Fastjson序列化原理
关键字: fastjson stackoverflow
本文使用的版本是 1.2.32
fastjson 是阿里开源的Json格式化工具库。在项目中使用了fastjson，然后出现了一个奇怪的bug。程序在序列化的时候递归调用了我调用序列化函数的函数。简单点说就是序列化中递归地调用了自己，最后stackoverflow。
下面是是使用的代码：
public class Host {
private String name;
public Host() {}
public String getName() {
return name;
}
public void setName(String name) {
this.name = name;
}
public static Host factory(byte [] bytes) {
return JSON.parseObjec(bytes, Host.class);
}
public byte[] getJson() {
return JSON.toJSONBytes(this);
}
}
然后在程序中某处使用byte []bytes = host.getJson()，出现的错误大概如下：
java.lang.StackOverflowError
at com.alibaba.fastjson.serializer.JSONSerializer.setContext(JSONSerializer.java:113)
at com.alibaba.fastjson.serializer.JSONSerializer.setContext(JSONSerializer.java:109)
at com.alibaba.fastjson.serializer.ASMSerializer_1_Host.write(Unknown Source)
at com.alibaba.fastjson.serializer.JSONSerializer.write(JSONSerializer.java:275)
at com.alibaba.fastjson.JSON.toJSONBytes(JSON.java:679)
at com.alibaba.fastjson.JSON.toJSONBytes(JSON.java:605)
at com.alibaba.fastjson.JSON.toJSONBytes(JSON.java:598)
at xxx.Host.getBytes(Host.java:38)
at com.alibaba.fastjson.serializer.ASMSerializer_1_Host.write(Unknown Source)
at com.alibaba.fastjson.serializer.JSONSerializer.write(JSONSerializer.java:275)
at com.alibaba.fastjson.JSON.toJSONBytes(JSON.java:679)
at com.alibaba.fastjson.JSON.toJSONBytes(JSON.java:605)
at com.alibaba.fastjson.JSON.toJSONBytes(JSON.java:598)
at xxx.Host.getBytes(Host.java:38)
分析调用堆栈发现fastjson在生成的serializer.ASMSerializer_1_Host中调用了Host.getJson()导致了递归。排除自己的错误后，就将代码定位到了fastjson中，应该是fastjson中出了问题。然后开始调试代码：
public static byte[] toJSONBytes(Object object, SerializeConfig config, int defaultFeatures, SerializerFeature... features) {
SerializeWriter out = new SerializeWriter(null, defaultFeatures, features);
try {
JSONSerializer serializer = new JSONSerializer(out, config);
serializer.write(object);
return out.toBytes(IOUtils.UTF8);
} finally {
out.close();
}
}
按照栈调用顺序来看，出错点应该在serializer.write(object)内部，继续深入：
public final void write(Object object) {
if (object == null) {
out.writeNull();
return;
}Class clazz = object.getClass();
ObjectSerializer writer = getObjectWriter(clazz);
try {
writer.write(this, object, null, null, 0);
} catch (IOException e) {
throw new JSONException(e.getMessage(), e);
}
}
这里发现通过getObjectWriter(clazz)取得了host的writer，想必就是自动生成的ASMSerializer_1_Host实例。本来想进入writer.write中观察，没有源代码只好放弃。然后将目标放到getObjectWriter中，看看在writer实例构造过程中能不能找到点线索。
【Fastjson序列化原理】经过几层跳转，来到了真正的getObjectWriter中：
private ObjectSerializer getObjectWriter(Class clazz, boolean create) {
ObjectSerializer writer = serializers.get(clazz);
if (writer == null) {
try {
final ClassLoader classLoader = Thread.currentThread().getContextClassLoader();
for (Object o : ServiceLoader.load(AutowiredObjectSerializer.class, classLoader)) {
if (!(o instanceof AutowiredObjectSerializer)) {
continue;
}AutowiredObjectSerializer autowired = (AutowiredObjectSerializer) o;
for (Type forType : autowired.getAutowiredFor()) {
put(forType, autowired);
}
}
} catch (ClassCastException ex) {
// skip
}writer = serializers.get(clazz);
}if (writer == null) {
final ClassLoader classLoader = JSON.class.getClassLoader();
if (classLoader != Thread.currentThread().getContextClassLoader()) {
try {
for (Object o : ServiceLoader.load(AutowiredObjectSerializer.class, classLoader)) {if (!(o instanceof AutowiredObjectSerializer)) {
continue;
}AutowiredObjectSerializer autowired = (AutowiredObjectSerializer) o;
for (Type forType : autowired.getAutowiredFor()) {
put(forType, autowired);
}
}
} catch (ClassCastException ex) {
// skip
}writer = serializers.get(clazz);
}
}if (writer == null) {
if (Map.class.isAssignableFrom(clazz)) {
put(clazz, MapSerializer.instance);
} else if (List.class.isAssignableFrom(clazz)) {
put(clazz, ListSerializer.instance);
} else if (Collection.class.isAssignableFrom(clazz)) {
put(clazz, CollectionCodec.instance);
} else if (Date.class.isAssignableFrom(clazz)) {
put(clazz, DateCodec.instance);
} else if (JSONAware.class.isAssignableFrom(clazz)) {
put(clazz, JSONAwareSerializer.instance);
} else if (JSONSerializable.class.isAssignableFrom(clazz)) {
