2023-09-04
go backSo, you're an ECE150 student at the University of Waterloo. Maybe you have a bit of experience programming, maybe not. Regardless, prepare yourself for some bad programming habits. Unfortunately, the course seems somewhat rife with opportunity to confuse yourself and end up with some pretty bad coding mistakes. Worse, even if you manage to do decently in the course, you may still completely misunderstand how it works.
This leads to cargo cult programming, where we worship the idea of setting a
pointer in C++ to nullptr after deleteing it, without really ever knowing
why.
The need for this struck me after I found myself trying to clear up what
delete in C++ actually does and what memory "safety" actually means.
You'll certainly hear about "memory leaks" or "segfaults" in programming courses, but what do they actually refer to? And, more importantly, why are they actually a problem?
Before we start, we need to get a few things down. First we need to understand what the heap and the stack are, what "scopes" are in programming, and then what
Well, here you go again. Sorry to bother you with another explanation of something that's quite confusing.
Basically, the stack is a place to temporarily store things. Whenever you declare a normal variable, you use the stack. The stack exists for each different "scope" that you go to.
Woah there, a scope? What's a scope?
Here's a quick C example that I'll annotate. Notice that it could be compiled (turned into computer code) into C++. This is because C++ is mostly C plus other things.
#include <stdio.h>
void g(void) {
int a = 6;
}
int main(void) {
int a = 4;
int b = 3;
g();
printf("%d", a); // this will print 4
return 0;
}
Okay! Let's go through this example.
First we declare that we're including stdio.h. All you need to know is that
that's the code where printf() is held. We can only print if we include this
file. You might ask: why would I need to include something in order to use a
print function? Especially if you've started with Python, where print() is a
piece of the base language. Basically, it's because C is a simple language.
This means it often runs much faster, because it doesn't need to be prepared to
do as much as something like Python does.
Next you'll notice the void keyword. It means that the function g doesn't
take anything and doesn't return anything either. Sometimes you'll see people
just write void g(), but technically this isn't good practice in C, because
it interprets it as "you can give me as many arguments as you want!". Try it
yourself!
From there, we declare an integer variable a and set it equal to four, declare
an integer b equal to 3, call the g() function, and then print a out.
When we call g(), we go into the scope of the function g(). The one rule
about this is that we can access anything in a "higher" scope than the current
one. We don't in my example, but we could totally write printf("%d", b); and
have it print out 3. This is because we have all the declared variables from
earlier here. This is a really important concept (and one that will motivate
this post later). We can access anything before it, but writing things inside
it have no effect when we exit the function call. As soon as we reach the
end of g(), we go back to main() and lose everything we have inside this
function.
Okay, wait, we lose it all? I thought we could use functions to do things for us and then hand it back?
Yes, yes you can. But not for everything! Only specific things. Some things can
be copied trivially. For example, an integer can be returned at the end of a
function, but something like an array cannot (in C and C++). This is because an
array is a block of memory; not just one individual piece.
Now that we have lexical scopes down, let's talk about the heap.
The heap is the place where we can place things that we want to live until we tell them not to. Basically,
Unfortunately, ECE150 (and from what I know, other courses) don't often put much emphasis on what memory usage and safety in C++ actually consists of. I'll start with describing (for a novice, not someone more experienced) how memory works in C, which is a more simple language than C++. If you hear people complain about C++, it's often because it has a lot of features. Some of these features are good, some of these features are confusing, but most of the features have difficulties with compatibility between them.
In C, we use a function called malloc to access memory on the heap.
#include <stdio.h>
#include <stdlib.h>
int main(void) {
int* array = (int*)malloc(5*sizeof(int));
array[0] = 0;
array[1] = 1;
array[2] = 2;
array[3] = 342;
array[4] = 23;
for (int i = 0; i < 5; i++) {
printf("%d", array[i]);
}
array[5] = 43; // this is a memory error!
return 0;
}
Inside of malloc, we pass 5 times the "sizeof" an integer in C. This allows us
to specify how much we want to store in the memory. malloc will actually take
just the number of bytes you want to store and then give you a pointer to that
block of memory.
The (int *) tells malloc that we want it to refer to a pointer to an int.
printfWhat's that weird %d I see in there? What does that mean?
Glad you asked! It's a "character code", which tells printf how you want to
display the data. Keep in mind that things like the characters that a computer
uses to print out "Hello, World!" are actually just numbers themselves.
(They're called ASCII if you use a simple latin character set without accents or other things.) This means that
sometimes we want to see what number the character corresponds to and therefore
need to tell the function whether we want to see it as a character or as a
number. In this case, we're asking for %d, which means "show me this data as
an integer number". Another example would be %f, which means "show me this
data as a floating point number (a number with decimal point like 5.4)".
new do?Well, when do I need dynamically allocated things? Great question!
Dynamically allocated memory is useful when dealing with big objects (a large array), that you can't actually copy between places. Theoretically you can just treat it as being on the stack, but this is quite suboptimal due to the size of the stack.
Another situation would be for resizing arrays. With statically allocated arrays, such as this char arr[] = "Hello";, you cannot resize it even if you want to. However, using dynamic allocation, you can resize using realloc, a function that finds you a new block of memory to store everything and puts the old contents of the memory there.
Keep in mind that some programming paradigms avoid using the heap (where dynamic memory goes) because of the concern that dynamic allocation is much more likely to cause issues when programming mistakes are made than static allocation.
You probably won't need to worry about that style, but NASA does tend to use it.
delete actually do?Delete in C++ calls the destructor of an object. Sometimes these objects are built in and sometimes they're defined by the programmer. Usually calling delete on them just means that there are objects inside them that need to be freed them.
Note that if you do not delete your memory, it will be leaked from your process. Due to how operating systems work, once the process (the executable) is finished, the operating system will clean it up for you, but on microcontrollers, it is very likely to just crash if you don't clean it up yourself.
Also, if you try to allocate too much at the same time and don't delete any objects you no longer need, you can trivially get yourself an "OOM" error. (Out of memory.) Believe me, it's not good and it would make the computer scientists of yesteryear cry.
null and nullptrhttps://stackoverflow.com/questions/1282295/what-exactly-is-nullptr
This is a really useful link for learning the complications with NULL and nullptr. Essentially, nullptr is a whole different type, whereas NULL is just a macro for 0. This means you can use NULL as an argument for a function that takes in an integer. Essentially, always opt to use nullptr in C++, since it is "type-safe". In C, it is generally more traditional to use NULL, even though NULL is not actually a whole different type.