if-else or just declaration-if in conditional value

Question

Suppose I have a variable that depends on a condition. From the efficiency perspective, should I use

int s;
if (d > 2)
{
   s = -1;
}
else
{
   s = 1;
}

or just

int s = 1;
if (d > 2)
{
   s = -1;
}

Can someone explain the difference? Is there any difference between a compiled language (e.g. C) vs an interpreted one (e.g. Python)?

Notice that this question is related with a previous question of mine, and the reason I'm asking is the same: it is very frequent in any programming language, and I always end up asking myself what should I use.

It completely depends on the compiler/interpreter. It's more important to use the one that makes the most sense when you're reading the code. If your program is too slow because of this type of syntactical decision, don't worry about the syntax, worry about the design. — xaxxon, Jun 19 '13 at 07:54
try to generate the assembly code and see the difference. you can generate the assembly code from C code with `gcc -S` — MOHAMED, Jun 19 '13 at 07:54
Odd that you need to ask it again despite various responses that pretty much answer what you're asking now. — devnull, Jun 19 '13 at 08:05
possible duplicate of [If-Else-Return or just if-Return?](http://stackoverflow.com/questions/9191388/if-else-return-or-just-if-return) — devnull, Jun 19 '13 at 08:05
@devnull, I'm sorry, for me was not clear that the answers in that question also answer this one. Nevetheless, I'm probability not so experienced as you in programming, so feel free to close this one if you don't agree with it. — Jorge Leitao, Jun 19 '13 at 08:08

David Ranieri · Answer 1 · 2013-06-19T08:08:09.287

1

No difference, with optimization gcc -S -O2 you give the same output:

    .file   "demo.c"
    .section    .rodata.str1.1,"aMS",@progbits,1
.LC0:
    .string "%d"
    .section    .text.startup,"ax",@progbits
    .p2align 4,,15
    .globl  main
    .type   main, @function
main:
.LFB11:
    .cfi_startproc
    subq    $8, %rsp
    .cfi_def_cfa_offset 16
    movl    $-1, %esi
    movl    $.LC0, %edi
    xorl    %eax, %eax
    call    printf
    xorl    %eax, %eax
    addq    $8, %rsp
    .cfi_def_cfa_offset 8
    ret
    .cfi_endproc
.LFE11:
    .size   main, .-main
    .ident  "GCC: (Debian 4.7.2-5) 4.7.2"
    .section    .note.GNU-stack,"",@progbits

edited Jun 19 '13 at 08:08

answered Jun 19 '13 at 07:57

David Ranieri

39,972
7
52
94

You're right unwind, edited, printing value also gives same output – David Ranieri Jun 19 '13 at 08:08
This assembler code is not really useful without the source code, especially showing whether `d` is initialized. – Armali Jun 20 '13 at 07:00
`d` is intialized as `int d = 5;` source code is the one posted by OP with a `printf("%d", s);` at the end – David Ranieri Jun 20 '13 at 07:52

jxh · Answer 2 · 2013-06-19T09:14:12.840

1

If you really have tons of code like this where you are initializing to 1 or -1, and it really gets run tons of times so it can affect performance, you can use a branchless version instead:

int s = 1 - 2*(d > 2);

More generally, you could set up an array for initialization values:

const int s_ini[2] = { 1, -1 };
int s = s_ini[d > 2];

edited Jun 19 '13 at 09:14

answered Jun 19 '13 at 08:09

jxh

69,070
8
110
193

Any reason you made `s_ini` `static`? My gut is telling me that this way might sometimes inhibit it being optimized away. – Benjamin Bannier Jun 19 '13 at 09:11

score 1 · Answer 3 · answered Jun 19 '13 at 09:01

This is a case where I personally would prefer the (sometimes dreaded) ternary operator ?:, like so:

const int s = (d < 2) ? -1 : 1;

Since:

This allows s to be const, which in my opinion is a huge win. Of course this assumes the variable is only read after being initialized.
It (like other solutions) only writes to s once, which can be faster.

I avoided "clever tricks" with with the computation.

Sylvain Leroux · Answer 4 · 2013-06-20T07:23:32.827

From my point of view, this is not a question of performance. Most compiler will produce the same code, and if they don't, that shouldn't have any noticeable performance impact. Unless you are working on very very (very) tight environment where each CPU cycle count. Which shouldn't be the case, unless you are a time traveler just arriving from the seventies'...

More seriously, as it was noticed several times in the various comments, this is more a matter of readability and communication. After that, it is hard to say that one form is better than the other.

My personal habit is to use if ... else ... when I want to put emphasis on the fact there is clearly two different path of equal "weight" in the program. I use the alternative set to default; if ... when I want to say "Well, that if is a special case".

Since this is related to an other of your questions, you might consider the form without else as some kind of guard clause, like a if ... return ...; return ...

score 0 · Answer 5 · answered Jun 19 '13 at 07:57

The performance difference isn't really important, but your first notation is most clear for others programmers.

int s;
if (d > 2)
   s = -1;
else
   s = 1;

When a programmer will se "int s = 1;" he will probably don't understand why you initialize this variable at 1, for put it a -1 two lignes after.

score 0 · Answer 6 · answered Jun 19 '13 at 08:01

There is no right and wrong in such cases. It depends. I believe that even if there might be a slight difference in run time, you should almost always prefer code maintainability.

In this case a default is better as you must pass all the conditions to return successfully.

    int rc = FAIL;

    /* do something */

    if (some-cond)
        goto end;

    /* do something */

    if (some-cond)
        goto end;

    rc = SUCCESS;

end:
    /* clean-up */
    return rc;

score 0 · Answer 7 · answered Jun 19 '13 at 08:44

0

From the readability of code, I think the former one is better. If you want to get better performance, I think you can do as @jxh answers.

answered Jun 19 '13 at 08:44

Charles0429

1,406
5
15
31

if-else or just declaration-if in conditional value

7 Answers7