[MUSIC PLAYING] 

DOUG LLOYD: Pointers, here we are. This is probably going to be the most difficult topic that we talk about in CS50. And if you've read anything about pointers before you might be a little bit intimidating going into this video. It's true the pointers do allow you the ability to perhaps screw up pretty badly when you're working with variables, and data, and causing your program to crash. But they're actually really useful and they allow us a really great way to pass data back and forth between functions, that we're otherwise unable to do. 

And so what we really want to do here is train you to have good pointer discipline, so that you can use pointers effectively to make your programs that much better. As I said pointers give us a different way to pass data between functions. Now if you recall from an earlier video, when we were talking about variable scope, I mentioned that all the data that we pass between functions in C is passed by value. And I may not have used that term, what I meant there was that we are passing copies of data. When we pass a variable to a function, we're not actually passing the variable to the function, right? We're passing a copy of that data to the function. The function does what it will and it calculates some value, and maybe we use that value when it gives it back. 

There was one exception to this rule of passing by value, and we'll come back to what that is a little later on in this video. If we use pointers instead of using variables, or instead of using the variables themselves or copies of the variables, we can now pass the variables around between functions in a different way. This means that if we make a change in one function, that change will actually take effect in a different function. Again, this is something that we couldn't do previously, and if you've ever tried to swap the value of two variables in a function, you've noticed this problem sort of creeping up, right? 

If we want to swap X and Y, and we pass them to a function called swap, inside of the function swap the variables do exchange values. One becomes two, two becomes one, but we don't actually change anything in the original function, in the caller. Because we can't, we're only working with copies of them. With pointers though, we can actually pass X and Y to a function. That function can do something with them. And those variables values can actually change. So that's quite a change in our ability to work with data. 

Before we dive into pointers, I think it's worth taking a few minutes to go back to basics here. And have a look at how computer memory works because these two subjects are going to actually be pretty interrelated. As you probably know, on your computer system you have a hard drive or perhaps a solid state drive, some sort of file storage location. It's usually somewhere in the neighborhood of 250 gigabytes to maybe a couple of terabytes now. And it's where all of your files ultimately live, even when your computer is shut off, you can turn it back on and you'll find your files are there again when you reboot your system. But disk drives, like a hard disk drive, an HDD, or a solid state drive, an SSD, are just storage space. 

We can't actually do anything with the data that is in hard disk, or in a solid state drive. In order to actually change data or move it around, we have to move it to RAM, random access memory. Now RAM, you have a lot less of in your computer. You may have somewhere in the neighborhood of 512 megabytes if you have an older computer, to maybe two, four, eight, 16, possibly even a little more, gigabytes of RAM. So that's much smaller, but that's where all of the volatile data exists. That's where we can change things. But when we turn our computer off, all of the data in RAM is destroyed. 

So that's why we need to have hard disk for the more permanent location of it, so that it exists- it would be really bad if every time we turned our computer off, every file in our system was obliterated. So we work inside of RAM. And every time we're talking about memory, pretty much, in CS50, we're talking about RAM, not hard disk. 

So when we move things into memory, it takes up a certain amount of space. All of the data types that we've been working with take up different amounts of space in RAM. So every time you create an integer variable, four bytes of memory are set aside in RAM so you can work with that integer. You can declare the integer, change it, assign it to a value 10 incremented by one, so on and so on. All that needs to happen in RAM, and you get four bytes to work with for every integer that you create. 

Every character you create gets one byte. That's just how much space is needed to store a character. Every float, a real number, gets four bytes unless it's a double precision floating point number, which allows you to have more precise or more digits after the decimal point without losing precision, which take up eight bytes of memory. Long longs, really big integers, also take up eight bytes of memory. How many bytes of memory do strings take up? Well let's put a pin in that question for now, but we'll come back to it. 

So back to this idea of memory as a big array of byte-sized cells. That's really all it is, it's just a huge array of cells, just like any other array that you're familiar with and see, except every element is one byte wide. And just like an array, every element has an address. Every element of an array has an index, and we can use that index to do so-called random access on the array. We don't have to start at the beginning of the array, iterate through every single element thereof, to find what we're looking for. We can just say, I want to get to the 15th element or the 100th element. And you can just pass in that number and get the value you're looking for. 

