This question already has answers here:
Why are these constructs using pre and post-increment undefined behavior?
(14 answers)
Closed 4 years ago.
can someone explain to me why this code prints the decremented amount? Since (4 <= 3) is false, shouldn't it proceed to print "Hi"?
int main()
{
int amount = 4;
if((amount%5) <= (--amount))
printf("%d", amount);
else
printf("Hi");
}
Turns out this code is undefined. You've got the expression
(amount%5) <= (--amount)
But the key question is, does the amount%5 part use the old value of amount, or the new, after being decremented by the --amount part? And the answer is: We have no idea. There's no way to tell. There's no rule in the C language to tell us.
You might think the code has to be evaluated left-to-right, that the amount%5 part has to happen first, before the --amount part has a chance to happen. But that's not the way it works. Since it's undefined, anything could happen.
If it's evaluated more or less left-to-right, then we have
if(amount%5 <= --amount)
which evaluates to
if(4%5 <= --amount)
which evaluates to
if(4 <= --amount)
which evaluates to
if(4 <= 3)
which is false, so you might expect it to print "Hi". But, if it's evaluated more or less right-to-left, it's
if(amount%5 <= 3)
which evaluates to
if(3%5 <= 3)
which evaluates to
if(3 <= 3)
which is true.
(You'll notice that I have removed some extra parentheses, since they don't do anything.)
I hasten to add that these are not the only two possibilities. The rules say that once an expression is undefined, literally anything can happen -- the behavior doesn't have to be anything we might consider "reasonable" at all.
So why is it undefined? In this case it's because you have a variable -- amount -- which appears twice in the expression, and where one of those appearances is a modification (the --amount part), and where the other one uses the variable's value. The C Standard explicitly tells us that in this case (a) we can't tell whether the use of amount uses the old or the modified value, and furthermore that (b) this flaw in the expression -- this simultaneous modification and use -- renders the whole expression (and actually the whole program it's in) undefined.
So how do you fix it? The answer is simple, and it's the punchline of an old joke: Don't do that. Don't write expressions where a single variable (or other object) is both modified and used. It's fine to write --amount, just don't do it in an expression that's also using the value of the variable amount somewhere else.
Questions like this come up often. The canonical SO answers, which others are already redirecting you to, are collected under the question Why are these constructs (using ++) undefined behavior in C?. There's more I could say about the rules concerning undefined expressions, but you can find those details over at that question.
Related
I am getting an output of 24 which is the factorial for 4, but I should be getting the output for 5 factorial which is 120
#include <stdio.h>
int factorial(int number){
if(number==1){
return number;
}
return number*factorial(--number);
}
int main(){
int a=factorial(5);
printf("%d",a);
}
Your program suffers from undefined behavior.
In the first call to factorial(5), where you have
return number * factorial(--number);
you imagine that this is going to compute
5 * factorial(4);
But that's not guaranteed!
What if the compiler looks at it in a different order?
What it if works on the right-hand side first?
What if it first does the equivalent of:
temporary_result = factorial(--number);
and then does the multiplication:
return number * temporary_result;
If the compiler does it in that order, then temporary_result will be factorial(4), and it'll return 4 times that, which won't be 5!. Basically, if the compiler does it in that order -- and it might! -- then number gets decremented "too soon".
You might not have imagined that the compiler could do things this way.
You might have imagined that the expression would always be "parsed left to right".
But those imaginations are not correct.
(See also this answer for more discussion on order of evaluation.)
I said that the expression causes "undefined behavior", and this expression is a classic example. What makes this expression undefined is that there's a little too much going on inside it.
The problem with the expression
return number * factorial(--number);
is that the variable number is having its value used within it, and that same variable number is also being modified within it. And this pattern is, basically, poison.
Let's label the two spots where number appears, so that we can talk about them very clearly:
return number * factorial(--number);
/* A */ /* B */
At spot A we take the value of the variable number.
At spot B we modify the value of the variable number.
But the question is, at spot A, do we get the "old" or the "new" value of number?
