What is faster: many ifs, or else if?

Question

I'm iterating through an array and sorting it by values into days of the week.

In order to do it I'm using many if statements. Does it make any difference to the processing speed if I use many ifs, versus a set of else if statements?

Answer 1

Yes, use an else if, consider the following code:

if(predicateA){
  //do Stuff
}
if(predicateB){
  // do more stuff
}

of

if(predicateA){
  //
}
else if(predicateB){
  //
}

in the second case if predicateA is true, predicateB (and any further predicates) will not need to be evaluated (and so the whole code will execute faster), whereas in the first example if predicateA is true, predicateB will still always be evaluated, and you may also get some unexpected suprises if predicateA and predicateB are not mutually exclusive.

Answer 2

I doubt that a micro optimization like this will make a measurable difference in your code.

Your sorting algorithm is more likely to be the source of a performance problem. Which sorting algorithm you choose will be critical, not many "ifs" versus "else if".

UPDATE:

The points made by others about "else if" being a better choice, due to its early exit and exclusive logic characteristics, suggest that it should be preferred over "if" in this case.

But the point about algorithm choice still stands - unless your data set is very small.

It's obvious that O(log n) would be better than O(n^2), but size of dataset matters as well. If you have only a few elements, you might not notice the difference. In that case, coding an inefficient method in the cleanest, most readable, most easily understandable at a glance could be your best bet.

Answer 3

You can have a look at phpbench

But to be honest if you want to optimize at this level, you might want to learn something else than php.

Answer 4

To be honest I don't think it would matter which way you do it in terms of performance, I doubt you would see any difference. I would recommend using a switch statement which isn't a performance enhancment, simply syntactically nicer:

switch ($day) 
{
    case "Monday":
        // do something with Monday
        break;
    case "Tuesday":
        // do something with Tuesday
        break;
    case "Wednesday":
        // do something with Wednesday
        break;
}

Answer 5

else if would be faster in the sense that you compare until you hit a condition that resolves to true, and you skip the rest of the ifs.

Also consider reordering the compares in order of descending frequency.

And using the switch statement depending on the datatype of the object you are comparing.

However at this point, as duffymo has suggested, you would be micro optimizing. The performance gain will never be as significant if you haven't chosen the right sorting algorithm for the job first.

Answer 6

I made a benchmark if there's a true difference between successive if() and if() then a few elseif()

I put a big string and did about 20 strpos() each time (x100 000) with the two methods and it showed this result :

Try 1 : 0.5094 (including elseif)
Try 2 : 0.6700 (including only if)

There's no doubt. I already knew sucessive elseif() were faster, even though there's a return in the middle ; it's still good to put some statistics in the answer.

Answer 7

If the values are integers you may achieve an optimisation by using a table lookup. E.g. say you have 256 values that map into 7 days somehow, you could set up an array with 256 cells and each cell contained the day of week you wanted. Then instead of:

if ( value == 0 ) {
  dayofweek = 1;
} else if ( value == 1 ) {
  dayofweek = 2;
} else if ( value == 2 ) {
  dayofweek = 3;
} else if ...

.. you could have..

dayofweek = lookuparray[value];

Of course, if you use this technique, then you should check the bounds of value first.

Answer 8

I would put another vote in for opting for a switch() statement instead.

Answer 9

In general, "else if" style can be faster because in the series of ifs, every condition is checked one after the other; in an "else if" chain, once one condition is matched, the rest are bypassed.

The fastest would be a table dispatch, which is what a switch statement gets optimized into when there are enough cases in it (if there are few cases in a switch, it gets translated into a series of if-else checks in the resulting machine code).

Answer 10

The decision to use many if-statements or one if-elseif-elseif... should not rely on performance, since this decision involves the program flow massively.

I doubt that you can switch from many if-statements to a big if-elseif without loosing functionality.

Its a design question, not a perfomance one.

Answer 11

This question is specially interesting when the if block returns thus finishing the method. It also applies directly to the way comparators in Java work.

I've thus run each method (bellow) 250.000.000 times and the results are as follows:

two values   if/else    - 6.43 millis
three values if/else/if - 8.66 millis
three values if/if      - 9.01 millis

While the worst case takes 1.4 times longer than the best one do notice that this is the aggregate sum of iterating each one of these methods 250 million times. Assuming that it would take 100ms for a human to perceive delay and that the worst/better difference is 2.58 millis it would imply that you would need almost a trillion (1000 * 1000 millions) iterations to perceive the difference between different methods.

Summing it up: use if-else it's one of those cases where the fastest option is also the one with more legibility and less error-prone.

// methods used to measure difference between if and if/else

/** equality is not important **/
private static int comparatorOfIfElse(int a, int b) {
    if(a < b) return -1;
    else return  1;
}

/** equality is taken into account using if/else **/
private static int comparatorOfIfElseIf(int a, int b) {
    if(a < b) return -1;
    else if(a > b) return  1;
    return 0;
}

/** equality is taken into account using only if **/
private static int comparatorOfIf(int a, int b) {
    if(a < b) return -1;
    if(a > b) return  1;
    return 0;
}