put(clazz, JSONSerializableSerializer.instance);
} else if (JSONStreamAware.class.isAssignableFrom(clazz)) {
put(clazz, MiscCodec.instance);
} else if (clazz.isEnum() || (clazz.getSuperclass() != null && clazz.getSuperclass().isEnum())) {
JSONType jsonType = clazz.getAnnotation(JSONType.class);
if (jsonType != null && jsonType.serializeEnumAsJavaBean()) {
put(clazz, createJavaBeanSerializer(clazz));
} else {
put(clazz, EnumSerializer.instance);
}
} else if (clazz.isArray()) {
Class componentType = clazz.getComponentType();
ObjectSerializer compObjectSerializer = getObjectWriter(componentType);
put(clazz, new ArraySerializer(componentType, compObjectSerializer));
} else if (Throwable.class.isAssignableFrom(clazz)) {
SerializeBeanInfo beanInfo = TypeUtils.buildBeanInfo(clazz, null, propertyNamingStrategy);
beanInfo.features |= SerializerFeature.WriteClassName.mask;
put(clazz, new JavaBeanSerializer(beanInfo));
} else if (TimeZone.class.isAssignableFrom(clazz) || Map.Entry.class.isAssignableFrom(clazz)) {
put(clazz, MiscCodec.instance);
} else if (Appendable.class.isAssignableFrom(clazz)) {
put(clazz, AppendableSerializer.instance);
} else if (Charset.class.isAssignableFrom(clazz)) {
put(clazz, ToStringSerializer.instance);
} else if (Enumeration.class.isAssignableFrom(clazz)) {
put(clazz, EnumerationSerializer.instance);
} else if (Calendar.class.isAssignableFrom(clazz) //
|| XMLGregorianCalendar.class.isAssignableFrom(clazz)) {
put(clazz, CalendarCodec.instance);
} else if (Clob.class.isAssignableFrom(clazz)) {
put(clazz, ClobSeriliazer.instance);
} else if (TypeUtils.isPath(clazz)) {
put(clazz, ToStringSerializer.instance);
} else if (Iterator.class.isAssignableFrom(clazz)) {
put(clazz, MiscCodec.instance);
} else {
String className = clazz.getName();
if (className.startsWith("java.awt.") //
&& AwtCodec.support(clazz) //
) {
// awt
if (!awtError) {
try {
put(Class.forName("java.awt.Color"), AwtCodec.instance);
put(Class.forName("java.awt.Font"), AwtCodec.instance);
put(Class.forName("java.awt.Point"), AwtCodec.instance);
put(Class.forName("java.awt.Rectangle"), AwtCodec.instance);
} catch (Throwable e) {
awtError = true;
// skip
}
}
returnAwtCodec.instance;
}// jdk8
if ((!jdk8Error) //
&& (className.startsWith("java.time.") //
|| className.startsWith("java.util.Optional") //
|| className.equals("java.util.concurrent.atomic.LongAdder")
|| className.equals("java.util.concurrent.atomic.DoubleAdder")
)) {
try {
put(Class.forName("java.time.LocalDateTime"), Jdk8DateCodec.instance);
put(Class.forName("java.time.LocalDate"), Jdk8DateCodec.instance);
put(Class.forName("java.time.LocalTime"), Jdk8DateCodec.instance);
put(Class.forName("java.time.ZonedDateTime"), Jdk8DateCodec.instance);
put(Class.forName("java.time.OffsetDateTime"), Jdk8DateCodec.instance);
put(Class.forName("java.time.OffsetTime"), Jdk8DateCodec.instance);
put(Class.forName("java.time.ZoneOffset"), Jdk8DateCodec.instance);
put(Class.forName("java.time.ZoneRegion"), Jdk8DateCodec.instance);
put(Class.forName("java.time.Period"), Jdk8DateCodec.instance);
put(Class.forName("java.time.Duration"), Jdk8DateCodec.instance);
put(Class.forName("java.time.Instant"), Jdk8DateCodec.instance);
put(Class.forName("java.util.Optional"), OptionalCodec.instance);
put(Class.forName("java.util.OptionalDouble"), OptionalCodec.instance);
put(Class.forName("java.util.OptionalInt"), OptionalCodec.instance);
put(Class.forName("java.util.OptionalLong"), OptionalCodec.instance);
put(Class.forName("java.util.concurrent.atomic.LongAdder"), AdderSerializer.instance);
put(Class.forName("java.util.concurrent.atomic.DoubleAdder"), AdderSerializer.instance);
writer = serializers.get(clazz);
if (writer != null) {
return writer;
}
} catch (Throwable e) {
// skip
jdk8Error = true;
}
}if ((!oracleJdbcError) //
&& className.startsWith("oracle.sql.")) {
try {
put(Class.forName("oracle.sql.DATE"), DateCodec.instance);
put(Class.forName("oracle.sql.TIMESTAMP"), DateCodec.instance);
writer = serializers.get(clazz);
if (writer != null) {
return writer;
}
} catch (Throwable e) {
// skip
oracleJdbcError = true;