Do we get it before or after spot B has modified it?
And the answer, as I already said, is: we don't know. There is no rule in C to tell us.
Again, you might have thought there was a rule about left-to-right evaluation, but there isn't. Because there's no rule that says how an expression like this should be parsed, a compiler can do anything it wants. It can parse it the "right" way, or the "wrong" way, or it can do something even more bizarre and unexpected. (And, really, there's no "right" or "wrong" way to parse an undefined expression like this in the first place.)
The solution to this problem is: Don't do that!
Don't write expressions where one variable (like number) is both used and modified.
In this case, as you've already discovered, there's a simple fix:
return number * factorial(number - 1);
Now, we're not actually trying to modify the value of the variable number (as the expression --number did), we're just subtracting 1 from it before passing the smaller value off to the recursive call.
So now, we're not breaking the rule, we're not using and modifying number in the same expression.
We're just using its value twice, and that's fine.
For more (much more!) on the subject of undefined behavior in expressions like these, see Why are these constructs using pre and post-increment undefined behavior?
How to find the factorial of a number;
function factorial(n) {
if(n == 0 || n == 1 ) {
return 1;
}else {
return n * factorial(n-1);
}
//return newnum;
}
console.log(factorial(3))
My question pertains to function calls in general, but I thought of it
while I was writing a priority queue using a heap. Just to give some context (not that it matters much) my heap stores items top to bottom left to right and I represent the heap as an array of structures. Upon inserting a new item, I just put it in the last place in the heap and then call the function "fix_up" at the bottom which will move the item to the proper place in the heap. I am wondering if instead of doing...
fix_up(pQueue->heap, pQueue->size);
pQueue->size++;
...I could just do...
fix_up(pQueue->heap, pQueue->size++);
I am unsure as to if this is ok for a few reasons.
1) Since pQueue->size is in the function call, I'm not even sure if it's actually pQueue->size or rather a copy of the integer stored in pQueue->size. If it was a copy then obviously I wouldn't be adding 1 to the actual pQueue->size so there'd be no point in doing this.
2) Since it's a function call, it is going to then go into the function fix_up and execute all the code there. I am wondering if this would have an unintended consequence of maybe when it went to fix_up it would get incremented by 1 and my index would be 1 higher than I intended while executing fix_up? Or would it do what it's supposed to do and wait until after fix_up had finished executing?
3) Even if it is ok, is it considered a good coding practice for C?
Status priority_queue_insert(PRIORITY_QUEUE hQueue, int priority_level, int data_item)
{
Priority_queue *pQueue = (Priority_queue*)hQueue;
Item *temp_heap;
int i;
/*Resize if necessary*/
if (pQueue->size >= pQueue->capacity) {
temp_heap = (Item*)malloc(sizeof(Item) * pQueue->capacity * 2);
if (temp_heap == NULL)
return FAILURE;
for (i = 0; i < pQueue->size; i++)
temp_heap[i] = pQueue->heap[i];
pQueue->capacity *= 2;
}
/*Either resizing was not necessary or it successfully resized*/
pQueue->heap[pQueue->size].key = priority_level;
pQueue->heap[pQueue->size].data = data_item;
/*Now it is placed as the last item in the heap. Fixup as necessary*/
fix_up(pQueue->heap, pQueue->size);
pQueue->size++;
//continue writing function code here
}
Yes you can.
However, you cannot do this:
foo(myStruct->size++, myStruct->size)
The reason is that the C standard does not say in which order the arguments should be evaluated. This would lead to undefined behavior.
1) Since pQueue->size is in the function call, I'm not even sure if it's actually pQueue->size or rather a copy of the integer stored in pQueue->size. If it was a copy then obviously I wouldn't be adding 1 to the actual pQueue->size so there'd be no point in doing this.
