在HDFS上使用NIO进行mapreduce(mapreduce using NIO on HDFS)

 phone和org的内容是完全不同的数据结构，并且您将无法完全反序列化为像您在示例中一样的同质格式。 最好的选择是至少部分反序列化为一个struct ： 
 
type data struct {
    Phone []map[string]string
    Org []map[string]map[string]string
}
 
 这至少会反序列化所有的数据，但它仍然有点混乱; 一片地图的地图并不是一个很好的数据结构。 这个问题并不清楚，但是如果任何一个领域都是固定的，那么您可能也希望将这些领域的类型编码，例如： 
 
type data struct {
    Phone []map[string]string
    Org []struct{
        Current map[string]string
    }
}
 
 然后您可以反序列化为这种类型并更容易使用它： 
 
var person data
json.Unmarshal([]byte(ii), &person)
fmt.Printf("%v", person.Phone)
fmt.Printf("%v", person.Org[0].Current)
 
 工作场地示例： https ： //play.golang.org/p/5W-7RzPimZj 
 
 请注意，我必须更正JSON中的错误，因"org"之前缺少逗号而无效。 

The contents of phone and org are completely different data structures, and you won't be able to cleanly deserialize both into a homogenous format like you've got in the example. The best option is to at least partially deserialize into a struct:
 
type data struct {
    Phone []map[string]string
    Org []map[string]map[string]string
}
 
This will at least deserialize all of the data, but it's still a bit messy; a slice of maps of maps is not a great data structure to work with. It's not clear from the question, but if any of the fields are fixed, you might want to codify those in types as well, for example:
 
type data struct {
    Phone []map[string]string
    Org []struct{
        Current map[string]string
    }
}
 
You can then deserialize to this type and use it much more easily:
 
var person data
json.Unmarshal([]byte(ii), &person)
fmt.Printf("%v", person.Phone)
fmt.Printf("%v", person.Org[0].Current)
 
Working playground example here: https://play.golang.org/p/5W-7RzPimZj
 
Note that I had to correct an error in the JSON, it is invalid due to a missing comma before "org".