这个问题中的许多评论者不理解PHP中的数组是如何工作的。从数组文档:

键可以是整数，也可以是字符串。如果一个键是一个整数的标准表示，它将被这样解释(即。“8”解释为8，“08”解释为“08”)。key中的浮点数被截断为整数。在PHP中，索引数组和关联数组类型是相同的类型，它们都可以包含整数和字符串索引。

换句话说，没有数组键“8”这样的东西，因为它总是(无声地)转换为整数8。因此，试图区分整数和数字字符串是不必要的。

如果你想要最有效的方法来检查一个数组的非整数键，而不需要复制数组的一部分(像array_keys()那样)或全部(像foreach那样):

function keyedNext( &$arr, &$k){
    $k = key($arr);
    return next($arr);
}

for ($k = key(reset($my_array)); is_int($k); keyedNext($my_array,$k))
    $onlyIntKeys = is_null($k);

这是因为key()在当前数组位置无效时返回NULL，并且NULL永远不会是一个有效的键(如果你试图使用NULL作为数组键，它会被无声地转换为“”)。

我注意到这个问题有两种常用的方法:一种使用array_values()，另一种使用key()。为了找出哪个更快，我写了一个小程序:

$arrays = Array(
  'Array #1' => Array(1, 2, 3, 54, 23, 212, 123, 1, 1),
  'Array #2' => Array("Stack", 1.5, 20, Array(3.4)),
  'Array #3' => Array(1 => 4, 2 => 2),
  'Array #4' => Array(3.0, "2", 3000, "Stack", 5 => "4"),
  'Array #5' => Array("3" => 4, "2" => 2),
  'Array #6' => Array("0" => "One", 1.0 => "Two", 2 => "Three"),
  'Array #7' => Array(3 => "asdf", 4 => "asdf"),
  'Array #8' => Array("apple" => 1, "orange" => 2),
);

function is_indexed_array_1(Array &$arr) {
  return $arr === array_values($arr);
}

function is_indexed_array_2(Array &$arr) {
  for (reset($arr), $i = 0; key($arr) === $i++; next($arr))
    ;
  return is_null(key($arr));
}

// Method #1
$start = microtime(true);
for ($i = 0; $i < 1000; $i++) {
  foreach ($arrays as $array) {
    $dummy = is_indexed_array_1($array);
  }
}
$end = microtime(true);
echo "Time taken with method #1 = ".round(($end-$start)*1000.0,3)."ms\n";

// Method #2
$start = microtime(true);
for ($i = 0; $i < 1000; $i++) {
  foreach ($arrays as $array) {
    $dummy = is_indexed_array_2($array);
  }
}
$end = microtime(true);
echo "Time taken with method #1 = ".round(($end-$start)*1000.0,3)."ms\n";

该程序在CentOS上的PHP 5.2上的输出如下:

方法#1所花费的时间= 10.745ms 方法#2所花费的时间= 18.239ms

PHP 5.3上的输出产生了类似的结果。显然，使用array_values()要快得多。

下面是我使用的方法:

function is_associative ( $a )
{
    return in_array(false, array_map('is_numeric', array_keys($a)));
}

assert( true === is_associative(array(1, 2, 3, 4)) );

assert( false === is_associative(array('foo' => 'bar', 'bar' => 'baz')) );

assert( false === is_associative(array(1, 2, 3, 'foo' => 'bar')) );

注意，这并不包括特殊情况，如:

$a = array( 1, 2, 3, 4 );

unset($a[1]);

assert( true === is_associative($a) );

对不起，我帮不了你。对于大小合适的数组，它也有一定的性能，因为它不会产生不必要的拷贝。正是这些小细节使得Python和Ruby编写起来更加方便……: P

我的解决方案是获得如下数组的键，并检查是否键不是整数:

private function is_hash($array) {
    foreach($array as $key => $value) {
        return ! is_int($key);
    }
    return false;
}

像下面这样获取哈希数组的array_keys是错误的:

array_keys(array(
       "abc" => "gfb",
       "bdc" => "dbc"
       )
);

将输出:

array(
       0 => "abc",
       1 => "bdc"
)

因此，将它与评分最高的答案中提到的一系列数字进行比较并不是一个好主意。它总是说它是一个哈希数组如果你试着比较键和范围。

我认为下面两个函数是最好的方法去检查'如果一个数组是结合或数字'。由于'numeric'可能只表示数字键或顺序数字键，下面列出了两个函数来检查这两种情况:

function is_indexed_array(&$arr) {
  for (reset($arr); is_int(key($arr)); next($arr));
  return is_null(key($arr));
}

function is_sequential_array(&$arr, $base = 0) {
  for (reset($arr), $base = (int) $base; key($arr) === $base++; next($arr));
  return is_null(key($arr));
}

The first function checks if each key is an integer value. The second function checks if each key is an integer value and in addition checks if all keys are sequential starting at $base, which defaults to 0 and thus can be omitted if you do not need to specify another base value. key($my_array) returns null if the read pointer is moved past the end of the array, which is what ends the for loop and makes the statement after the for loop return true if all keys were integer. If not, the loop ends prematurely because a key is of type string, and the statement after the for loop will return false. The latter function in addition adds one to $base after each compare, to be able to check if the next key is of the correct value. The strict compare makes it also check if the key is of type integer. The $base = (int) $base part in the first section of the for loop can be left out when $base is omitted or if you make sure it is only called using an integer. But since I can't be sure for everybody, I left it in. The statement is executed only once, anyway. I think these are the most efficient solutions:

Memory wise: No copying of data or key ranges. Doing an array_values or array_keys may seem shorter (less code) but keep in mind what goes on in the background once you make that call. Yes there are more (visible) statements than in some other solutions, but that is not what counts, is it? Time wise: Besides the fact that copying/extracting data and/or keys also takes time, this solution is more efficient than doing a foreach. Again a foreach may seem more efficient to some because it is shorter in notation, but in the background foreach also calls reset, key and next to do it's looping. But in addition it also calls valid to check the end condition, which is avoided here due to the combination with the integer check.

请记住，数组键只能是整数或字符串，严格的数字字符串，如“1”(但不是“01”)将被转换为整数。如果您希望数组是顺序的，这使得检查整数键成为除了计数之外唯一需要的操作。当然，如果is_indexed_array返回false，则可以将该数组视为关联数组。我说“见过”，因为事实上它们都是。

解决这个问题的一种方法是利用json_encode，它已经有自己的内部方法来区分关联数组和索引数组，以便输出正确的JSON。

可以通过检查编码后返回的第一个字符是{(关联数组)还是[(索引数组)来做到这一点。

// Too short :)
function is_assoc($arr) {
    ksort($arr);
    return json_encode($arr)[0] === '{';
}