更快的实现:使用String.regionMatches()
使用regexp可能相对较慢。如果你只是想检查一种情况,(慢)没有关系。但如果你有一个数组或一个包含成千上万个字符串的集合,事情就会变得非常缓慢。
下面给出的解决方案既不使用正则表达式,也不使用toLowerCase()(这也很慢,因为它创建了另一个字符串,并在检查后丢弃它们)。
解决方案构建在String.regionMatches()方法上,该方法似乎是未知的。它检查2个String区域是否匹配,但重要的是它还有一个重载,带有一个方便的ignoreCase参数。
public static boolean containsIgnoreCase(String src, String what) {
final int length = what.length();
if (length == 0)
return true; // Empty string is contained
final char firstLo = Character.toLowerCase(what.charAt(0));
final char firstUp = Character.toUpperCase(what.charAt(0));
for (int i = src.length() - length; i >= 0; i--) {
// Quick check before calling the more expensive regionMatches() method:
final char ch = src.charAt(i);
if (ch != firstLo && ch != firstUp)
continue;
if (src.regionMatches(true, i, what, 0, length))
return true;
}
return false;
}
速度分析
这种速度分析并不意味着是火箭科学,只是对不同方法的速度有多快的粗略描述。
我比较了5种方法。
Our containsIgnoreCase() method.
By converting both strings to lower-case and call String.contains().
By converting source string to lower-case and call String.contains() with the pre-cached, lower-cased substring. This solution is already not as flexible because it tests a predefiend substring.
Using regular expression (the accepted answer Pattern.compile().matcher().find()...)
Using regular expression but with pre-created and cached Pattern. This solution is already not as flexible because it tests a predefined substring.
结果(通过调用该方法1000万次):
我们的方法是670毫秒
2x toLowerCase() and contains(): 2829 ms
1x toLowerCase()和contains(),缓存子字符串:2446毫秒
Regexp: 7180 ms
缓存模式的Regexp: 1845毫秒
表格中的结果:
RELATIVE SPEED 1/RELATIVE SPEED
METHOD EXEC TIME TO SLOWEST TO FASTEST (#1)
------------------------------------------------------------------------------
1. Using regionMatches() 670 ms 10.7x 1.0x
2. 2x lowercase+contains 2829 ms 2.5x 4.2x
3. 1x lowercase+contains cache 2446 ms 2.9x 3.7x
4. Regexp 7180 ms 1.0x 10.7x
5. Regexp+cached pattern 1845 ms 3.9x 2.8x
我们的方法比使用小写和contains()快4倍,比使用正则表达式快10倍,即使Pattern是预缓存的也快3倍(并且失去了检查任意子字符串的灵活性)。
分析测试代码
如果你对分析是如何执行的感兴趣,下面是完整的可运行应用程序:
import java.util.regex.Pattern;
public class ContainsAnalysis {
// Case 1 utilizing String.regionMatches()
public static boolean containsIgnoreCase(String src, String what) {
final int length = what.length();
if (length == 0)
return true; // Empty string is contained
final char firstLo = Character.toLowerCase(what.charAt(0));
final char firstUp = Character.toUpperCase(what.charAt(0));
for (int i = src.length() - length; i >= 0; i--) {
// Quick check before calling the more expensive regionMatches()
// method:
final char ch = src.charAt(i);
if (ch != firstLo && ch != firstUp)
continue;
if (src.regionMatches(true, i, what, 0, length))
return true;
}
return false;
}
// Case 2 with 2x toLowerCase() and contains()
public static boolean containsConverting(String src, String what) {
return src.toLowerCase().contains(what.toLowerCase());
}
// The cached substring for case 3
private static final String S = "i am".toLowerCase();
// Case 3 with pre-cached substring and 1x toLowerCase() and contains()
public static boolean containsConverting(String src) {
return src.toLowerCase().contains(S);
}
// Case 4 with regexp
public static boolean containsIgnoreCaseRegexp(String src, String what) {
return Pattern.compile(Pattern.quote(what), Pattern.CASE_INSENSITIVE)
.matcher(src).find();
}
// The cached pattern for case 5
private static final Pattern P = Pattern.compile(
Pattern.quote("i am"), Pattern.CASE_INSENSITIVE);
// Case 5 with pre-cached Pattern
public static boolean containsIgnoreCaseRegexp(String src) {
return P.matcher(src).find();
}
// Main method: perfroms speed analysis on different contains methods
// (case ignored)
public static void main(String[] args) throws Exception {
final String src = "Hi, I am Adam";
final String what = "i am";
long start, end;
final int N = 10_000_000;
start = System.nanoTime();
for (int i = 0; i < N; i++)
containsIgnoreCase(src, what);
end = System.nanoTime();
System.out.println("Case 1 took " + ((end - start) / 1000000) + "ms");
start = System.nanoTime();
for (int i = 0; i < N; i++)
containsConverting(src, what);
end = System.nanoTime();
System.out.println("Case 2 took " + ((end - start) / 1000000) + "ms");
start = System.nanoTime();
for (int i = 0; i < N; i++)
containsConverting(src);
end = System.nanoTime();
System.out.println("Case 3 took " + ((end - start) / 1000000) + "ms");
start = System.nanoTime();
for (int i = 0; i < N; i++)
containsIgnoreCaseRegexp(src, what);
end = System.nanoTime();
System.out.println("Case 4 took " + ((end - start) / 1000000) + "ms");
start = System.nanoTime();
for (int i = 0; i < N; i++)
containsIgnoreCaseRegexp(src);
end = System.nanoTime();
System.out.println("Case 5 took " + ((end - start) / 1000000) + "ms");
}
}