Whatever argument you're sending to a function, it will be evaluated before the function starts to execute. So
T var = expr;
foo(var);
is always equivalent to
foo(expr);
2) Since it's a function call, it is going to then go into the function fix_up and execute all the code there. I am wondering if this would have an unintended consequence of maybe when it went to fix_up it would get incremented by 1 and my index would be 1 higher than I intended while executing fix_up? Or would it do what it's supposed to do and wait until after fix_up had finished executing?
See above
3) Even if it is ok, is it considered a good coding practice for C?
Somewhat subjective, and a bit OT for this site, but I'll answer it anyway from my personal view. In general, I would try to avoid it.
Though, the other posts already answer this question, but none of them talk about role of Sequence Point, in this particular case, which can greatly help in clarifying OP's doubt.
From this [emphasis mine]:
There is a sequence point after the evaluation of all function arguments and of the function designator, and before the actual function call.
From this [emphasis mine]:
Increment operators initiate the side-effect of adding the value 1 of appropriate type to the operand. Decrement operators initiate the side-effect of subtracting the value 1 of appropriate type from the operand. As with any other side-effects, these operations complete at or before the next sequence point.
Also, the post increment operator increase the value of operand by 1 but the value of the expression is the operand's original value prior to the increment operation.
So, in this statement:
fix_up(pQueue->heap, pQueue->size++);
the value of pQueue->size will be increased by 1 before the fix_up() function call but the argument value will be the original value prior to the increment operation.
Yes you can use it directly in the expression you pass as argument.
A statement like
fix_up(pQueue->heap, pQueue->size++);
is somewhat equivalent to
{
int old_value = pQueue->size;
pQueue->size = pQueue->size + 1;
fix_up(pQueue->heap, old_value);
}
A note about the "equivalent" example above. Since the order of evaluation of arguments to function calls is not specified, the actual increment of pQueue->size could happen after the call to fix_up. And it also means that using pQueue->size more than once in the same call would lead to undefined behavior.
Yeah you can use it in function calls, but please note that your two examples are not equivalent. The pQueue->heap argument may be evaluated before or after pQueue->size++ and you can't know or rely on the order. Consider this example :
int func (void)
{
static int x = 0;
x++;
return x;
}
printf("%d %d", func(), func());
This will print 1 2 or 2 1 and we can't know which we'll get. The compiler need not evalute the function parameters consistently throughout the program. So if we add a second printf("%d %d", func(), func()); we could get something like 1 2 4 3 as output.
The importance here is to not write code which relies on order of evaluation. Which is the same reason as mixing ++ with other operations or side-effects in the same expression is bad practice. It can even lead to undefined behavior in some cases.
To answer your questions:
1) Since pQueue->size is in the function call, I'm not even sure if it's actually pQueue->size or rather a copy of the integer stored in pQueue->size. If it was a copy then obviously I wouldn't be adding 1 to the actual pQueue->size so there'd be no point in doing this.
The ++ is applied to the variable in the caller, so this isn't an issue. The local copy of the variable happens during function call, independently of the ++. However, the result of a ++ operation is not a so-called "lvalue" (addressable data), so this code is not valid:
void func (int* a);
...
func(&x++);
++ takes precedence and is evaluted first. The result is not an lvalue and cannot have its address taken.
2) Since it's a function call, it is going to then go into the function fix_up and execute all the code there. I am wondering if this would have an unintended consequence of maybe when it went to fix_up it would get incremented by 1 and my index would be 1 higher than I intended while executing fix_up? Or would it do what it's supposed to do and wait until after fix_up had finished executing?
This isn't an issue unless the function modifies the original variable through a global pointer or such. In that case you would have problems. For example
int* ptr;
void func (int a)
{
*ptr = 1;
}
int x=5;
ptr = &x;
func(x++);
This is very questionable code and x will be 1 after the line func(x++); and not 6 as one might have expected. This is because the function call expression is evaluated and finished before the function call.
3) Even if it is ok, is it considered a good coding practice for C?
It will work ok in your case but it is bad practice. Specifically, mixing the ++ or -- operators together with other operators in the same expression is bad (although common) practice, since it has a high potential for bugs and tends to make code less readable.
