这是一个稍微不同风格的递归解决方案:

public static int countOccurrences(String haystack, char needle)
{
    return countOccurrences(haystack, needle, 0);
}

private static int countOccurrences(String haystack, char needle, int accumulator)
{
    if (haystack.length() == 0) return accumulator;
    return countOccurrences(haystack.substring(1), needle, haystack.charAt(0) == needle ? accumulator + 1 : accumulator);
}

可以在一行代码中使用split()函数

int noOccurence=string.split("#",-1).length-1;

好的，受Yonatan解决方案的启发，这里有一个纯递归的库方法-唯一使用的库方法是length()和charAt()，它们都不做任何循环:

public static int countOccurrences(String haystack, char needle)
{
    return countOccurrences(haystack, needle, 0);
}

private static int countOccurrences(String haystack, char needle, int index)
{
    if (index >= haystack.length())
    {
        return 0;
    }

    int contribution = haystack.charAt(index) == needle ? 1 : 0;
    return contribution + countOccurrences(haystack, needle, index+1);
}

递归是否算作循环取决于您使用的确切定义，但这可能是您能得到的最接近的定义。

我不知道现在大多数jvm是否都做尾递归…当然，如果不是这样，对于适当长的字符串就会出现同名堆栈溢出。

int count = (line.length() - line.replace("str", "").length())/"str".length();

我对此的“惯用语”是:

int count = StringUtils.countMatches("a.b.c.d", ".");

既然已经是通用语言了，为什么还要自己写呢?

Spring Framework的线性程序是:

int occurance = StringUtils.countOccurrencesOf("a.b.c.d", ".");

灵感来自Jon Skeet，一个非循环版本，不会吹你的堆栈。如果你想使用fork-join框架，这也是一个有用的起点。

public static int countOccurrences(CharSequeunce haystack, char needle) {
    return countOccurrences(haystack, needle, 0, haystack.length);
}

// Alternatively String.substring/subsequence use to be relatively efficient
//   on most Java library implementations, but isn't any more [2013].
private static int countOccurrences(
    CharSequence haystack, char needle, int start, int end
) {
    if (start == end) {
        return 0;
    } else if (start+1 == end) {
        return haystack.charAt(start) == needle ? 1 : 0;
    } else {
        int mid = (end+start)>>>1; // Watch for integer overflow...
        return
            countOccurrences(haystack, needle, start, mid) +
            countOccurrences(haystack, needle, mid, end);
    }
}

(免责声明:未经测试，未经编译，不合理。)

也许最好的(单线程，不支持代理对)编写方法是:

public static int countOccurrences(String haystack, char needle) {
    int count = 0;
    for (char c : haystack.toCharArray()) {
        if (c == needle) {
           ++count;
        }
    }
    return count;
}