Question

I’m hitting a problem in JSON I’m getting back from a Bugzilla server because it sometimes returns “text” : {} and sometimes “text” : “blah blah blah”. Bugzilla returns the former if no description was given for a bug. I’m mystified why it doesn’t come back as the much more sensible “text” : “” but it does and that’s it.

If I have a String named text in the target object for Gson, it objects when it sees the {} case because it says that’s an object and not a String:

Exception in thread "main" com.google.gson.JsonParseException: The 
JsonDeserializer StringTypeAdapter failed to deserialized json object {} given 
the type class java.lang.String

Any suggestions on how I can make Gson parse this?

Answer 1

Gson requires custom deserialization for the situation in the original question. Following is one such example.

input.json:

[
  {
    "text":"some text"
  },
  {
    "text":{}
  }
]

Foo.java:

import java.io.FileReader;
import java.lang.reflect.Type;

import com.google.gson.Gson;
import com.google.gson.GsonBuilder;
import com.google.gson.JsonDeserializationContext;
import com.google.gson.JsonDeserializer;
import com.google.gson.JsonElement;
import com.google.gson.JsonParseException;

public class Foo
{
  public static void main(String[] args) throws Exception
  {
    GsonBuilder gsonBuilder = new GsonBuilder();
    gsonBuilder.registerTypeAdapter(String.class, new StringDeserializer());
    Gson gson = gsonBuilder.create();
    Thing[] things = gson.fromJson(new FileReader("input.json"), Thing[].class);
    System.out.println(gson.toJson(things));
  }
}

class Thing
{
  String text;
}

class StringDeserializer implements JsonDeserializer<String>
{
  @Override
  public String deserialize(JsonElement json, Type typeOfT, JsonDeserializationContext context)
      throws JsonParseException
  {
    if (json.isJsonPrimitive()) return json.getAsString();
    return "";
  }
}

Output:

[{"text":"some text"},{"text":""}]

Using instead a custom deserializer for the Thing.class type would of course be possible. This would have the benefit of not adding additional processing for every String, but then you’d be stuck with “manual” processing all of the other attributes of Thing.