Similarly every location in memory has an address. So your memory might look something like this. Here's a very small chunk of memory, this is 20 bytes of memory. The first 20 bytes because my addresses there at the bottom are 0, 1, 2, 3, and so on all the way up to 19. And when I declare variables and when I start to work with them, the system is going to set aside some space for me in this memory to work with my variables. So I might say, char c equals capital H. And what's going to happen? Well the system is going to set aside for me one byte. In this case it chose byte number four, the byte at address four, and it's going to store the letter capital H in there for me. If I then say int speed limit equals 65, it's going to set aside four bytes of memory for me. And it's going to treat those four bytes as a single unit because what we're working with is an integer here. And it's going to store 65 in there. 

Now already I'm kind of telling you a bit of a lie, right, because we know that computers work in binary. They don't understand necessarily what a capital H is or what a 65 is, they only understand binary, zeros and ones. And so actually what we're storing in there is not the letter H and the number 65, but rather the binary representations thereof, which look a little something like this. And in particular in the context of the integer variable, it's not going to just spit it into, it's not going to treat it as one four byte chunk necessarily, it's actually going to treat it as four one byte chunks, which might look something like this. And even this isn't entirely true either, because of something called an endianness, which we're not going to get into now, but if you're curious about, you can read up on little and big endianness. But for the sake of this argument, for the sake of this video, let's just assume that is, in fact, how the number 65 would be represented in memory on every system, although it's not entirely true. 

But let's actually just get rid of all binary entirely, and just think about as H and 65, it's a lot easier to think about it like that as a human being. All right, so it also seems maybe a little random that I've- my system didn't give me bytes 5, 6, 7, and 8 to store the integer. There's a reason for that, too, which we won't get into right now, but suffice it to say that what the computer is doing here is probably a good move on its part. To not give me memory that's necessarily back to back. Although it's going to do it now if I want to get another string, called surname, and I want to put Lloyd in there. I'm going to need to fit one character, each letter of that's going to require one character, one byte of memory. So if I could put Lloyd into my array like this I'm pretty good to go, right? What's missing? 

Remember that every string we work with in C ends with backslash zero, and we can't omit that here, either. We need to set aside one byte of memory to hold that so we know when our string has ended. So again this arrangement of the way things appear in memory might be a little random, but it actually is how most systems are designed. To line them up on multiples of four, for reasons again that we don't need to get into right now. But this, so suffice it to say that after these three lines of code, this is what memory might look like. If I need memory locations 4, 8, and 12 to hold my data, this is what my memory might look like. 

And just be particularly pedantic here, when we're talking about memory addresses we usually do so using hexadecimal notations. So why don't we convert all of these from decimal to hexadecimal notation just because that's generally how we refer to memory. So instead of being 0 through 19, what we have is zero x zero through zero x1 three. Those are the 20 bytes of memory that we have or we're looking at in this image right here. 

So all of that being said, let's step away from memory for a second and back to pointers. Here is the most important thing to remember as we start working with pointers. A pointer is nothing more than an address. I'll say it again because it's that important, a pointer is nothing more than an address. Pointers are addresses to locations in memory where variables live. Knowing that it becomes hopefully a little bit easier to work with them. Another thing I like to do is to have sort of diagrams visually representing what's happening with various lines of code. And we'll do this a couple of times in pointers, and when we talk about dynamic memory allocation as well. Because I think that these diagrams can be particularly helpful. 