Your original code with pQueue->size++; on a line of it's own is superior in every way - stick with that. Contrary to popular belief, when writing C, you get no bonus points for "most operators on a single line". You may however get bugs and maintenance problems.
I saw this question in a test in which we have to tell the output of the following code.
#include<stdio.h>
int main(){
int k = 0;
while(+(+k--)!=0)
k=k++;
printf("%d\n", k);
return 0;
}
The output is -1. I am unsure why this is the answer, though.
What does the expression +(+k--) mean in C?
This code is deeply, perhaps deliberately, confusing. It contains a narrowly-averted instance of the dread undefined behavior. It's hard to know whether the person who constructed this question was being very, very clever or very, very stupid. And the "lesson" this code might purport to teach or quiz you about -- namely, that the unary plus operator doesn't do much -- is not one that's important enough, I would think, to deserve this kind of subversive misdirection.
There are two confusing aspects of the code, the strange condition:
while(+(+k--)!=0)
and the demented statement it controls:
k=k++;
I'm going to cover the second part first.
If you have a variable like k that you want to increment by 1, C gives you not one, not two, not three, but four different ways to do it:
k = k + 1
k += 1
++k
k++
Despite this bounty (or perhaps because of it), some programmers get confused and cough out contortions like
k = k++;
If you can't figure out what this is supposed to do, don't worry: no one can. This expression contains two different attempts to alter k's value (the k = part, and the k++ part), and because there's no rule in C to say which of the attempted modifications "wins", an expression like this is formally undefined, meaning not only that it has no defined meaning, but that the whole program containing it is suspect.
Now, if you look very carefully, you'll see that in this particular program, the line k = k++ doesn't actually get executed, because (as we're about to see) the controlling condition is initially false, so the loop runs 0 times. So this particular program might not actually be undefined -- but it's still pathologically confusing.
See also these canonical SO answers to all questions concerning Undefined Behavior of this sort.
But you didn't ask about the k=k++ part. You asked about the first confusing part, the +(+k--)!=0 condition. This looks strange, because it is strange. No one would ever write such code in a real program. So there's not much reason to learn how to understand it. (Yes, it's true, exploring the boundaries of a system can help you learn about its fine points, but there's a line in my book between imaginative, thought-provoking explorations versus dunderheaded, abusive explorations, and this expression is pretty clearly on the wrong side of that line.)
Anyway, let's examine +(+k--)!=0. (And after doing so, let's forget all about it.) Any expression like this has to be understood from the inside out. I presume you know what
k--
does. It takes k's current value and "returns" it to the rest of the expression, and it more or less simultaneously decrements k, that is, it stores the quantity k-1 back into k.
But then what does the + do? This is unary plus, not binary plus. It's just like unary minus. You know that binary minus does subtraction: the expression
a - b
subtracts b from a. And you know that unary minus negates things: the expression
-a
gives you the negative of a. What unary + does is... basically nothing. +a gives you a's value, after changing positive values to positive and negative values to negative. So the expression
+k--
gives you whatever k-- gave you, that is, k's old value.
But we're not done, because we have
+(+k--)
This just takes whatever +k-- gave you, and applies unary + to it again. So it gives you whatever +k-- gave you, which was whatever k-- gave you, which was k's old value.
So in the end, the condition
while(+(+k--)!=0)
does exactly the same thing as the much more ordinary condition
while(k-- != 0)
would have done. (It also does the same thing as the even more complicated-looking condition while(+(+(+(+k--)))!=0) would have done. And those parentheses aren't really necessary; it also does the same thing as while(+ +k--!=0) would have done.)
Even figuring out what the "normal" condition
while(k-- != 0)
does is kind of tricky. There are sort of two things going on in this loop: As the loop runs potentially multiple times, we're going to:
keep doing k--, to make k smaller and smaller, but also
keep doing the body of the loop, whatever that does.
