如何编写一个JSON解析器

dy223 9年前

编写一个JSON解析器实际上就是一个函数,它的输入是一个表示JSON的字符串,输出是结构化的对应到语言本身的数据结构。

和XML相比,JSON本身结构非常简单,并且仅有几种数据类型,以Java为例,对应的数据结构是:

  • "string":Java的String;
  • number:Java的Long或Double;
  • true/false:Java的Boolean;
  • null:Java的null;
  • [array]:Java的List<Object>或Object[];
  • {"key":"value"}:Java的Map<String, Object>。

解析JSON和解析XML类似,最终都是解析为内存的一个对象。出于效率考虑,使用流的方式几乎是唯一选择,也就是解析器只从头扫描一遍JSON字符串,就完整地解析出对应的数据结构。

本质上解析器就是一个状态机,只要按照JSON定义的格式(参考http://www.json.org,正确实现状态转移即可。但是为了简化代码,我们也没必要完整地实现一个字符一个字符的状态转移。

解析器的输入应该是一个字符流,所以,第一步是获得Reader,以便能不断地读入下一个字符。

在解析的过程中,我们经常要根据下一个字符来决定状态跳转,此时又涉及到回退的问题,就是某些时候不能用next()取下一个字符,而是用peek()取下一个字符,但字符流的指针不移动。所以,Reader接口不能满足这个需求,应当进一步封装一个CharReader,它可以实现:

  • char next():读取下一个字符,移动Reader指针;
  • char peek():读取下一个字符,不移动Reader指针;
  • String next(int size):读取指定的N个字符并移动指针;
  • boolean hasMore():判断流是否结束。

JSON解析比其他文本解析要简单的地方在于,任何JSON数据类型,只需要根据下一个字符即可确定,仔细总结可以发现,如果peek()返回的字符是某个字符,就可以期望读取的数据类型:

  • {:期待一个JSON object;
  • ::期待一个JSON object的value;
  • ,:期待一个JSON object的下一组key-value,或者一个JSON array的下一个元素;
  • [:期待一个JSON array;
  • t:期待一个true;
  • f:期待一个false;
  • n:期待一个null;
  • ":期待一个string;
  • 0~9:期待一个number。

但是单个字符要匹配的状态太多了,需要进一步把字符流变为Token,可以总结出如下几种Token:

  • END_DOCUMENT:JSON文档结束;
  • BEGIN_OBJECT:开始一个JSON object;
  • END_OBJECT:结束一个JSON object;
  • BEGIN_ARRAY:开始一个JSON array;
  • END_ARRAY:结束一个JSON array;
  • SEP_COLON:读取一个冒号;
  • SEP_COMMA:读取一个逗号;
  • STRING:一个String;
  • BOOLEAN:一个true或false;
  • NUMBER:一个number;
  • NULL:一个null。

然后,将CharReader进一步封装为TokenReader,提供以下接口:

  • Token readNextToken():读取下一个Token;
  • boolean readBoolean():读取一个boolean;
  • Number readNumber():读取一个number;
  • String readString():读取一个string;
  • void readNull():读取一个null。

由于JSON的Object和Array可以嵌套,在读取过程中,使用一个栈来存储Object和Array是必须的。每当我们读到一个BEGIN_OBJECT时,就创建一个Map并压栈;每当读到一个BEGIN_ARRAY时,就创建一个List并压栈;每当读到一个END_OBJECT和END_ARRAY时,就弹出栈顶元素,并根据新的栈顶元素判断是否压栈。此外,读到Object的Key也必须压栈,读到后面的Value后将Key-Value压入栈顶的Map。

如果读到END_DOCUMENT时,栈恰好只剩下一个元素,则读取正确,将该元素返回,读取结束。如果栈剩下不止一个元素,则JSON文档格式不正确。

最后,JsonReader的核心解析代码parse()就是负责从TokenReader中不断读取Token,根据当前状态操作,然后设定下一个Token期望的状态,如果与期望状态不符,则JSON的格式无效。起始状态被设定为STATUS_EXPECT_SINGLE_VALUE | STATUS_EXPECT_BEGIN_OBJECT | STATUS_EXPECT_BEGIN_ARRAY,即期望读取到单个value、{或[。循环的退出点是读取到END_DOCUMENT时。

