JavaScript JSON Cookbook

You'll need a way to edit the JavaScript and run it in your browser. In this example, and nearly all examples in this book, we'll use Google Chrome for this. You can download Google Chrome at https://www.google.com/chrome/browser. Once you install Google Chrome, you'll want to activate the JavaScript console by clicking on the Customize and control Doodle Chrome icon on the right-hand side, which looks like this:

Then, you'll want to go to More Tools | JavaScript console. You should see a JavaScript console on the side of the web page, like this:

If you prefer key commands, you can also use Ctrl + Shift + J on Windows and Linux, or control + option + J on a Macintosh.

From here, you can enter JavaScript on the lower right-hand corner and press Enter (return on a Mac OS X system) to evaluate the JavaScript.

Modern web browsers, such as Chrome, define a JSON object in the JavaScript runtime that can convert the string data containing JSON to JavaScript objects, and convert a JavaScript object to JSON. Here's a simple example:

>var json = '{"call":"KF6GPE","type":"l","time":
"1399371514","lasttime":"1418597513","lat":37.17667,"lng":
-122.14650,"result" : "ok" }';
<- "{ "call":"KF6GPE","type":"l","time":"1399371514",
"lasttime":"1418597513","lat":37.17667,"lng":-122.14650,
"result" : "ok" }"
>var object = JSON.parse(json);
<- Object {call:"KF6GPE",type:"l",time:"1399371514",
lasttime:"1418597513",lat:37.17667, lng:-122.14650,result: "ok"}
> object.result
<- "ok"
>var newJson = JSON.stringify(object);
<- "{ "call":"KF6GPE","type":"l","time":"1399371514",
"lasttime":"1418597513","lat": 37.17667,"lng": -122.14650,
"result" : "ok" }"

Chrome and other modern web browsers define the JSON object, which has methods to convert between strings containing JSON and JavaScript objects.

In the previous example, we begin by setting the value of the json variable to a simple JSON expression consisting of one attribute result with the value ok. The JavaScript interpreter returns the resulting value of the variable json.

The next line uses the JSON method parse to convert the JSON string referenced by json into a JavaScript object:

>var object = JSON.parse(json);
<- Object { call:"KF6GPE", type:"l", time:"1399371514", lasttime:"1418597513", lat:37.17667, lng:-122.14650, result: "ok"}

You can then access any of the values in the object, just as you would any other JavaScript object; it is, after all, just an object:

> object.result;
<- "ok"

Finally, if you need to convert an object to JSON, you can do that with the JSON method stringify:

>var newJson = JSON.stringify(object);
<- "{ "call":"KF6GPE","type":"l","time":"1399371514",
"lasttime":"1418597513","lat": 37.17667,"lng": -122.14650,
"result" : "ok" }"

You should know two things about these methods. First of all, parse will throw an exception if the JSON you pass is malformed, or isn't JSON at all:

>JSON.parse('{"result" = "ok" }')
<- VM465:2 Uncaught SyntaxError: Unexpected token =

The errors aren't very helpful but better than nothing if you're debugging JSON sent by a less-than-fully compliant and debugged JSON encoder.

Second, very old web browsers may not have a JSON object with these methods. In that case, you can use the JavaScript function eval after wrapping the JSON in parenthesis, like this:

>eval('('+json+')')
<- Object {result: "ok"}

The eval function evaluates the string you pass as JavaScript, and the JSON notation is really just a subset of JavaScript. However, you should avoid using eval whenever you can for a few reasons. First, it's often slower than the methods provided by the JSON object. Second, it's not safe; your string might contain malicious JavaScript that can crash or otherwise subvert your JavaScript application, which is not a threat you should take lightly. Use the JSON object whenever it's available. Third, you can use the parse and stringify methods to handle simple values, such as Booleans, numbers, and strings; you're not limited to the key-value pairs in the previous example. If all I wanted to do was pass a Boolean (such as "the transaction succeeded!"), I might just write the following:

var jsonSuccess = 'true';
<- "true"
> var flag = JSON.parse(jsonSuccess);

Finally, it's worth pointing out that both the parse and stringify methods to JSON take an optional replacer function, which is invoked on every key and value in the object being serialized or deserialized. You can use this function to perform on-the-fly data conversions as the JSON is being parsed; for example, you can use it to convert between the string representation of a date and the number of seconds since midnight at the start of the epoch, or to correct the capitalization of strings. I could use a replacer function for either side of the transformation, as shown in the following code, to make the call field lowercase:

> var object = JSON.parse(json, function(k, v) {
  if ( k == 'call') return v.toLowerCase();
});
<- Object { call:"kf6gpe", type:"l", time:"1399371514",
lasttime:"1418597513", lat:37.17667, lng:-122.14650, result: "ok"}

You can also return undefined to remove an item from the results; to omit the type field from the JSON I generate, I can execute the following:

> var newJson = JSON.stringify(object, function (k, v) {
  if k == 'type') return undefined;
});
<- "{ "call":"KF6GPE","time":"1399371514","lasttime":
"1418597513","lat": 37.17667,"lng": -122.14650, "result" : "ok" 
}"

To use JsonCpp, you need to first go to the website and download the zip file with the entire library. Once you do so, you need to integrate it with your application's source code.