}
}if ((!springfoxError) //
&& className.equals("springfox.documentation.spring.web.json.Json")) {
try {
put(Class.forName("springfox.documentation.spring.web.json.Json"), //
SwaggerJsonSerializer.instance);
writer = serializers.get(clazz);
if (writer != null) {
return writer;
}
} catch (ClassNotFoundException e) {
// skip
springfoxError = true;
}
}if ((!guavaError) //
&& className.startsWith("com.google.common.collect.")) {
try {
put(Class.forName("com.google.common.collect.HashMultimap"), //
GuavaCodec.instance);
put(Class.forName("com.google.common.collect.LinkedListMultimap"), //
GuavaCodec.instance);
put(Class.forName("com.google.common.collect.ArrayListMultimap"), //
GuavaCodec.instance);
put(Class.forName("com.google.common.collect.TreeMultimap"), //
GuavaCodec.instance);
writer = serializers.get(clazz);
if (writer != null) {
return writer;
}
} catch (ClassNotFoundException e) {
// skip
guavaError = true;
}
}if (className.equals("net.sf.json.JSONNull")) {
try {
put(Class.forName("net.sf.json.JSONNull"), //
MiscCodec.instance);
} catch (ClassNotFoundException e) {
// skip
}
writer = serializers.get(clazz);
if (writer != null) {
return writer;
}
}if (TypeUtils.isProxy(clazz)) {
Class superClazz = clazz.getSuperclass();
ObjectSerializer superWriter = getObjectWriter(superClazz);
put(clazz, superWriter);
return superWriter;
}if (create) {
put(clazz, createJavaBeanSerializer(clazz));
}
}writer = serializers.get(clazz);
}
return writer;
}
简单扫描代码逻辑，发现writer是通过serializers.get(clazz)获取的。而代码中分别从Thread.currentThread().getContextClassLoader、JSON.class.getClassLoader以及最后对一下常见类分析来填充serializers。最后一种办法的末尾，走到了：
put(clazz, createJavaBeanSerializer(clazz));
可以发现逻辑是实在找不到，使用createJavaBeanSerializer(clazz)来创建clazz对应的writer。看来我们的目标应该是这个createJavaBeanSerializer函数，所以进一步深入：
private final ObjectSerializer createJavaBeanSerializer(Class clazz) {
SerializeBeanInfo beanInfo = TypeUtils.buildBeanInfo(clazz, null, propertyNamingStrategy, fieldBased);
if (beanInfo.fields.length == 0 && Iterable.class.isAssignableFrom(clazz)) {
return MiscCodec.instance;
}return createJavaBeanSerializer(beanInfo);
}
首先调用TypeUtils.buildBeanInfo来生成SerializerBeanInfo。
public static SerializeBeanInfo buildBeanInfo(Class beanType //
, Map aliasMap //
, PropertyNamingStrategy propertyNamingStrategy //
, boolean fieldBased //
) {JSONType jsonType = beanType.getAnnotation(JSONType.class);
// fieldName,field ，先生成fieldName的快照，减少之后的findField的轮询
Map fieldCacheMap = new HashMap();
ParserConfig.parserAllFieldToCache(beanType, fieldCacheMap);
List fieldInfoList = fieldBased
? computeGettersWithFieldBase(beanType, aliasMap, false, propertyNamingStrategy) //
: computeGetters(beanType, jsonType, aliasMap, fieldCacheMap, false, propertyNamingStrategy);
FieldInfo[] fields = new FieldInfo[fieldInfoList.size()];
fieldInfoList.toArray(fields);
String[] orders = null;
final int features;
String typeName = null;
if (jsonType != null) {
orders = jsonType.orders();
typeName = jsonType.typeName();
if (typeName.length() == 0) {
typeName = null;
}
features = SerializerFeature.of(jsonType.serialzeFeatures());
} else {
features = 0;
}FieldInfo[] sortedFields;
List sortedFieldList;
if (orders != null && orders.length != 0) {
sortedFieldList = fieldBased
? computeGettersWithFieldBase(beanType, aliasMap, true, propertyNamingStrategy) //
: computeGetters(beanType, jsonType, aliasMap,fieldCacheMap, true, propertyNamingStrategy);
} else {
sortedFieldList = new ArrayList(fieldInfoList);
Collections.sort(sortedFieldList);
}
sortedFields = new FieldInfo[sortedFieldList.size()];
sortedFieldList.toArray(sortedFields);
if (Arrays.equals(sortedFields, fields)) {
sortedFields = fields;
}return new SerializeBeanInfo(beanType, jsonType, typeName, features, fields, sortedFields);
}
其中parserAllFieldToCache将字段保存起来，减少访问次数。紧接着设置fieldInfoList的值，此时fieldBase为false，所以进入了computeGetters。
public static List computeGetters(Class clazz, //
JSONType jsonType, //
Map aliasMap, //
Map fieldCacheMap, //
boolean sorted, //
PropertyNamingStrategy propertyNamingStrategy //
) {
Map fieldInfoMap = new LinkedHashMap();
for (Method method : clazz.getMethods()) {