public class JsonReader {        TokenReader reader;        public Object parse() {          Stack stack = new Stack();          int status = STATUS_EXPECT_SINGLE_VALUE | STATUS_EXPECT_BEGIN_OBJECT | STATUS_EXPECT_BEGIN_ARRAY;          for (;;) {              Token currentToken = reader.readNextToken();              switch (currentToken) {              case BOOLEAN:                  if (hasStatus(STATUS_EXPECT_SINGLE_VALUE)) {                      // single boolean:                      Boolean bool = reader.readBoolean();                      stack.push(StackValue.newJsonSingle(bool));                      status = STATUS_EXPECT_END_DOCUMENT;                      continue;                  }                  if (hasStatus(STATUS_EXPECT_OBJECT_VALUE)) {                      Boolean bool = reader.readBoolean();                      String key = stack.pop(StackValue.TYPE_OBJECT_KEY).valueAsKey();                      stack.peek(StackValue.TYPE_OBJECT).valueAsObject().put(key, bool);                      status = STATUS_EXPECT_COMMA | STATUS_EXPECT_END_OBJECT;                      continue;                  }                  if (hasStatus(STATUS_EXPECT_ARRAY_VALUE)) {                      Boolean bool = reader.readBoolean();                      stack.peek(StackValue.TYPE_ARRAY).valueAsArray().add(bool);                      status = STATUS_EXPECT_COMMA | STATUS_EXPECT_END_ARRAY;                      continue;                  }                  throw new JsonParseException("Unexpected boolean.", reader.reader.readed);                case NULL:                  if (hasStatus(STATUS_EXPECT_SINGLE_VALUE)) {                      // single null:                      reader.readNull();                      stack.push(StackValue.newJsonSingle(null));                      status = STATUS_EXPECT_END_DOCUMENT;                      continue;                  }                  if (hasStatus(STATUS_EXPECT_OBJECT_VALUE)) {                      reader.readNull();                      String key = stack.pop(StackValue.TYPE_OBJECT_KEY).valueAsKey();                      stack.peek(StackValue.TYPE_OBJECT).valueAsObject().put(key, null);                      status = STATUS_EXPECT_COMMA | STATUS_EXPECT_END_OBJECT;                      continue;                  }                  if (hasStatus(STATUS_EXPECT_ARRAY_VALUE)) {                      reader.readNull();                      stack.peek(StackValue.TYPE_ARRAY).valueAsArray().add(null);                      status = STATUS_EXPECT_COMMA | STATUS_EXPECT_END_ARRAY;                      continue;                  }                  throw new JsonParseException("Unexpected null.", reader.reader.readed);                case NUMBER:                  if (hasStatus(STATUS_EXPECT_SINGLE_VALUE)) {                      // single number:                      Number number = reader.readNumber();                      stack.push(StackValue.newJsonSingle(number));                      status = STATUS_EXPECT_END_DOCUMENT;                      continue;                  }                  if (hasStatus(STATUS_EXPECT_OBJECT_VALUE)) {                      Number number = reader.readNumber();                      String key = stack.pop(StackValue.TYPE_OBJECT_KEY).valueAsKey();                      stack.peek(StackValue.TYPE_OBJECT).valueAsObject().put(key, number);                      status = STATUS_EXPECT_COMMA | STATUS_EXPECT_END_OBJECT;                      continue;                  }                  if (hasStatus(STATUS_EXPECT_ARRAY_VALUE)) {                      Number number = reader.readNumber();                      stack.peek(StackValue.TYPE_ARRAY).valueAsArray().add(number);                      status = STATUS_EXPECT_COMMA | STATUS_EXPECT_END_ARRAY;                      continue;                  }                  throw new JsonParseException("Unexpected number.", reader.reader.readed);                case STRING:                  if (hasStatus(STATUS_EXPECT_SINGLE_VALUE)) {                      // single string:                      String str = reader.readString();                      stack.push(StackValue.newJsonSingle(str));                      status = STATUS_EXPECT_END_DOCUMENT;                      continue;                  }                  if (hasStatus(STATUS_EXPECT_OBJECT_KEY)) {                      String str = reader.readString();                      stack.push(StackValue.newJsonObjectKey(str));                      status = STATUS_EXPECT_COLON;                      continue;                  }                  if (hasStatus(STATUS_EXPECT_OBJECT_VALUE)) {                      String str = reader.readString();                      String key = stack.pop(StackValue.TYPE_OBJECT_KEY).valueAsKey();                      stack.peek(StackValue.TYPE_OBJECT).valueAsObject().put(key, str);                      status = STATUS_EXPECT_COMMA | STATUS_EXPECT_END_OBJECT;                      continue;                  }                  if (hasStatus(STATUS_EXPECT_ARRAY_VALUE)) {                      String str = reader.readString();                      stack.peek(StackValue.TYPE_ARRAY).valueAsArray().add(str);                      status = STATUS_EXPECT_COMMA | STATUS_EXPECT_END_ARRAY;                      continue;                  }                  throw new JsonParseException("Unexpected char \'\"\'.", reader.reader.readed);                case SEP_COLON: // :                  if (status == STATUS_EXPECT_COLON) {                      status = STATUS_EXPECT_OBJECT_VALUE | STATUS_EXPECT_BEGIN_OBJECT | STATUS_EXPECT_BEGIN_ARRAY;                      continue;                  }                  throw new JsonParseException("Unexpected char \':\'.", reader.reader.readed);                case SEP_COMMA: // ,                  if (hasStatus(STATUS_EXPECT_COMMA)) {                      if (hasStatus(STATUS_EXPECT_END_OBJECT)) {                          status = STATUS_EXPECT_OBJECT_KEY;                          continue;                      }                      if (hasStatus(STATUS_EXPECT_END_ARRAY)) {                          status = STATUS_EXPECT_ARRAY_VALUE | STATUS_EXPECT_BEGIN_ARRAY | STATUS_EXPECT_BEGIN_OBJECT;                          continue;                      }                  }                  throw new JsonParseException("Unexpected char \',\'.", reader.reader.readed);                case END_ARRAY:                  if (hasStatus(STATUS_EXPECT_END_ARRAY)) {                      StackValue array = stack.pop(StackValue.TYPE_ARRAY);                      if (stack.isEmpty()) {                          stack.push(array);                          status = STATUS_EXPECT_END_DOCUMENT;                          continue;                      }                      int type = stack.getTopValueType();                      if (type == StackValue.TYPE_OBJECT_KEY) {                          // key: [ CURRENT ] ,}                          String key = stack.pop(StackValue.TYPE_OBJECT_KEY).valueAsKey();                          stack.peek(StackValue.TYPE_OBJECT).valueAsObject().put(key, array.value);                          status = STATUS_EXPECT_COMMA | STATUS_EXPECT_END_OBJECT;                          continue;                      }                      if (type == StackValue.TYPE_ARRAY) {                          // xx, xx, [CURRENT] ,]                          stack.peek(StackValue.TYPE_ARRAY).valueAsArray().add(array.value);                          status = STATUS_EXPECT_COMMA | STATUS_EXPECT_END_ARRAY;                          continue;                      }                  }                  throw new JsonParseException("Unexpected char: \']\'.", reader.reader.readed);                case END_OBJECT:                  if (hasStatus(STATUS_EXPECT_END_OBJECT)) {                      StackValue object = stack.pop(StackValue.TYPE_OBJECT);                      if (stack.isEmpty()) {                          // root object:                          stack.push(object);                          status = STATUS_EXPECT_END_DOCUMENT;                          continue;                      }                      int type = stack.getTopValueType();                      if (type == StackValue.TYPE_OBJECT_KEY) {                          String key = stack.pop(StackValue.TYPE_OBJECT_KEY).valueAsKey();                          stack.peek(StackValue.TYPE_OBJECT).valueAsObject().put(key, object.value);                          status = STATUS_EXPECT_COMMA | STATUS_EXPECT_END_OBJECT;                          continue;                      }                      if (type == StackValue.TYPE_ARRAY) {                          stack.peek(StackValue.TYPE_ARRAY).valueAsArray().add(object.value);                          status = STATUS_EXPECT_COMMA | STATUS_EXPECT_END_ARRAY;                          continue;                      }                  }                  throw new JsonParseException("Unexpected char: \'}\'.", reader.reader.readed);                case END_DOCUMENT:                  if (hasStatus(STATUS_EXPECT_END_DOCUMENT)) {                      StackValue v = stack.pop();                      if (stack.isEmpty()) {                          return v.value;                      }                  }                  throw new JsonParseException("Unexpected EOF.", reader.reader.readed);                case BEGIN_ARRAY:                  if (hasStatus(STATUS_EXPECT_BEGIN_ARRAY)) {                      stack.push(StackValue.newJsonArray(this.jsonArrayFactory.createJsonArray()));                      status = STATUS_EXPECT_ARRAY_VALUE | STATUS_EXPECT_BEGIN_OBJECT | STATUS_EXPECT_BEGIN_ARRAY | STATUS_EXPECT_END_ARRAY;                      continue;                  }                  throw new JsonParseException("Unexpected char: \'[\'.", reader.reader.readed);                case BEGIN_OBJECT:                  if (hasStatus(STATUS_EXPECT_BEGIN_OBJECT)) {                      stack.push(StackValue.newJsonObject(this.jsonObjectFactory.createJsonObject()));                      status = STATUS_EXPECT_OBJECT_KEY | STATUS_EXPECT_BEGIN_OBJECT | STATUS_EXPECT_END_OBJECT;                      continue;                  }                  throw new JsonParseException("Unexpected char: \'{\'.", reader.reader.readed);              }          }      }  }

详细源码请参考:https://github.com/michaelliao/jsonstream
来自:http://www.liaoxuefeng.com/article/0014211269349633dda29ee3f29413c91fa65c372585f23000