Documentation states that "seconds" must be positive. This is not correct, 0 is possible.
Rather, "seconds" must be non-negative.(PHP 5, PHP 7, PHP 8)
time_nanosleep — Delay for a number of seconds and nanoseconds
   Delays program execution for the given number of
   seconds and nanoseconds.
  
secondsMust be a non-negative integer.
nanosecondsMust be a non-negative integer less than 1 billion.
Note: On Windows, the system may sleep longer that the given number of nanoseconds, depending on the hardware.
   Returns true on success or false on failure.
  
If the delay was interrupted by a signal, an associative array will be returned with the components:
seconds - number of seconds remaining in
      the delay
     
    nanoseconds - number of nanoseconds
      remaining in the delay
     
    
Example #1 time_nanosleep() example
<?php
// Careful! This won't work as expected if an array is returned
if (time_nanosleep(0, 500000000)) {
    echo "Slept for half a second.\n";
}
// This is better:
if (time_nanosleep(0, 500000000) === true) {
    echo "Slept for half a second.\n";
}
// And this is the best:
$nano = time_nanosleep(2, 100000);
if ($nano === true) {
    echo "Slept for 2 seconds, 100 microseconds.\n";
} elseif ($nano === false) {
    echo "Sleeping failed.\n";
} elseif (is_array($nano)) {
    $seconds = $nano['seconds'];
    $nanoseconds = $nano['nanoseconds'];
    echo "Interrupted by a signal.\n";
    echo "Time remaining: $seconds seconds, $nanoseconds nanoseconds.";
}
?>
Documentation states that "seconds" must be positive. This is not correct, 0 is possible.
Rather, "seconds" must be non-negative.You should take into account, if you use the function replacement down here, the CPU will be in use of 99% for the time of execution...
(A little bit better in this situation is to let the 'full seconds' go by a normal sleep command (makes the thread sleep!, and uses minimum cpu))
<?php
    //THIS IS THE FUNCTION WE ARE TALKIN ABOUT
    function timeWait($microtime)
    {
//optimizations added by me [start]
//sleep the full seconds
sleep(intval($microtime));
//set the microtime to only resleep the last part of the nanos
$microtime = $microtime - intval($microtime);
//optimizations added by me [end]
        $timeLimit = $microtime + array_sum(explode(" ",microtime()));
        while(array_sum(explode(" ",microtime())) < $timeLimit)
        {/*DO NOTHING*/}
        return(true);
    }
 
    //THIS IS HOW WE CAN USE IT
    echo "Process started at " . date("H:i:s") . " and " . current(explode(" ",microtime())) . " nanoseconds.<br>";
    timeWait(5.5); //With this call the system will wait 5 seconds and a half. You can use either integer or float.
    echo "Process completed at " . date("H:i:s") . " and " . current(explode(" ",microtime())) . " nanoseconds.";
 ?>Just glancing at this - and the note from over a year ago with a implementation for windows.. with 5.0.0 and higher it would be simplier to just do something like......
<?php
if (!function_exists('time_nanosleep')) {
function time_nanosleep($seconds, $nanoseconds) {
sleep($seconds);
usleep(round($nanoseconds/100));
return true;
}
}
?>
....off the top of my head - obviously simple enough there should be no mistakes.. but those are the ones that always seem to get ya :( .....A response to the note below:
Your function is also useless, as the WinNT 32 kernel only functions at a minimum of about 10+ ms (1,000 us), rendering usleep() useless, because usleep uses the C function which is provided by the system (in this case, kernel32.dll).
You'll want to use a function that does not rely on the kernel, but rather something made for precise measurement:
<?php
function usleep_win( $micro_seconds )
{
    if ( @function_exists( "socket_create" ) && @function_exists( "socket_select" ) )
    {
        $false = NULL;
        $socket = array( socket_create( AF_INET, SOCK_RAW, $false ) );
        socket_select( $false, $false, $socket, 0, $micro_seconds );
        return true;
    }
    else
    {
        return false;
    }
}
?>
This function will allow to you sleep for a specified microsecond, although I have measured it to be off by ~5 us.
Again, most of this depends on the hardware in your system. If you _REALLY_ need to be precise to < 10 us, you shouldn't be using WinNT anyways!