How you integrate it with your application's source code differs from platform to platform, but the general process is this:

Once you do this, you should have access to the JsonCpp interface in any file that includes the json/json.h header.

Here's a simple C++ application that uses JsonCpp to convert between std::string containing some simple JSON and a JSON object:

#include <string>
#include <iostream>
#include "json/json.h"

using namespace std;

int main(int argc, _TCHAR* argv[])
{
  Json::Reader reader;
  Json::Value root;

  string json = "{\"call\": \"KF6GPE\",\"type\":\"l\",\"time\":
  \"1399371514\",\"lasttime\":\"1418597513\",\"lat\": 37.17667,
  \"lng\": -122.14650,\"result\":\"ok\"}";

  bool parseSuccess = reader.parse(json, root, false);

  if (parseSuccess)
  {
    const Json::Value resultValue = root["result"];
    cout << "Result is " << resultValue.asString() << "\n";
  }

  Json::StyledWriter styledWriter;
  Json::FastWriter fastWriter;
  Json::Value newValue;
  newValue["result"] = "ok";

  cout << styledWriter.write(newValue) << "\n";
  cout << fastWriter.write(newValue) << "\n";

  return 0;
}

This example begins by including the necessary includes, including json/json.h, which defines the interface to JsonCpp. We explicitly reference the std namespace for brevity, although don't do so for the Json namespace, in which JsonCpp defines all of its interfaces.

The JsonCpp implementation defines Json::Reader and Json::Writer, specifying the interfaces to JSON readers and writers, respectively. In practice, the Json::Reader interface is also the implementation of a JSON class that can read JSON, returning its values as Json::Value. The Json::Writer variable just defines an interface; you'll want to use a subclass of it such as Json::FastWriter or Json::StyledWriter to create JSON from Json::Value objects.

The previous listing begins by defining Json::Reader and Json::Value; we'll use the reader to read the JSON we define on the next line and store its value in the Json::Value variable root. (Presumably your C++ application would get its JSON from another source, such as a web service or local file.)

Parsing JSON is as simple as calling the reader's parse function, passing the JSON and Json::Value into which it will write the JSON values. It returns a Boolean, which will be true if the JSON parsing succeeds.

The Json::Value class represents the JSON object as a tree; individual values are referenced by the attribute name in the original JSON, and the values are the values of those keys, accessible through methods such as asString, which returns the value of the object as a native C++ type. These methods of Json::Value includes the following:

In addition, the class provides operator[], letting you access array elements.

You can also query a Json::Value object to determine its type using one of these methods:

At any rate, our code uses asString to fetch the std::string value encoded as the result attribute, and writes it to the console.

The code then defines Json::StyledWriter and Json::FastWriter to create some pretty-printed JSON and unformatted JSON in strings, as well as a single Json::Value object to contain our new JSON. Assigning content to the JSON value is simple because it overrides the operator[] and operator[]= methods with the appropriate implementations to convert standard C++ types to JSON objects. So, the following line of code creates a single JSON attribute/value pair with the attribute set to result, and the value set to ok (although this code doesn't show it, you can create trees of JSON attribute-value pairs by assigning JSON objects to other JSON objects):

newValue["result"] = "ok";

We first use StyledWriter and then FastWriter to encode the JSON value in newValue, writing each string to the console.

Of course, you can also pass single values to JsonCpp; there's no reason why you can't execute the following code if all you wanted to do was pass a double-precision number:

Json::Reader reader;
Json::Value piValue;

string json = "3.1415";
bool parseSuccess = reader.parse(json, piValue, false);
  double pi = piValue.asDouble();

For the documentation for JsonCpp, you can install doxygen from http://www.stack.nl/~dimitri/doxygen/ and run it over the doc folder of the main JsonCpp distribution.

There are other JSON conversion implementations for C++, too. For a complete list, see the list at http://json.org/.

This example uses the built-in JSON serializer and deserializer in the System.Web.Extensions assembly, one of the many .NET libraries that are available. If you've installed a recent version of Visual Studio (see https://www.visualstudio.com/en-us/downloads/visual-studio-2015-downloads-vs.aspx), it should be available. All you need to do to use this assembly is include it in the assemblies your application references in Visual Studio by right-clicking the References item in your project, choosing Add Reference, and scrolling down to System.Web.Extensions in the Framework Assemblies list.

