. Also, I don't do C++ much because I am really not good at it. I have not made the effort to get used to how C++ works, and every time I try it's changed again anyway. :P

 

I added a new bbcode:

Inline rendering of [c]sample code[/c] => sample code.