fastjson存在乱序的问题

现象及原因

通常来讲，在使用json数据格式时一般不需要要求数据有序。但凡事都有例外，针对查询时序数据这样一个场景，就必须要求服务器端返回的数据是按时间有序的，否则前端在进行数据展示时就会有问题。
项目架构如下：

数据从OpenTSDB中查询出来的时候是有序的：

[{
    "metrc":"cpu.usage",
    "dps": {
        "123456": 12,
        "123457": 13,
        "123458": 23,
        "123459": 32
    }
}]

执行如下操作：

JSONObject.parseArray(json)

结果查看对应的JSON数组中的map数据是乱序的，可能的结果如下：

[{
    "metrc":"cpu.usage",
    "dps": {
        "123457": 13,
        "123456": 12,
        "123459": 32,
        "123458": 23
    }
}]

原本希望时序数据是按时间Key有序的，但是经过fastjson解析之后就会出现Key乱序。实际上，这个问题是fastjson本身的bug，详见：https://github.com/alibaba/fastjson/issues/660 。

解决办法

如下以解析从OpenTSDB中查询返回的时序数据为例。

1.升级fastjson版本

fastjson从1.2.3版本开始，在解析json对象时可以指定Feature.OrderedField参数，这样解析的结果就不会乱序。

public static void main(String[] args) {
    // 模拟从OpenTSDB中查询返回的时序数据
    String str = "[{"metric":"temperature","tags":{"device_id":"device-12312-14","dt_name":"dsdsdsd"},"aggregateTags":[],"dps":{"1538210186542":30,"1538210191574":83,"1538210196597":41,"1538210201624":56,"1538210206654":20,"1538210211677":25,"1538210216700":54,"1538210221740":36,"1538210226773":89,"1538210231813":8,"1538210236847":34,"1538210241882":83,"1538210246916":96,"1538210251952":42,"1538210257002":6,"1538210262038":87,"1538210267076":19,"1538210272108":44,"1538210277139":84,"1538210282176":41,"1538210287216":57,"1538210292254":26,"1538210297283":64}}]";

    // 直接使用fastjson的接口实现有序解析
    JSONArray array = JSONArray.parseObject(str.getBytes(), JSONArray.class, Feature.OrderedField);
    System.out.println(array.getJSONObject(0).getJSONObject("dps").toJSONString());
}

实际上，追踪一下fastjson的实现源码发现，当传递参数Feature.OrderedField时，底层正是使用LinkedHashMap来实现Key有序的（LinkedHashMap是按插入顺序排序）：

public JSONObject(int initialCapacity, boolean ordered){
	if (ordered) {
		// 使用LinkedHashMap保证json对象的key是按照插入顺序有序的
    	map = new LinkedHashMap<String, Object>(initialCapacity);
	} else {
		map = new HashMap<String, Object>(initialCapacity);
	}
}

2.手动排序

除了可以直接通过fastjson的接口在解析时就实现有序，还可以对解析结果进行手动排序。

public static void main(String[] args) {
    // 模拟从OpenTSDB中查询返回的时序数据
    String str = "[{"metric":"temperature","tags":{"device_id":"device-12312-14","dt_name":"dsdsdsd"},"aggregateTags":[],"dps":{"1538210186542":30,"1538210191574":83,"1538210196597":41,"1538210201624":56,"1538210206654":20,"1538210211677":25,"1538210216700":54,"1538210221740":36,"1538210226773":89,"1538210231813":8,"1538210236847":34,"1538210241882":83,"1538210246916":96,"1538210251952":42,"1538210257002":6,"1538210262038":87,"1538210267076":19,"1538210272108":44,"1538210277139":84,"1538210282176":41,"1538210287216":57,"1538210292254":26,"1538210297283":64}}]";

    // 解析之后手动排序
    JSONArray array = JSONArray.parseArray(str);
    System.out.println(array.toJSONString());
    JSONObject json = array.getJSONObject(0);
    // 不传递参数Feature.OrderedField时解析得到的json对象key是无序的，本质上是一个HashMap结构
    Map<String, Object> map = json.getJSONObject("dps").getInnerMap();
    // 通过TreeMap对Key进行排序
    map = sortMapByKey(map);
    JSONObject dps = new JSONObject();
    dps.put("dps", map);
    System.out.println(dps.toJSONString());
}

private static Map<String, Object> sortMapByKey(Map<String, Object> map) {
    if (map == null || map.isEmpty()) {
        return null;
    }
    Map<String, Object> sortMap = new TreeMap<String, Object>(new MapKeyComparator());
    sortMap.putAll(map);
    return sortMap;
}

private static class MapKeyComparator implements Comparator<String> {
    @Override
    public int compare(String str1, String str2) {
        return str1.compareTo(str2);
    }
} 

【参考】
https://dzone.com/articles/hashmap-vs-treemap-vs HashMap vs. TreeMap vs. HashTable vs. LinkedHashMap