Java6 String.substring()方法的内存泄露
substring(start,end)在Java编程里面经常使用，没想到如果使用不当，会出现内存泄露。
要了解substring()，最好的方法便是查看源码(jdk6)：
1 /** 2 * <blockquote><pre> 3 * "hamburger".substring(4, 8) returns "urge" 4 * "smiles".substring(1, 5) returns "mile" 5 * </pre></blockquote> 6 * 7 * @param beginIndex the beginning index, inclusive. 8 * @param endIndex the ending index, exclusive. 9 * @return the specified substring. 10 * @exception IndexOutOfBoundsException if the 11 * <code>beginIndex</code> is negative, or 12 * <code>endIndex</code> is larger than the length of 13 * this <code>String</code> object, or 14 * <code>beginIndex</code> is larger than 15 * <code>endIndex</code>. 16 */ 17 public String substring(int beginIndex, int endIndex) { 18 if (beginIndex < 0) { 19 throw new StringIndexOutOfBoundsException(beginIndex); 20 } 21 if (endIndex > count) { 22 throw new StringIndexOutOfBoundsException(endIndex); 23 } 24 if (beginIndex > endIndex) { 25 throw new StringIndexOutOfBoundsException(endIndex - beginIndex); 26 } 27 return ((beginIndex == 0) && (endIndex == count)) ? this : 28 new String(offset + beginIndex, endIndex - beginIndex, value); 29 }
插一句，这段substring()的源代码，为如何编写api提供了很好的一个例子，让我想起了老赵的一篇文章，对参数的判断，异常的处理，思路上有点接近。
值得注意的是，如果调用substring(i,i)的话（即beginIndex==endIndex）或者是substring(stringLength)（即是beginIndex==字符串长度），并不会抛出异常，而是会返回一个空的字符串，因为new String(offset + beginIndex , 0 , value)。
言归正传，真正创建字符串的，是一个String(int,in,char[])的构造函数，源代码如下：
1 // Package private constructor which shares value array for speed. 2 String(int offset, int count, char value[]) { 3 this.value = value; 4 this.offset = offset; 5 this.count = count; 6 }
Java里的字符串，其实是由三个私有变量定义：
public final class String implements java.io.Serializable, Comparable<String>, CharSequence { /** The value is used for character storage. */ private final char value[]; /** The offset is the first index of the storage that is used. */ private final int offset; /** The count is the number of characters in the String. */ private final int count; }
当为字符串分配内存时，char数组存储字符，offset=0，count=字符串长度。问题在于，由substring(start,end)调用构造函数String(int,in,char[])时，实际上是改变offset和count的位置达到取得子字符串的目的，而子字符串里的value[]数组，仍然指向原字符串。假设原字符串s有1GB，且我们需要的是s.substring(1,10)这样一段小的字符串，但由于substring()里的value[]数组仍然指向1GB的原字符串，导致原字符串无法在GC中释放，从而产生了内存泄露。
但为什么要这样设计呢？由于String是不可变的(immutable)，基于这种共享同一个字符数组的设计有以下好处：
调用substring()时无需复制数组，可重用value[]数组；且substring()的运行是常数时间，非线性，性能得到提高（这也是第二段代码注释的意思：share values for speed）。
而劣势，便是可能会产生内存泄露（实际上，Oracle早有人提出这个bug:http://bugs.sun.com/view_bug.do?bug_id=4513622）。
如何避免这个问题呢？有一个变通的方案，通过一个构造函数，复制一段数组：
1 /** 2 * Initializes a newly created {@code String} object so that it represents 3 * the same sequence of characters as the argument; in other words, the 4 * newly created string is a copy of the argument string. Unless an 5 * explicit copy of {@code original} is needed, use of this constructor is 6 * unnecessary since Strings are immutable. 7 * 8 * @param original 9 * A {@code String} 10 */ 11 public String(String original) { 12 int size = original.count; 13 char[] originalValue = original.value; 14 char[] v; 15 if (originalValue.length > size) { 16 // The array representing the String is bigger than the new 17 // String itself. Perhaps this constructor is being called 18 // in order to trim the baggage, so make a copy of the array. 19 int off = original.offset; 20 v = Arrays.copyOfRange(originalValue, off, off+size); 21 } else { 22 // The array representing the String is the same 23 // size as the String, so no point in making a copy. 24 v = originalValue; 25 } 26 this.offset = 0; 27 this.count = size; 28 this.value = v; 29 } 30 31 //smalStr no longer holds the value[] of 1GB 32 String smallStr = new String(s.substring(1,10));
上面的构造方法，重新复制了一段数组给v，然后再将v给字符串的数组，从而避免内存泄露。
在Java7里，String的实现已经改变，substring()方法的实现，由原来的共享数组变成了传统的拷贝，杜绝了内存泄露的同时也将运行时间由常数变成了线性：
1 public String substring(int beginIndex, int endIndex) { 2 if (beginIndex < 0) { 3 throw new StringIndexOutOfBoundsException(beginIndex); 4 } 5 if (endIndex > value.length) { 6 throw new StringIndexOutOfBoundsException(endIndex); 7 } 8 int subLen = endIndex - beginIndex; 9 if (subLen < 0) { 10 throw new StringIndexOutOfBoundsException(subLen); 11 } 12 return ((beginIndex == 0) && (endIndex == value.length)) ? this 13 : new String(value, beginIndex, subLen); 14 }
/** * Allocates a new {@code String} that contains characters from a subarray * of the character array argument. The {@code offset} argument is the * index of the first character of the subarray and the {@code count} * argument specifies the length of the subarray. The contents of the * subarray are copied; subsequent modification of the character array does * not affect the newly created string. * * @param value * Array that is the source of characters * * @param offset * The initial offset * * @param count * The length * * @throws IndexOutOfBoundsException * If the {@code offset} and {@code count} arguments index * characters outside the bounds of the {@code value} array */ public String(char value[], int offset, int count) { if (offset < 0) { throw new StringIndexOutOfBoundsException(offset); } if (count < 0) { throw new StringIndexOutOfBoundsException(count); } // Note: offset or count might be near -1>>>1. if (offset > value.length - count) { throw new StringIndexOutOfBoundsException(offset + count); } this.value = Arrays.copyOfRange(value, offset, offset+count); }
这个构造函数，每次都会复制数组，实现与Java6并不一样。至于哪个好哪个坏，其实很难说清楚。
据说有一种Rope的数据结构，可以更加高效地处理字符串，得好好看看。
参考：
http://javarevisited.blogspot.hk/2011/10/how-substring-in-java-works.html
http://eyalsch.wordpress.com/2009/10/27/stringleaks/
http://blog.zhaojie.me/2013/03/string-and-rope-1-string-in-dotnet-and-java.html