我想要一个优雅的解决方案来解决这个问题：
- 不使用任何外部库
- 使用Reflection 访问字段，包括超类字段
- 使用递归遍历对象图，每次调用只有一个堆栈帧
- 使用IdentityHashMap 处理向后引用并避免无限递归
- 适当地处理基元、自动装箱、CharSequences、枚举和空值
- 允许您选择是否解析静态字段
- 足够简单，可以根据格式偏好进行修改
我编写了以下实用程序类：
import java.lang.reflect.Array;
import java.lang.reflect.Field;
import java.lang.reflect.Modifier;
import java.util.IdentityHashMap;
import java.util.Map.Entry;
import java.util.TreeMap;
/**
* Utility class to dump {@code Object}s to string using reflection and recursion.
*/
public class StringDump {
/**
* Uses reflection and recursion to dump the contents of the given object using a custom, JSON-like notation (but not JSON). Does not format static fields.<p>
* @see #dump(Object, boolean, IdentityHashMap, int)
* @param object the {@code Object} to dump using reflection and recursion
* @return a custom-formatted string representing the internal values of the parsed object
*/
public static String dump(Object object) {
return dump(object, false, new IdentityHashMap<Object, Object>(), 0);
}
/**
* Uses reflection and recursion to dump the contents of the given object using a custom, JSON-like notation (but not JSON).<p>
* Parses all fields of the runtime class including super class fields, which are successively prefixed with "{@code super.}" at each level.<p>
* {@code Number}s, {@code enum}s, and {@code null} references are formatted using the standard {@link String#valueOf()} method.
* {@code CharSequences}s are wrapped with quotes.<p>
* The recursive call invokes only one method on each recursive call, so limit of the object-graph depth is one-to-one with the stack overflow limit.<p>
* Backwards references are tracked using a "visitor map" which is an instance of {@link IdentityHashMap}.
* When an existing object reference is encountered the {@code "sysId"} is printed and the recursion ends.<p>
*
* @param object the {@code Object} to dump using reflection and recursion
* @param isIncludingStatics {@code true} if {@code static} fields should be dumped, {@code false} to skip them
* @return a custom-formatted string representing the internal values of the parsed object
*/
public static String dump(Object object, boolean isIncludingStatics) {
return dump(object, isIncludingStatics, new IdentityHashMap<Object, Object>(), 0);
}
private static String dump(Object object, boolean isIncludingStatics, IdentityHashMap<Object, Object> visitorMap, int tabCount) {
if (object == null ||
object instanceof Number || object instanceof Character || object instanceof Boolean ||
object.getClass().isPrimitive() || object.getClass().isEnum()) {
return String.valueOf(object);
}
StringBuilder builder = new StringBuilder();
int sysId = System.identityHashCode(object);
if (object instanceof CharSequence) {
builder.append("\"").append(object).append("\"");
}
else if (visitorMap.containsKey(object)) {
builder.append("(sysId#").append(sysId).append(")");
}
else {
visitorMap.put(object, object);
StringBuilder tabs = new StringBuilder();
for (int t = 0; t < tabCount; t++) {
tabs.append("\t");
}
if (object.getClass().isArray()) {
builder.append("[").append(object.getClass().getName()).append(":sysId#").append(sysId);
int length = Array.getLength(object);
for (int i = 0; i < length; i++) {
Object arrayObject = Array.get(object, i);
String dump = dump(arrayObject, isIncludingStatics, visitorMap, tabCount + 1);
builder.append("\n\t").append(tabs).append("\"").append(i).append("\":").append(dump);
}
builder.append(length == 0 ? "" : "\n").append(length == 0 ? "" : tabs).append("]");
}
else {
// enumerate the desired fields of the object before accessing
TreeMap<String, Field> fieldMap = new TreeMap<String, Field>(); // can modify this to change or omit the sort order
StringBuilder superPrefix = new StringBuilder();
for (Class<?> clazz = object.getClass(); clazz != null && !clazz.equals(Object.class); clazz = clazz.getSuperclass()) {
Field[] fields = clazz.getDeclaredFields();
for (int i = 0; i < fields.length; i++) {
Field field = fields[i];
if (isIncludingStatics || !Modifier.isStatic(field.getModifiers())) {
fieldMap.put(superPrefix + field.getName(), field);
}
}
superPrefix.append("super.");
}
builder.append("{").append(object.getClass().getName()).append(":sysId#").append(sysId);
for (Entry<String, Field> entry : fieldMap.entrySet()) {
String name = entry.getKey();
Field field = entry.getValue();
String dump;
try {
boolean wasAccessible = field.isAccessible();
field.setAccessible(true);
Object fieldObject = field.get(object);
field.setAccessible(wasAccessible); // the accessibility flag should be restored to its prior ClassLoader state
dump = dump(fieldObject, isIncludingStatics, visitorMap, tabCount + 1);
}
catch (Throwable e) {
dump = "!" + e.getClass().getName() + ":" + e.getMessage();
}
builder.append("\n\t").append(tabs).append("\"").append(name).append("\":").append(dump);
}
builder.append(fieldMap.isEmpty() ? "" : "\n").append(fieldMap.isEmpty() ? "" : tabs).append("}");
}
}
return builder.toString();
}
}
我在多个课程中对其进行了测试，对我来说它非常有效。例如，尝试使用它来转储主线程：
public static void main(String[] args) throws Exception {
System.out.println(dump(Thread.currentThread()));
}
编辑
自从写这篇文章以来，我有理由创建这个算法的迭代版本。递归版本的深度受限于总堆栈帧，但您可能有理由转储一个非常大的对象图。为了处理我的情况，我修改了算法以使用堆栈数据结构代替运行时堆栈。此版本具有时间效率，并且受堆大小而不是堆栈帧深度的限制。
您可以下载并使用iterative version here。