String methodName = method.getName();
int ordinal = 0, serialzeFeatures = 0, parserFeatures = 0;
String label = null;
if (Modifier.isStatic(method.getModifiers())) {
continue;
}if (method.getReturnType().equals(Void.TYPE)) {
continue;
}if (method.getParameterTypes().length != 0) {
continue;
}if (method.getReturnType() == ClassLoader.class) {
continue;
}if (method.getName().equals("getMetaClass")
&& method.getReturnType().getName().equals("groovy.lang.MetaClass")) {
continue;
}JSONField annotation = method.getAnnotation(JSONField.class);
if (annotation == null) {
annotation = getSuperMethodAnnotation(clazz, method);
}if (annotation != null) {
if (!annotation.serialize()) {
continue;
}ordinal = annotation.ordinal();
serialzeFeatures = SerializerFeature.of(annotation.serialzeFeatures());
parserFeatures = Feature.of(annotation.parseFeatures());
if (annotation.name().length() != 0) {
String propertyName = annotation.name();
if (aliasMap != null) {
propertyName = aliasMap.get(propertyName);
if (propertyName == null) {
continue;
}
}FieldInfo fieldInfo = new FieldInfo(propertyName, method, null, clazz, null, ordinal,
serialzeFeatures, parserFeatures, annotation, null, label);
fieldInfoMap.put(propertyName, fieldInfo);
continue;
}if (annotation.label().length() != 0) {
label = annotation.label();
}
}if (methodName.startsWith("get")) {
if (methodName.length() < 4) {
continue;
}if (methodName.equals("getClass")) {
continue;
}if (methodName.equals("getDeclaringClass") && clazz.isEnum()) {
continue;
}char c3 = methodName.charAt(3);
String propertyName;
if (Character.isUpperCase(c3) //
|| c3 > 512 // for unicode method name
) {
if (compatibleWithJavaBean) {
propertyName = decapitalize(methodName.substring(3));
} else {
propertyName = Character.toLowerCase(methodName.charAt(3)) + methodName.substring(4);
}
propertyName = getPropertyNameByCompatibleFieldName(fieldCacheMap, methodName,propertyName,3);
} else if (c3 == '_') {
propertyName = methodName.substring(4);
} else if (c3 == 'f') {
propertyName = methodName.substring(3);
} else if (methodName.length() >= 5 && Character.isUpperCase(methodName.charAt(4))) {
propertyName = decapitalize(methodName.substring(3));
} else {
continue;
}boolean ignore = isJSONTypeIgnore(clazz, propertyName);
if (ignore) {
continue;
}
//假如bean的field很多的情况一下，轮询时将大大降低效率
Field field = ParserConfig.getFieldFromCache(propertyName, fieldCacheMap);
if (field == null && propertyName.length() > 1) {
char ch = propertyName.charAt(1);
if (ch >= 'A' && ch <= 'Z') {
String javaBeanCompatiblePropertyName = decapitalize(methodName.substring(3));
field = ParserConfig.getFieldFromCache(javaBeanCompatiblePropertyName, fieldCacheMap);
}
}JSONField fieldAnnotation = null;
if (field != null) {
fieldAnnotation = field.getAnnotation(JSONField.class);
if (fieldAnnotation != null) {
if (!fieldAnnotation.serialize()) {
continue;
}ordinal = fieldAnnotation.ordinal();
serialzeFeatures = SerializerFeature.of(fieldAnnotation.serialzeFeatures());
parserFeatures = Feature.of(fieldAnnotation.parseFeatures());
if (fieldAnnotation.name().length() != 0) {
propertyName = fieldAnnotation.name();
if (aliasMap != null) {
propertyName = aliasMap.get(propertyName);
if (propertyName == null) {
continue;
}
}
}if (fieldAnnotation.label().length() != 0) {
label = fieldAnnotation.label();
}
}
}if (aliasMap != null) {
propertyName = aliasMap.get(propertyName);
if (propertyName == null) {
continue;
}
}if (propertyNamingStrategy != null) {
propertyName = propertyNamingStrategy.translate(propertyName);
}FieldInfo fieldInfo = new FieldInfo(propertyName, method, field, clazz, null, ordinal, serialzeFeatures, parserFeatures,
annotation, fieldAnnotation, label);
fieldInfoMap.put(propertyName, fieldInfo);
}if (methodName.startsWith("is")) {
if (methodName.length() < 3) {
continue;
}if (method.getReturnType() != Boolean.TYPE
&& method.getReturnType() != Boolean.class) {
continue;
}char c2 = methodName.charAt(2);
String propertyName;
if (Character.isUpperCase(c2)) {
if (compatibleWithJavaBean) {
propertyName = decapitalize(methodName.substring(2));
} else {