So if I say for example, int k in my code, what is happening? Well what's basically happening is I'm getting memory set aside for me, but I don't even like to think about it like that, I like to think about it like a box. I have a box and it's colored green because I can put integers in green boxes. If it was a character I might have a blue box. But I always say, if I'm creating a box that can hold integers that box is colored green. And I take a permanent marker and I write k on the side of it. So I have a box called k, into which I can put integers. So when I say int k, that's what happens in my head. If I say k equals five, what am I doing? Well, I'm putting five in the box, right. This is pretty straightforward, if I say int k, create a box called k. If I say k equals 5, put five into the box. Hopefully that's not too much of a leap. Here's where things go a little interesting though. If I say int*pk, well even if I don't know what this necessarily means, it's clearly got something to do with an integer. So I'm going to color this box green-ish, I know it's got something to do with an integer, but it's not an integer itself, because it's an int star. There's something slightly different about it. So an integer's involved, but otherwise it's not too different from what we were talking about. It's a box, its got a label, it's wearing a label pk, and it's capable of holding int stars, whatever those are. They have something to do with integers, clearly. Here's the last line though. If I say pk=&k, whoa, what just happened, right? So this random number, seemingly random number, gets thrown into the box there. All that is, is pk gets the address of k. So I'm sticking where k lives in memory, its address, the address of its bytes. All I'm doing is I'm saying that value is what I'm going to put inside of my box called pk. And because these things are pointers, and because looking at a string like zero x eight zero c seven four eight two zero is probably not very meaningful. When we generally visualize pointers, we actually do so as pointers. Pk gives us the information we need to find k in memory. So basically pk has an arrow in it. And if we walk the length of that arrow, imagine it's something you can walk on, if we walk along the length of the arrow, at the very tip of that arrow, we will find the location in memory where k lives. And that's really important because once we know where k lives, we can start to work with the data inside of that memory location. Though we're getting a teeny bit ahead of ourselves for now. 

So what is a pointer? A pointer is a data item whose value is a memory address. That was that zero x eight zero stuff going on, that was a memory address. That was a location in memory. And the type of a pointer describes the kind of data you'll find at that memory address. So there's the int star part right. If I follow that arrow, it's going to lead me to a location. And that location, what I will find there in my example, is a green colored box. It's an integer, that's what I will find if I go to that address. The data type of a pointer describes what you will find at that memory address. So here's the really cool thing though. Pointers allow us to pass variables between functions. And actually pass variables and not pass copies of them. Because if we know exactly where in memory to find a variable, we don't need to make a copy of it, we can just go to that location and work with that variable. So in essence pointers sort of make a computer environment a lot more like the real world, right. 

So here's an analogy. Let's say that I have a notebook, right, and it's full of notes. And I would like you to update it. You are a function that updates notes, right. In the way we've been working so far, what happens is you will take my notebook, you'll go to the copy store, you'll make a Xerox copy of every page of the notebook. You'll leave my notebook back on my desk when you're done, you'll go and cross out things in my notebook that are out of date or wrong, and then you'll pass back to me the stack of Xerox pages that is a replica of my notebook with the changes that you've made to it. And at that point, it's up to me as the calling function, as the caller, to decide to take your notes and integrate them back into my notebook. So there's a lot of steps involved here, right. Like wouldn't it be better if I just say, hey, can you update my notebook for me, hand you my notebook, and you take things and literally cross them out and update my notes in my notebook. And then give me my notebook back. That's kind of what pointers allow us to do, they make this environment a lot more like how we operate in reality. 

All right so that's what a pointer is, let's talk about how pointers work in C, and how we can start to work with them. So there's a very simple pointer in C called the null pointer. The null pointer points to nothing. This probably seems like it's actually not a very useful thing, but as we'll see a little later on, the fact that this null pointer exists actually really can come in handy. And whenever you create a pointer, and you don't set its value immediately- an example of setting its value immediately will be a couple slides back where I said pk equals & k, pk gets k's address, as we'll see what that means, we'll see how to code that shortly- if we don't set its value to something meaningful immediately, you should always set your pointer to point to null. You should set it to point to nothing. 

That's very different than just leaving the value as it is and then declaring a pointer and just assuming it's null because that's rarely true. So you should always set the value of a pointer to null if you don't set its value to something meaningful immediately. You can check whether a pointer's value is null using the equality operator (==), just like you compare any integer values or character values using (==) as well. It's a special sort of constant value that you can use to test. So that was a very simple pointer, the null pointer. Another way to create a pointer is to extract the address of a variable you've already created, and you do this using the & operator address extraction. Which we've already seen previously in the first diagram example I showed. So if x is a variable that we've already created of type integer, then &x is a pointer to an integer. &x is- remember, & is going to extract the address of the thing on the right. And since a pointer is just an address, than &x is a pointer to an integer whose value is where in memory x lives. It's x's address. So &x is the address of x. Let's take this one step further and connect to something I alluded to in a prior video. If arr is an array of doubles, then &arr square bracket i is a pointer to a double. OK. arr square bracket i, if arr is an array of doubles, then arr square bracket i is the i-th element of that array, and &arr square bracket i is where in memory the i-th element of arr exists. 

