concat php strings; using . operator or double quotes

Question

Edit - My question is not stricktly limited to preformance, I would also like to know the pitfalls of each and if there is a condition where one should be used over the other.

Which is better to use to concat a string in PHP?

Option A: Use the . operator to concat the strings

$string1 = "hello ";
$string2 = $string1 . "world !";

Option B: Use double quotes

$string1 = "hello ";
$string2 = "$string1 world !";

I realize that both will in fact do the same thing, and in my personal development I prefer to use the . operator. My question only arises because i've read that the . operator forces php to re-concatenate with each new string, so in the example:

$string1 = "hello ";
$string2 = "world";
$string3 = $string1.$string2." !";

would actually run slower than

$stirng1 = "hello";
$string2 = "world";
$string3 = "$string1 $string2 !";

Reference: PHP Language Operators > Strings

Answer 1

I think before you start worrying about it, you need to see if it is even worth thinking about. I did think about it, and wrote the following tiny script and ran it to see what the benchmarks were like.

For each loop, I made 100,000 passes. Now I didn't print my strings anywhere so if the PHP optimizer takes all of my work away because of that, then I apologize. However looking at these results, you are looking at a difference of about 0.00001 second for each.

Before you optimize for anything other than readability, use a profiler and see where your hotspots are. If you run tens of millions of concatenations, then you may have an argument. But with 1000, you are still talking about a difference of 0.01 seconds. I'm sure you could save more than 0.01 seconds just by optimizing SQL queries and the like.

My evidence is below....

Here's what I ran:

<?php
for($l = 0; $l < 5; $l++)
  {
    echo "Pass " .$l. ": \n";
    $starta = microtime(1);
    for( $i = 0; $i < 100000; $i++)
      {
    $a = md5(rand());
    $b = md5(rand());
    $c = "$a $b".' Hello';
      }
    $enda = microtime(1);

    $startb = microtime(1);
    for( $i = 0; $i < 100000; $i++)
      {
    $a = md5(rand());
    $b = md5(rand());
    $c = $a . ' ' . $b . ' Hello';
      }
    $endb = microtime(1);


    echo "\tFirst method: " . ($enda - $starta) . "\n";
    echo "\tSecond method: " . ($endb - $startb) . "\n";
  }

Here are the results:

Pass 0: 
    First method: 1.3060460090637
    Second method: 1.3552670478821
Pass 1: 
    First method: 1.2648279666901
    Second method: 1.2579910755157
Pass 2: 
    First method: 1.2534148693085
    Second method: 1.2467019557953
Pass 3: 
    First method: 1.2516458034515
    Second method: 1.2479140758514
Pass 4: 
    First method: 1.2541329860687
    Second method: 1.2839770317078

Answer 2

Concatenation is almost always faster than interpolation, but the difference is rarely significant enough to warrant caring. That said, I prefer concatenation because it allows easier editing when (for example) you want to change a string to a method or function call. I.e., from:

$s1 = 'this ' . $thing . ' with a thing';

To:

$s1 = 'this ' . blarg($thing) . ' with a thing';

Edit: When I say, "Concatenation is almost always faster than interpolation," what I mean is, I have actually benchmarked many various forms of it, and I'm not just guessing, or reiterating somebody else's post. It's easy to do, try it.