But we do the k-- part right away, before (or in the process of) deciding whether to take another trip through the loop. And remember that k-- "returns" the old value of k, before decrementing it. In this program, the initial value of k is 0. So k-- is going to "return" the old value 0, then update k to -1. But then the rest of the condition is != 0 -- and it's not true that 0 != 0. That is, 0 is equal to 0, so we won't make any trips through the loop, so we won't try to execute the problematic statement k=k++ at all.
In other words, in this particular loop, although I said that "there are sort of two things going on", it turns out that thing 1 happens one time, but thing 2 happens zero times.
At any rate, I hope it's now adequately clear why this poor excuse for a program ends up printing -1 as the final value of k. Normally, I don't like to answer quiz questions like this -- it feels like cheating -- but in this case, since I fundamentally disagree with the whole point of the exercise, I don't mind.
At first glance it looks like this code invokes undefined behavior however that is not the case.
First let's format the code correctly:
#include<stdio.h>
int main(){
int k = 0;
while(+(+k--)!=0)
k=k++;
printf("%d\n", k);
return 0;
}
So now we can see that the statement k=k++; is inside of the loop.
Now let's trace the program:
When the loop condition is first evaluated, k has the value 0. The expression k-- has the current value of k, which is 0, and k is decremented as a side effect. So after this statement the value of k is -1.
The leading + on this expression has no effect on the value, so +k-- evaluated to 0 and similarly +(+k--) evaluates to 0.
Then the != operator is evaluated. Since 0!=0 is false, the body of the loop is not entered. Had the body been entered, you would invoke undefined behavior because k=k++ both reads and writes k without a sequence point. But the loop is not entered, so no UB.
Finally the value of k is printed which is -1.
Here's a version of this that shows operator precedence:
+(+(k--))
The two unary + operators don't do anything, so this expression is exactly equivalent to k--. The person that wrote this most likely was trying to mess with your mind.
I will take an exam and my teacher asks weird C syntax rules. Like:
int q=5;
for(q=-2;q=-5;q+=3) { //assignment in condition part??
printf("%d",q); //prints -5
break;
}
Or
int d[][3][2]={4,5,6,7,8,9,10,11,12,13,14,15,16};
int i=-1;
int j;
j=d[i++][++i][++i];
printf("%d",j); //prints 4?? why j=d[0][0][0] ?
Or
extern int a;
int main() {
do {
do {
printf("%o",a); //prints 12
} while(!1);
} while(0);
return 0;
}
int a=10;
I could not find it rules any site or book. Really absurd and uncommon. Where can I find?
To me it seems that your teacher is asking questions which invole undefined behavior.
If you tell him that this is incorrect, you're directly confronting him.
However, you could do the following:
Compile the code on different platforms
Compile the code with different compilers
Compile the code with different versions of the same compiler
Build a matrix with the results. You'll find out that they differ
Show the results to your teacher ans ask him to explain why that happens
That way you do not say that he's wrong, you're just showing some facts and you're showing that you're willing to learn and work.
Do that a long before the exam so that the teacher can look into it and think about his questions so that he can change the exam in time.
I could not find it rules any site or book. Where can I find?
See Where do I find the current C or C++ standard documents?. If you have a good library at university, they should own a copy.
Concerning for(q=-2;q=-5;q+=3) {, all you need to do is to break this down into its components. q=-2 is ran first, then q=-5 is tested, and if that is not 0 (which it isn't since it's an expression with value -5), then the loop body runs once. Then break forces a premature exit from an otherwise infinite loop. The expression then q+=3 is never reached.
The behaviour of d[i++][++i][++i] is undefined. Tell your teacher that, tactfully.
The "%o" format denotes octal output. a is set to 10 in decimal which is 12 in octal. Your code would be clearer if you had written:
int a=012; // octal constant.
The online version of the C language standard has what you need (and is what I will be referring to in this answer); just bear in mind is is a language definition and not a tutorial, and as such may not be easy to read for someone who doesn't have a lot of experience yet.