So what's the implication here? An arrays name, the implication of this whole thing, is that an array's name is actually itself a pointer. You've been working with pointers all along every time that you've used an array. Remember from the example on variable scope, near the end of the video I present an example where we have a function called set int and a function called set array. And your challenge to determine whether or not, or what the values that we printed out the end of the function, at the end of the main program. 

If you recall from that example or if you've watched the video, you know that when you- the call to set int effectively does nothing. But the call to set array does. And I sort of glossed over why that was the case at the time. I just said, well it's an array, it's special, you know, there's a reason. The reason is that an array's name is really just a pointer, and there's this special square bracket syntax that make things a lot nicer to work with. And they make the idea of a pointer a lot less intimidating, and that's why they're sort of presented in that way. But really arrays are just pointers. And that's why when we made a change to the array, when we passed an array as a parameter to a function or as an argument to a function, the contents of the array actually changed in both the callee and in the caller. Which for every other kind of variable we saw was not the case. So that's just something to keep in mind when you're working with pointers, is that the name of an array actually a pointer to the first element of that array. 

OK so now we have all these facts, let's keep going, right. Why do we care about where something lives. Well like I said, it's pretty useful to know where something lives so you can go there and change it. Work with it and actually have the thing that you want to do to that variable take effect, and not take effect on some copy of it. This is called dereferencing. We go to the reference and we change the value there. So if we have a pointer and it's called pc, and it points to a character, then we can say *pc and *pc is the name of what we'll find if we go to the address pc. What we'll find there is a character and *pc is how we refer to the data at that location. So we could say something like *pc=D or something like that, and that means that whatever was at memory address pc, whatever character was previously there, is now D, if we say *pc=D. 

So here we go again with some weird C stuff, right. So we've seen * previously as being somehow part of the data type, and now it's being used in a slightly different context to access the data at a location. I know it's a little confusing and that's actually part of this whole like, why pointers have this mythology around them as being so complex, is kind of a syntax problem, honestly. But * is used in both contexts, both as part of the type name, and we'll see a little later something else, too. And right now is the dereference operator. So it goes to the reference, it accesses the data at the location of the pointer, and allows you to manipulate it at will. 

Now this is very similar to visiting your neighbor, right. If you know what your neighbor lives, you're not hanging out with your neighbor. You know you happen to know where they live, but that doesn't mean that by virtue of having that knowledge you are interacting with them. If you want to interact with them, you have to go to their house, you have to go to where they live. And once you do that, then you can interact with them just like you'd want to. And similarly with variables, you need to go to their address if you want to interact them, you can't just know the address. And the way you go to the address is to use *, the dereference operator. What do you think happens if we try and dereference a pointer whose value is null? Recall that the null pointer points to nothing. So if you try and dereference nothing or go to an address nothing, what do you think happens? Well if you guessed segmentation fault, you'd be right. If you try and dereference a null pointer, you suffer a segmentation fault. But wait, didn't I tell you, that if you're not going to set your value of your pointer to something meaningful, you should set to null? I did and actually the segmentation fault is kind of a good behavior. 

Have you ever declared a variable and not assigned its value immediately? So you just say int x; you don't actually assign it to anything and then later on in your code, you print out the value of x, having still not assigned it to anything. Frequently you'll get zero, but sometimes you might get some random number, and you have no idea where it came from. Similarly can things happen with pointers. When you declare a pointer int*pk for example, and you don't assign it to a value, you get four bytes for memory. Whatever four bytes of memory the system can find that have some meaningful value. And there might have been something already there that is no longer needed by another function, so you just have whatever data was there. 