Here's a simple application that deserializes some JSON, as a dictionary of attribute-object pairs:

using System;
using System.Collections.Generic;
using System.Web.Script.Serialization;

namespace JSONExample
{
    public class SimpleResult
    {
        public string result;
    }

    class Program
    {
        static void Main(string[] args)
        {
            JavaScriptSerializer serializer = 
            new System.Web.Script.Serialization.
            JavaScriptSerializer();

            string json = @"{ ""call"":""KF6GPE"",""type"":
""l"",""time"":""1399371514"",""lasttime"":""1418597513"",
""lat"": 37.17667,""lng\": -122.14650,""result"": ""ok"" }";

dynamic result = serializer.DeserializeObject(json);
            foreach (KeyValuePair<string, object> entry in result)
            {
                var key = entry.Key;
                var value = entry.Value as string;
Console.WriteLine(String.Format("{0} : {1}", 
key, value));
            }
            Console.WriteLine(serializer.Serialize(result));

            var anotherResult = new SimpleResult { result="ok" };
            Console.WriteLine(serializer.Serialize(
            anotherResult));
        }
    }
}

The System.Web.Extensions assembly provides the JavaScriptSerializer class in the System.Web.Script.Serialization namespace. This code begins by defining a simple class, SimpleResult, which we'll encode as JSON in our example.

The Main method first defines a JavaScriptSerializer instance, and then string containing our JSON. Parsing the JSON is as easy as calling the JavaScriptSerializer instance's DeserializeObject method, which returns an object whose type is determined at run-time based on the JSON you pass.

DeserializeObject returns a Dictionary of key-value pairs; the keys are the attributes in the JSON, and the values are objects representing the values of those attributes. In our example, we simply walk the keys and values in the dictionary, printing each. Because we know the type of the value in the JSON, we can simply cast it to the appropriate type (string, in this case) using the C# as keyword; if it wasn't string, we'd receive the value null. You can use as or the type inference of C# to determine the type of unknown objects in your JSON, making it easy to parse JSON for which you lack strict semantics.

The JavaScriptSerializer class also includes a Serialize method; you can either pass it as a dictionary of attribute-value pairs, as we do with our deserialized result, or you can pass it as an instance of a C# class. If you pass it as a class, it'll attempt to serialize the class by introspecting the class fields and values.

The JSON implementation that Microsoft provides is adequate for many purposes, but not necessarily the best for your application. Other developers have implemented better ones that typically use the same interface as the Microsoft implementation. One good choice is Newtonsoft's Json.NET, which you can get at http://json.codeplex.com/ or from NuGet in Visual Studio. It supports a wider variety of .NET platforms (including Windows Phone), LINQ queries, XPath-like queries against the JSON, and is faster than the Microsoft implementation. Using it is similar to using the Microsoft implementation: install the package from the Web or NuGet, add a reference of the assembly to your application, and then use the JsonSerializer class in the NewtonSoft.Json namespace. It defines the same SerializeObject and DeserializeObject methods that the Microsoft implementation does, making switching to this library easy. James Newton-King, the author of Json.NET, makes it available under the MIT license.

As with other languages, you can also carry primitive types through the deserialization and serialization process. For example, after evaluating the following code, the resulting dynamic variable piResult will contain a floating-point number, 3.14:

string piJson = "3.14";
dynamic piResult = serializer.DeserializeObject(piJson);

As I previously hinted, you can do this in a type-safe manner; we'll discuss more of this in Chapter 7, Using JSON in a Type-safe Manner. You'll do this using the generic method DeserializeObject<>, passing a type variable of the type you want to deserialize into.

First, you'll need to get Gson; you can do this by doing a read-only checkout of the repository using SVN over HTTP with SVN by using the following command:

svn checkout http://google-gson.googlecode.com/svn/trunk/google-gson-read-only

Of course, this assumes that you have a Java development kit (http://www.oracle.com/technetwork/java/javase/downloads/index.html) and SVN (TortoiseSVN is a good client for Windows available at http://tortoisesvn.net/downloads.html) installed. Many Java IDEs include support for SVN.

Once you check out the code, follow the instructions that come with it to build the Gson JAR file, and add the JAR file to your project.