propertyName = Character.toLowerCase(methodName.charAt(2)) + methodName.substring(3);
}
propertyName = getPropertyNameByCompatibleFieldName(fieldCacheMap, methodName,propertyName,2);
} else if (c2 == '_') {
propertyName = methodName.substring(3);
} else if (c2 == 'f') {
propertyName = methodName.substring(2);
} else {
continue;
}boolean ignore = isJSONTypeIgnore(clazz, propertyName);
if (ignore) {
continue;
}Field field = ParserConfig.getFieldFromCache(propertyName,fieldCacheMap);
if (field == null) {
field = ParserConfig.getFieldFromCache(methodName,fieldCacheMap);
}JSONField fieldAnnotation = null;
if (field != null) {
fieldAnnotation = field.getAnnotation(JSONField.class);
if (fieldAnnotation != null) {
if (!fieldAnnotation.serialize()) {
continue;
}ordinal = fieldAnnotation.ordinal();
serialzeFeatures = SerializerFeature.of(fieldAnnotation.serialzeFeatures());
parserFeatures = Feature.of(fieldAnnotation.parseFeatures());
if (fieldAnnotation.name().length() != 0) {
propertyName = fieldAnnotation.name();
if (aliasMap != null) {
propertyName = aliasMap.get(propertyName);
if (propertyName == null) {
continue;
}
}
}if (fieldAnnotation.label().length() != 0) {
label = fieldAnnotation.label();
}
}
}if (aliasMap != null) {
propertyName = aliasMap.get(propertyName);
if (propertyName == null) {
continue;
}
}if (propertyNamingStrategy != null) {
propertyName = propertyNamingStrategy.translate(propertyName);
}//优先选择get
if (fieldInfoMap.containsKey(propertyName)) {
continue;
}FieldInfo fieldInfo = new FieldInfo(propertyName, method, field, clazz, null, ordinal, serialzeFeatures, parserFeatures,
annotation, fieldAnnotation, label);
fieldInfoMap.put(propertyName, fieldInfo);
}
}Field[] fields = clazz.getFields();
computeFields(clazz, aliasMap, propertyNamingStrategy, fieldInfoMap, fields);
return getFieldInfos(clazz, sorted, fieldInfoMap);
}
这里针对clazz的每一个方法进行了判断，由于只有get和set开头的函数，所以只关心methodName.startsWith("get")分支。最后进入了getPropertyNameByCompatibleFieldName所在的分支，并将propertyName设置为对应get的属性名。在getPropertyNameByCompatibleFieldName函数中，而compatibleWithFieldName设置为false所以相当于跳过了。
private static String getPropertyNameByCompatibleFieldName(Map fieldCacheMap, String methodName,
String propertyName,int fromIdx) {
if (compatibleWithFieldName){
if (!fieldCacheMap.containsKey(propertyName)){
String tempPropertyName=methodName.substring(fromIdx);
returnfieldCacheMap.containsKey(tempPropertyName)?tempPropertyName:propertyName;
}
}
return propertyName;
}
继续分析，程序进入isJSONTypeIgnore根据注解判断是否跳过该字段，我的例子中不关心。紧接着调用了getFieldFromCache：
public static Field getFieldFromCache(String fieldName, Map fieldCacheMap) {
Field field = fieldCacheMap.get(fieldName);
if (field == null) {
field = fieldCacheMap.get("_" + fieldName);
}if (field == null) {
field = fieldCacheMap.get("m_" + fieldName);
}if (field == null) {
char c0 = fieldName.charAt(0);
if (c0 >= 'a' && c0 <= 'z') {
char[] chars = fieldName.toCharArray();
chars[0] -= 32;
// lower
String fieldNameX = new String(chars);
field = fieldCacheMap.get(fieldNameX);
}
}return field;
}
这里按照刚才取出的方法名来查找字段，如果失败则加上或者m之类的方法继续判断。返回继续分析，在做了部分如注解别名之类的处理后，将分析得到的结果生成一个FieldInfo，并保存在fieldInfoMap中。最后调用computeFields进一步处理一些public属性的fields数据。最后经过getFieldInfos处理后，将得到的List
private static void computeFields(
Class clazz, //
Map aliasMap, //
PropertyNamingStrategy propertyNamingStrategy, //
Map fieldInfoMap, //
Field[] fields) {for (Field field : fields) {
if (Modifier.isStatic(field.getModifiers())) {
continue;
}JSONField fieldAnnotation = field.getAnnotation(JSONField.class);
int ordinal = 0, serialzeFeatures = 0, parserFeatures = 0;
String propertyName = field.getName();
String label = null;
if (fieldAnnotation != null) {
if (!fieldAnnotation.serialize()) {
continue;
}ordinal = fieldAnnotation.ordinal();
serialzeFeatures = SerializerFeature.of(fieldAnnotation.serialzeFeatures());
parserFeatures = Feature.of(fieldAnnotation.parseFeatures());
if (fieldAnnotation.name().length() != 0) {
propertyName = fieldAnnotation.name();
}if (fieldAnnotation.label().length() != 0) {
label = fieldAnnotation.label();
}
}if (aliasMap != null) {
propertyName = aliasMap.get(propertyName);