What if you tried to do dereference some address that you don't- there were already bytes and information in there, that's now in your pointer. If you try and dereference that pointer, you might be messing with some memory that you didn't intend to mess with it all. And in fact you could do something really devastating, like break another program, or break another function, or do something malicious that you didn't intend to do at all. And so that's why it's actually a good idea to set your pointers to null if you don't set them to something meaningful. It's probably better at the end of the day for your program to crash then for it to do something that screws up another program or another function. That behavior is probably even less ideal than just crashing. And so that's why it's actually a good habit to get into to set your pointers to null if you don't set them to a meaningful value immediately, a value that you know and that you can safely the dereference. 

So let's come back now and take a look at the overall syntax of the situation. If I say int *p;, what have I just done? What I've done is this. I know the value of p is an address because all pointers are just addresses. I can dereference p using the * operator. In this context here, at the very top recall the * is part of the type. Int * is the data type. But I can dereference p using the * operator, and if I do so, if I go to that address, what will I find at that address? I will find an integer. So int*p is basically saying, p is an address. I can dereference p and if I do, I will find an integer at that memory location. 

OK so I said there was another annoying thing with stars and here's where that annoying thing with stars is. Have you ever tried to declare multiple variables of the same type on the same line of code? So for a second, pretend that the line, the code I actually have there in green isn't there and it just says int x,y,z;. What that would do is actually create three integer variables for you, one called x, one called y, and one called z. It's a way to do it without having to split onto three lines. 

Here's where stars get annoying again though, because the * is actually part of both the type name and part of the variable name. And so if I say int *px,py,pz, what I actually get is a pointer to an integer called px and two integers, py and pz. And that's probably not what we want, that's not good. 

So if I want to create multiple pointers on the same line, of the same type, and stars, what I actually need to do is say int *pa,*pb,*pc. Now having just said that and now telling you this, you probably will never do this. And it's probably a good thing honestly, because you might inadvertently omit a star, something like that. It's probably best to maybe declare pointers on individual lines, but it's just another one of those annoying syntax things with stars that make pointers so difficult to work with. Because it's just this syntactic mess you have to work through. With practice it does really become second nature. I still make mistakes with it still after programming for 10 years, so don't be upset if something happens to you, it's pretty common honestly. It's really kind of a flaw of the syntax. 

OK so I kind of promised that we would revisit the concept of how large is a string. Well if I told you that a string, we've really kind of been lying to you the whole time. There's no data type called string, and in fact I mentioned this in one of our earliest videos on data types, that string was a data type that was created for you in CS50.h. You have to #include CS50.h in order to use it. 

Well string is really just an alias for something called the char *, a pointer to a character. Well pointers, recall, are just addresses. So what is the size in bytes of a string? Well it's four or eight. And the reason I say four or eight is because it actually depends on the system, If you're using CS50 ide, char * is the size of a char * is eight, it's a 64-bit system. Every address in memory is 64 bits long. If you're using CS50 appliance or using any 32-bit machine, and you've heard that term 32-bit machine, what is a 32-bit machine? Well it just means that every address in memory is 32 bits long. And so 32 bits is four bytes. So a char * is four or eight bytes depending on your system. And indeed any data types, and a pointer to any data type, since all pointers are just addresses, are four or eight bytes. So let's revisit this diagram and let's conclude this video with a little exercise here. So here's the diagram we left off with at the very beginning of the video. So what happens now if I say *pk=35? So what does it mean when I say, *pk=35? Take a second. *pk. In context here, * is dereference operator. So when the dereference operator is used, we go to the address pointed to by pk, and we change what we find. So *pk=35 effectively does this to the picture. So it's basically syntactically identical to of having said k=35. 

One more. If I say int m, I create a new variable called m. A new box, it's a green box because it's going to hold an integer, and it's labeled m. If I say m=4, I put an integer into that box. If say pk=&m, how does this diagram change? Pk=&m, do you recall what the & operator does or is called? Remember that & some variable name is the address of a variable name. So what we're saying is pk gets the address of m. And so effectively what happens the diagram is that pk no longer points to k, but points to m. 

Again pointers are very tricky to work with and they take a lot of practice, but because of their ability to allow you to pass data between functions and actually have those changes take effect, getting your head around is really important. It probably is the most complicated topic we discuss in CS50, but the value that you get from using pointers far outweighs the complications that come from learning them. So I wish you the best of luck learning about pointers. I'm Doug Lloyd, this is CS50.