为什么$ arr [] ='x'比$ arr [0] ='x'快

Question

I have the following:

# a.php
for($i=1; $i<=5000000; $i++) {
    $arr = [];
    for($f = 1; $f <= 5; $f++) {
        $arr[$f] = 'a'; # <-- I am passing an index manually
    }
}

And this:

# b.php
for($i=1; $i<=5000000; $i++) {
    $arr = [];
    for($f = 1; $f <= 5; $f++) {
        $arr[] = 'a'; # <-- Note that I am not passing an index manually
    }
}

Why is b.php code faster than a.php code?...

In b.php I am not passing an index manually, so PHP calculates it (isn't this slower?), and a.php passes a defined index to that array, so I'm confused about this

Used npm's gnomon package for time measurement

~/$ php a.php | gnomon
   1.0981s   

     Total   1.0985s

~/$ php a.php | gnomon
   1.1350s   

     Total   1.1358s

~/$ php a.php | gnomon
   1.1664s   

     Total   1.1668s

~/$ php a.php | gnomon
   1.1105s   

     Total   1.1108s

~/$ php a.php | gnomon
   1.1074s   

     Total   1.1078s

~/$ php a.php | gnomon
   1.0969s   

     Total   1.0973s

~/$ php a.php | gnomon
   1.0872s   

     Total   1.0875s

~/$ php a.php | gnomon
   1.0992s   

     Total   1.0996s

And

~/$ php b.php | gnomon
   0.8960s   

     Total   0.8984s

~/$ php b.php | gnomon
   0.8859s   

     Total   0.8863s

~/$ php b.php | gnomon
   0.9031s   

     Total   0.9035s

~/$ php b.php | gnomon
   0.9078s   

     Total   0.9083s

~/$ php b.php | gnomon
   0.8880s   

     Total   0.8884s

~/$ php b.php | gnomon
   0.8945s   

     Total   0.8951s

~/$ php b.php | gnomon
   0.8891s   

     Total   0.8896s

~/$ php test.php | gnomon
   0.8843s   

     Total   0.8847s

Answer 1

In the first solution, php has to figure out what index must be used to set the new value and check if we are going to update existed element or add a new one.

In b.php new element is always put on the end of the array, the additional check of index is not required. This is basically how stack works.