if (propertyName == null) {
continue;
}
}if (propertyNamingStrategy != null) {
propertyName = propertyNamingStrategy.translate(propertyName);
}if (!fieldInfoMap.containsKey(propertyName)) {
FieldInfo fieldInfo = new FieldInfo(propertyName, null, field, clazz, null, ordinal, serialzeFeatures, parserFeatures,
null, fieldAnnotation, label);
fieldInfoMap.put(propertyName, fieldInfo);
}
}
}
分析到这里，可以发现在fieldInfoList中实际上值：name,json。看到这里相比也能猜出大概了，现在继续跟踪。回到buildBeanInfo中，将刚才得到的fieldInfoList构造为SerializeBeanInfo并返回。
public ObjectSerializer createJavaBeanSerializer(SerializeBeanInfo beanInfo) {
JSONType jsonType = beanInfo.jsonType;
if (jsonType != null) {
Class serializerClass = jsonType.serializer();
if (serializerClass != Void.class) {
try {
Object seralizer = serializerClass.newInstance();
if (seralizer instanceof ObjectSerializer) {
return (ObjectSerializer) seralizer;
}
} catch (Throwable e) {
// skip
}
}if (jsonType.asm() == false) {
asm = false;
}for (SerializerFeature feature : jsonType.serialzeFeatures()) {
if (SerializerFeature.WriteNonStringValueAsString == feature //
|| SerializerFeature.WriteEnumUsingToString == feature //
|| SerializerFeature.NotWriteDefaultValue =https://www.it610.com/article/= feature) {
asm = false;
break;
}
}
}Class clazz = beanInfo.beanType;
if (!Modifier.isPublic(beanInfo.beanType.getModifiers())) {
return new JavaBeanSerializer(beanInfo);
}boolean asm = this.asm && !fieldBased;
if (asm && asmFactory.classLoader.isExternalClass(clazz)
|| clazz == Serializable.class || clazz == Object.class) {
asm = false;
}if (asm && !ASMUtils.checkName(clazz.getSimpleName())) {
asm = false;
}if (asm) {
for(FieldInfo fieldInfo : beanInfo.fields){
Field field = fieldInfo.field;
if (field != null && !field.getType().equals(fieldInfo.fieldClass)) {
asm = false;
break;
}Method method = fieldInfo.method;
if (method != null && !method.getReturnType().equals(fieldInfo.fieldClass)) {
asm = false;
break;
}JSONField annotation = fieldInfo.getAnnotation();
if (annotation == null) {
continue;
}if ((!ASMUtils.checkName(annotation.name())) //
|| annotation.format().length() != 0
|| annotation.jsonDirect()
|| annotation.serializeUsing() != Void.class
|| annotation.unwrapped()
) {
asm = false;
break;
}for (SerializerFeature feature : annotation.serialzeFeatures()) {
if (SerializerFeature.WriteNonStringValueAsString == feature //
|| SerializerFeature.WriteEnumUsingToString == feature //
|| SerializerFeature.NotWriteDefaultValue =https://www.it610.com/article/= feature) {
asm = false;
break;
}
}
}
}if (asm) {
try {
ObjectSerializer asmSerializer = createASMSerializer(beanInfo);
if (asmSerializer != null) {
return asmSerializer;
}
} catch (ClassFormatError e) {
// skip
} catch (ClassCastException e) {
// skip
} catch (Throwable e) {
throw new JSONException("create asm serializer error, class "
+ clazz, e);
}
}return new JavaBeanSerializer(beanInfo);
}
经过处理后进入了createASMSerializer，其中调用createJavaBeanSerializer来创建具体的writer：
public JavaBeanSerializer createJavaBeanSerializer(SerializeBeanInfo beanInfo) throws Exception {
Class clazz = beanInfo.beanType;
if (clazz.isPrimitive()) {
throw new JSONException("unsupportd class " + clazz.getName());
}JSONType jsonType = clazz.getAnnotation(JSONType.class);
FieldInfo[] unsortedGetters = beanInfo.fields;
;
for (FieldInfo fieldInfo : unsortedGetters) {
if (fieldInfo.field == null //
&& fieldInfo.method != null //
&& fieldInfo.method.getDeclaringClass().isInterface()) {
return new JavaBeanSerializer(clazz);
}
}FieldInfo[] getters = beanInfo.sortedFields;
boolean nativeSorted = beanInfo.sortedFields == beanInfo.fields;
if (getters.length > 256) {
return new JavaBeanSerializer(clazz);
}for (FieldInfo getter : getters) {
if (!ASMUtils.checkName(getter.getMember().getName())) {
return new JavaBeanSerializer(clazz);
}
}String className = "ASMSerializer_" + seed.incrementAndGet() + "_" + clazz.getSimpleName();
String packageName = ASMSerializerFactory.class.getPackage().getName();
String classNameType = packageName.replace('.', '/') + "/" + className;
String classNameFull = packageName + "." + className;
ClassWriter cw = new ClassWriter();