Having said that, your teacher is throwing you a few foul balls. For example:
j=d[i++][++i][++i];
This statement results in undefined behavior for several reasons. The first several paragraphs of section 6.5 of the document linked above explain the problem, but in a nutshell:
Except in a few situations, C does not guarantee left-to-right evaluation of expressions; neither does it guarantee that side effects are applied immediately after evaluation;
Attempting to modify the value of an object more than once between sequence points1, or modifying and then trying to use the value of an object without an intervening sequence point, results in undefined behavior.
Basically, don't write anything of the form:
x = x++;
x++ * x++;
a[i] = i++;
a[i++] = i;
C does not guarantee that each ++i and i++ is evaluated from left to right, and it does not guarantee that the side effect of each evaluation is applied immediately. So the result of j[i++][++i][++i] is not well-defined, and the result will not be consistent over different programs, or even different builds of the same program2.
AND, on top of that, i++ evaluates to the current value of i; so clearly, your teacher's intent was for j[i++][++i][++i] to evaluate to j[-1][1][2], which would also result in undefined behavior since you're attempting to index outside of the array bounds.
This is why I hate, hate, hate it when teachers throw this kind of code at their students - not only is it needlessly confusing, not only does it encourage bad practice, but more often than not it's just plain wrong.
As for the other questions:
for(q=-2;q=-5;q+=3) { //assignment in condition part??
See sections 6.5.16 and 6.8.5.3. In short, an assignment expression has a value (the value of the left operand after any type conversions), and it can appear as part of a controlling expression in a for loop. As long as the result of the assignment is non-zero (as in the case above), the loop will execute.
printf("%o",a); //prints 12
See section 7.21.6.1. The o conversion specifier tells printf to format the integer value as octal: 1010 == 128
A sequence point is a point in a programs execution where an expression has been fully evaluated and any side effects have been applied. Sequence points occur at the ends of statements, between the evaluation of a function's parameters and the function call, after evaluating the left operand of the &&, ||, and ?: operators, and a few other places. See Annex C for the complete list.
Or even different runs of the same build, although in practice you won't see values change from run to run unless you're doing something really hinky.
This question already has answers here:
Why are these constructs using pre and post-increment undefined behavior?
(14 answers)
Closed 4 years ago.
#include<stdio.h>
main()
{
int a=10;
printf("\n %d %d", a, a++); //11 10
a=10;
printf("\n %d %d", a++, a); //10 11
a=10;
printf("\n %d %d %d ", a, a++,++a); //12 11 12
}
after running this I got the output given in comments. as far as I know first output is expected because execution of printf goes from right to left but could not understand second and third
Nothing goes "from right to left" in function argument evaluation. When function arguments are evaluated, the order of evaluation is unspecified and there are no sequence points between evaluating separate arguments. This means that there's absolutely no temporal ordering in this process. The arguments can be evaluated in any order, and the process of their evaluation can be intertwined in any way.
However, your code suffers from even worse problems. All three statements that call printf produce undefined behavior (UB) because they either make an attempt to modify the same object (a) twice without a sequence point between the modifications (the third call), or they attempt to modify an object and read it for an independent purpose (the first and the second call). So, it is too early to even mention the order of evaluation. Your code's behavior is undefined.
None of the outputs can really qualify as unexpected. All the arguments to a function are evaluated before entry to the function itself -- but the order of their evaluation relative to each other is unspecified, so all of these results are allowed. Officially, your last one (that has two separate instances of incrementing a) has undefined behavior, so it doesn't have to do anything sensible at all.
++a means increment a first, and then return evaluate the expression. (a changes, and the expression evaluates as a + 1)
a++ means evaluate a (so, erm, a), and then increment it. So a is passed, but the value of a is then (ie afterwards) changed to a+1.
You are invoking undefined behaviour by referencing both 'a' and 'a++' in the argument list.
It is not defined which order the arguments are evaluated in. Different compilers may choose different orders. A single compiler can choose different orders at different times.
Do not do it!
Function parameters are not evaluated in a defined order in C. Therefore one cannot tell beforehand if a or a++ will be evaluated first when calling printf.
See Parameter evaluation order before a function calling in C