To begin, you need to create a com.google.gson.Gson object. This class defines the interface you'll use to convert between JSON and Java:

Gson gson = new com.google.gson.Gson(); 
String json = "{\"call\": \"KF6GPE\", \"type\": \"l\", \"time\":
\"1399371514\", \"lasttime\": \"1418597513\", \"lat\": 37.17667,
\"lng\": -122.14650,\"result\":\"ok\"}";
com.google.gson.JsonObject result = gson.fromJson(json, 
JsonElement.class).getAsJsonObject(); 

The JsonObject class defines the top-level object for containing a JSON object; you use its get and add methods to get and set attributes, like this:

JsonElement result = result.get("result").getAsString();

The Gson library uses the JsonElement class to encapsulate a single JSON value; it has the following methods that let you get the value contained in JsonElement as a plain Java type:

You can also learn about the type in JsonElement using one of the following methods:

You can also convert instances of your classes directly to JSON, writing something like this:

public class SimpleResult {
    public String result;
}

// Elsewhere in your code…
Gson gson = new com.google.gson.Gson(); 
SimpleResult result = new SimpleResult;
result.result = "ok";
String json = gson.toJson(result);	

This defines a class SimpleResult, which we use to create a single instance, and then use the Gson object instance to convert to a string containing the JSON using the Gson method toJson.

Finally, because JsonElement encapsulates a single value, you can also handle simple values expressed in JSON, like this:

Gson gson = new com.google.gson.Gson(); 
String piJson = "3.14";
double result = gson.fromJson(piJson, 
JsonElement.class).getAsDouble(); 

This converts the primitive value 3.14 in JSON to a Java double.

Like the C# example, you can convert directly from JSON to a plain old Java object (POJO) in a type-safe manner. You'll see how to do this in Chapter 7, Using JSON in a Type-safe Manner.

There are other JSON conversion implementations for Java, too. For a complete list, see the list at http://json.org/.

To begin with, download the JSON module from CPAN and install it. Typically, you'll download the file, unpack it, and then run the following code on a system that already has Perl and make configured:

perl Makefile.PL 
make 
make install

Here's a simple example:

use JSON;
use Data::Dumper;
my $json = '{ "call":"KF6GPE","type":"l","time":"1399371514",
"lasttime":"1418597513","lat": 37.17667,"lng": -122.14650,
"result" : "ok" }';
my %result = decode_json($json);
print Dumper(result);
print encode_json(%result);

Let's look at the interface the JSON module provides.

The CPAN module defines the decode_json and encode_json methods to decode and encode JSON respectively. These methods interconvert between Perl objects, such as literal values and associative arrays, and strings containing JSON respectively.

The code begins by importing the JSON and Data::Dumper modules. Next, it defines a single string, $json, which contains the JSON we want to parse.

With the JSON in $json, we define %result to be the associative array containing the objects defined in the JSON, and dump the values in the hash on the next line.

Finally, we re-encode the hash as JSON and print the results to the terminal.

For more information and to download the JSON CPAN module, visit https://metacpan.org/pod/JSON.

Simply enter the following in your source code to be able to reference the JSON facility:

import json

Here's a simple example from the Python interpreter:

>>> import json
>>>json = '{ "call":"KF6GPE","type":"l","time":"1399371514",
"lasttime":"1418597513","lat": 37.17667,"lng": -122.14650,
"result" : "ok" }'
u'{"call":"KF6GPE","type":"l","time":"1399371514",
"lasttime":"1418597513","lat": 37.17667,"lng": -122.14650,
"result": "ok" }'
>>>result = json.loads(json)
{u'call':u'KF6GPE',u'type':u'l',u'time':u'1399371514',
u'lasttime':u'1418597513',u'lat': 37.17667,u'lng': -122.14650,u'result': u'ok'}
>>> result['result']
u'ok'
>>> print json.dumps(result)
{"call":"KF6GPE","type":"l","time":"1399371514",
"lasttime":"1418597513","lat": 37.17667,"lng": -122.14650,
"result":"ok"}
>>> print json.dumps(result, 
...                  indent=4)
{
"call":"KF6GPE",
"type":"l",
"time":"1399371514",
"lasttime":"1418597513",
"lat": 37.17667,
"lng": -122.14650,
    "result": "ok"
}

Let's look at loads and dumps further.

Python has great support for associative arrays through its object hierarchy. The json module offers a json object with loads and dumps method that convert from JSON in text strings to associative arrays, and from associative arrays to JSON in text strings. If you're familiar with the Python marshal and pickle modules, the interface is similar; you use the loads method to obtain a Python object from its JSON representation and the dumps method to convert an object into its JSON equivalent.

The previous listing does just this. It defines a variable j that contains our JSON, and then obtains a Python object result using json.loads. Fields in the JSON are accessible as named objects in the resulting Python object. (Note that we can't call our JSON string json because it would shadow the definition of the interface to the module.)

To convert to JSON, we use the json.dumps method. By default, dumps creates a compact machine-readable version of JSON with minimum whitespace; this is best used for over-the-wire transmissions or for storage in a file. When you're debugging your JSON, it helps to pretty-print it with indentation and some whitespace around separators; you can do this using the optional indent and separators arguments. The indent argument indicates the number of spaces that each successive nested object should be indented in the string, and separators indicates the separators between each object and between each attribute and value.

For more documentation on the json module, see the Python documentation at https://docs.python.org/2/library/json.html.