cw.visit(V1_5 //
, ACC_PUBLIC + ACC_SUPER //
, classNameType //
, JavaBeanSerializer //
, new String[] { ObjectSerializer } //
);
for (FieldInfo fieldInfo : getters) {
if (fieldInfo.fieldClass.isPrimitive() //
//|| fieldInfo.fieldClass.isEnum() //
|| fieldInfo.fieldClass == String.class) {
continue;
}new FieldWriter(cw, ACC_PUBLIC, fieldInfo.name + "_asm_fieldType", "Ljava/lang/reflect/Type;
") //
.visitEnd();
if (List.class.isAssignableFrom(fieldInfo.fieldClass)) {
new FieldWriter(cw, ACC_PUBLIC, fieldInfo.name + "_asm_list_item_ser_",
ObjectSerializer_desc) //
.visitEnd();
}new FieldWriter(cw, ACC_PUBLIC, fieldInfo.name + "_asm_ser_", ObjectSerializer_desc) //
.visitEnd();
}MethodVisitor mw = new MethodWriter(cw, ACC_PUBLIC, "", "(" + desc(SerializeBeanInfo.class) + ")V", null, null);
mw.visitVarInsn(ALOAD, 0);
mw.visitVarInsn(ALOAD, 1);
mw.visitMethodInsn(INVOKESPECIAL, JavaBeanSerializer, "", "(" + desc(SerializeBeanInfo.class) + ")V");
// init _asm_fieldType
for (int i = 0;
i < getters.length;
++i) {
FieldInfo fieldInfo = getters[i];
if (fieldInfo.fieldClass.isPrimitive() //
//|| fieldInfo.fieldClass.isEnum() //
|| fieldInfo.fieldClass == String.class) {
continue;
}mw.visitVarInsn(ALOAD, 0);
if (fieldInfo.method != null) {
mw.visitLdcInsn(com.alibaba.fastjson.asm.Type.getType(desc(fieldInfo.declaringClass)));
mw.visitLdcInsn(fieldInfo.method.getName());
mw.visitMethodInsn(INVOKESTATIC, type(ASMUtils.class), "getMethodType",
"(Ljava/lang/Class;
Ljava/lang/String;
)Ljava/lang/reflect/Type;
");
} else {
mw.visitVarInsn(ALOAD, 0);
mw.visitLdcInsn(i);
mw.visitMethodInsn(INVOKESPECIAL, JavaBeanSerializer, "getFieldType", "(I)Ljava/lang/reflect/Type;
");
}mw.visitFieldInsn(PUTFIELD, classNameType, fieldInfo.name + "_asm_fieldType", "Ljava/lang/reflect/Type;
");
}mw.visitInsn(RETURN);
mw.visitMaxs(4, 4);
mw.visitEnd();
boolean DisableCircularReferenceDetect = false;
if (jsonType != null) {
for (SerializerFeature featrues : jsonType.serialzeFeatures()) {
if (featrues == SerializerFeature.DisableCircularReferenceDetect) {
DisableCircularReferenceDetect = true;
break;
}
}
}// 0 write
// 1 writeNormal
// 2 writeNonContext
for (int i = 0;
i < 3;
++i) {
String methodName;
boolean nonContext = DisableCircularReferenceDetect;
boolean writeDirect = false;
if (i == 0) {
methodName = "write";
writeDirect = true;
} else if (i == 1) {
methodName = "writeNormal";
} else {
writeDirect = true;
nonContext = true;
methodName = "writeDirectNonContext";
}Context context = new Context(getters, beanInfo, classNameType, writeDirect,
nonContext);
mw = new MethodWriter(cw, //
ACC_PUBLIC, //
methodName, //
"(L" + JSONSerializer
+ ";
Ljava/lang/Object;
Ljava/lang/Object;
Ljava/lang/reflect/Type;
I)V", //
null, //
new String[] { "java/io/IOException" } //
);
{
Label endIf_ = new Label();
mw.visitVarInsn(ALOAD, Context.obj);
//serializer.writeNull();
mw.visitJumpInsn(IFNONNULL, endIf_);
mw.visitVarInsn(ALOAD, Context.serializer);
mw.visitMethodInsn(INVOKEVIRTUAL, JSONSerializer,
"writeNull", "()V");
mw.visitInsn(RETURN);
mw.visitLabel(endIf_);
}mw.visitVarInsn(ALOAD, Context.serializer);
mw.visitFieldInsn(GETFIELD, JSONSerializer, "out", SerializeWriter_desc);
mw.visitVarInsn(ASTORE, context.var("out"));
if ((!nativeSorted) //
&& !context.writeDirect) {if (jsonType == null || jsonType.alphabetic()) {
Label _else = new Label();
mw.visitVarInsn(ALOAD, context.var("out"));
mw.visitMethodInsn(INVOKEVIRTUAL, SerializeWriter, "isSortField", "()Z");
mw.visitJumpInsn(IFNE, _else);
mw.visitVarInsn(ALOAD, 0);
mw.visitVarInsn(ALOAD, 1);
mw.visitVarInsn(ALOAD, 2);
mw.visitVarInsn(ALOAD, 3);
mw.visitVarInsn(ALOAD, 4);
mw.visitVarInsn(ILOAD, 5);
mw.visitMethodInsn(INVOKEVIRTUAL, classNameType,
"writeUnsorted", "(L" + JSONSerializer
+ ";
Ljava/lang/Object;
Ljava/lang/Object;
Ljava/lang/reflect/Type;
I)V");
mw.visitInsn(RETURN);
mw.visitLabel(_else);
}
}// isWriteDoubleQuoteDirect
if (context.writeDirect && !nonContext) {
Label _direct = new Label();
Label _directElse = new Label();
mw.visitVarInsn(ALOAD, 0);
mw.visitVarInsn(ALOAD, Context.serializer);
mw.visitMethodInsn(INVOKEVIRTUAL, JavaBeanSerializer, "writeDirect", "(L" + JSONSerializer + ";
)Z");
mw.visitJumpInsn(IFNE, _directElse);
mw.visitVarInsn(ALOAD, 0);
mw.visitVarInsn(ALOAD, 1);
mw.visitVarInsn(ALOAD, 2);
mw.visitVarInsn(ALOAD, 3);
mw.visitVarInsn(ALOAD, 4);
mw.visitVarInsn(ILOAD, 5);
mw.visitMethodInsn(INVOKEVIRTUAL, classNameType,
"writeNormal", "(L" + JSONSerializer
+ ";
Ljava/lang/Object;
Ljava/lang/Object;
Ljava/lang/reflect/Type;
I)V");
mw.visitInsn(RETURN);
mw.visitLabel(_directElse);
mw.visitVarInsn(ALOAD, context.var("out"));
mw.visitLdcInsn(SerializerFeature.DisableCircularReferenceDetect.mask);
mw.visitMethodInsn(INVOKEVIRTUAL, SerializeWriter, "isEnabled", "(I)Z");
mw.visitJumpInsn(IFEQ, _direct);
mw.visitVarInsn(ALOAD, 0);
mw.visitVarInsn(ALOAD, 1);
mw.visitVarInsn(ALOAD, 2);
mw.visitVarInsn(ALOAD, 3);
mw.visitVarInsn(ALOAD, 4);
mw.visitVarInsn(ILOAD, 5);
mw.visitMethodInsn(INVOKEVIRTUAL, classNameType, "writeDirectNonContext",
"(L" + JSONSerializer + ";
Ljava/lang/Object;
Ljava/lang/Object;
Ljava/lang/reflect/Type;
I)V");
mw.visitInsn(RETURN);
mw.visitLabel(_direct);
}mw.visitVarInsn(ALOAD, Context.obj);
// obj
mw.visitTypeInsn(CHECKCAST, type(clazz));
// serializer
mw.visitVarInsn(ASTORE, context.var("entity"));
// obj
generateWriteMethod(clazz, mw, getters, context);
mw.visitInsn(RETURN);
mw.visitMaxs(7, context.variantIndex + 2);
mw.visitEnd();
}if (!nativeSorted) {
// sortField support
Context context = new Context(getters, beanInfo, classNameType, false,
DisableCircularReferenceDetect);
mw = new MethodWriter(cw, ACC_PUBLIC, "writeUnsorted",
"(L" + JSONSerializer + ";
Ljava/lang/Object;
Ljava/lang/Object;
Ljava/lang/reflect/Type;
I)V",
null, new String[] { "java/io/IOException" });
mw.visitVarInsn(ALOAD, Context.serializer);
mw.visitFieldInsn(GETFIELD, JSONSerializer, "out", SerializeWriter_desc);
mw.visitVarInsn(ASTORE, context.var("out"));
mw.visitVarInsn(ALOAD, Context.obj);
// obj
mw.visitTypeInsn(CHECKCAST, type(clazz));
// serializer
mw.visitVarInsn(ASTORE, context.var("entity"));
// objgenerateWriteMethod(clazz, mw, unsortedGetters, context);
mw.visitInsn(RETURN);
mw.visitMaxs(7, context.variantIndex + 2);
mw.visitEnd();
}// 0 writeAsArray
// 1 writeAsArrayNormal
// 2 writeAsArrayNonContext
for (int i = 0;
i < 3;
++i) {
String methodName;
boolean nonContext = DisableCircularReferenceDetect;
boolean writeDirect = false;
if (i == 0) {
methodName = "writeAsArray";
writeDirect = true;
} else if (i == 1) {
methodName = "writeAsArrayNormal";
} else {
writeDirect = true;
nonContext = true;
methodName = "writeAsArrayNonContext";
}Context context = new Context(getters, beanInfo, classNameType, writeDirect,
nonContext);
mw = new MethodWriter(cw, ACC_PUBLIC, methodName,
"(L" + JSONSerializer + ";
Ljava/lang/Object;
Ljava/lang/Object;
Ljava/lang/reflect/Type;
I)V",
null, new String[] { "java/io/IOException" });
mw.visitVarInsn(ALOAD, Context.serializer);
mw.visitFieldInsn(GETFIELD, JSONSerializer, "out", SerializeWriter_desc);
mw.visitVarInsn(ASTORE, context.var("out"));
mw.visitVarInsn(ALOAD, Context.obj);
// obj
mw.visitTypeInsn(CHECKCAST, type(clazz));
// serializer
mw.visitVarInsn(ASTORE, context.var("entity"));
// obj
generateWriteAsArray(clazz, mw, getters, context);
mw.visitInsn(RETURN);
mw.visitMaxs(7, context.variantIndex + 2);
mw.visitEnd();
}byte[] code = cw.toByteArray();
Class exampleClass = classLoader.defineClassPublic(classNameFull, code, 0, code.length);
Constructor constructor = exampleClass.getConstructor(SerializeBeanInfo.class);
Object instance = constructor.newInstance(beanInfo);
return (JavaBeanSerializer) instance;
}
到这里为止，我们的分析就可以结束了，实际上这里是根据fieldInfo，通过CodeGen技术生成一个writer实例。而getJson被简单当作了json属性的getter，所以在writer.write(object)中调用了getJson从而出现了递归。那么这个问题的简单解决办法就是将getJson换个名字，比如toJson。最后，在github的issue中也翻到了一个对应的问题，作者给出的答案就是换个名字。
使用fastjson序列化this的时候栈溢出
原文出处:FastJson 踩坑记录
推荐阅读
- 做一件事情的基本原理是什么（）
- 【读书笔记】贝叶斯原理
- SG平滑轨迹算法的原理和实现
- “写作宝典”《金字塔原理》之读书笔记
- Spring|Spring 框架之 AOP 原理剖析已经出炉！！！预定的童鞋可以识别下发二维码去看了
- Spring|Spring Boot 自动配置的原理、核心注解以及利用自动配置实现了自定义 Starter 组件
- Vue源码分析—响应式原理(二)
- MYSQL主从同步的实现
- (1)redis集群原理及搭建与使用(1)
- Git学习-笔记摘要