Klaxon is a library to parse JSON in Kotlin.
repositories {
jcenter()
}
dependencies {
implementation 'com.beust:klaxon:5.0.1'
}
Join the #klaxon
Slack channel.
Klaxon has different API's depending on your needs:
- An object binding API to bind JSON documents directly to your objects, and vice versa.
- A streaming API to process your JSON documents as they're being read.
- A low level API to manipulate JSON objects and use queries on them.
- A JSON path query API to extract specific parts of your JSON document while streaming.
These four API's cover various scenarios and you can decide which one to use based on whether you want to stream your document and whether you need to query it.
Streaming | Query | Manipulation | |
---|---|---|---|
Object binding API | No | No | Kotlin objects |
Streaming API | Yes | No | Kotlin objects and JsonObject/JsonArray |
Low level API | No | Yes | Kotlin objects |
JSON Path query API | Yes | Yes | JsonObject/JsonArray |
To use Klaxon's high level API, you define your objects inside a class. Klaxon supports all the classes you can define in Kotlin:
- Regular and
data
classes. - Mutable and immutable classes.
- Classes with default parameters.
For example:
class Person(val name: String, val age: Int)
Classes with default parameters are supported as well:
class Person (val name: String, var age: Int = 23)
Once you've specified your value class, you invoke the parse()
function, parameterized with that class:
val result = Klaxon()
.parse<Person>("""
{
"name": "John Smith",
}
""")
assert(result?.name == "John Smith")
assert(result.age == 23)
The @Json
annotation allows you to customize how the mapping between your JSON documents and
your Kotlin objects is performed. It supports the following attributes:
Use the name
attribute when your Kotlin property has a different name than the field found in your
JSON document:
data class Person(
@Json(name = "the_name")
val name: String
)
val result = Klaxon()
.parse<Person>("""
{
"the_name": "John Smith", // note the field name
"age": 23
}
""")
assert(result.name == "John Smith")
assert(result.age == 23)
You can set this boolean attribute to true
if you want certain properties of your value class not to be
mapped during the JSON parsing process. This is useful if you defined additional properties in your value classes.
class Ignored(val name: String) {
@Json(ignored = true)
val actualName: String get() = ...
}
In this example, Klaxon will not try to find a field called actualName
in your JSON document.
Note that you can achieve the same result by declaring these properties private
:
class Ignored(val name: String) {
private val actualName: String get() = ...
}
Additionally, if you want to declare a property private
but still want that property to be visible to
Klaxon, you can annotate it with @Json(ignored = false)
.
The index
attribute allows you to specify where in the JSON string the key should appear. This allows you to
specify that certain keys should appear before others:
class Data(
@Json(index = 1) val id: String,
@Json(index = 2) val name: String
)
println(Klaxon().toJsonString(Data("id", "foo")))
// displays { "id": "id", "name": "foo" }
whereas
class Data(
@Json(index = 2) val id: String,
@Json(index = 1) val name: String
)
println(Klaxon().toJsonString(Data("id", "foo")))
// displays { "name": "foo" , "id": "id" }
Properties that are not assigned an index will be displayed in a non deterministic order in the output JSON.
On top of using the @Json(name=...)
annotation to rename fields, you can implement a field renamer yourself that
will be applied to all the fields that Klaxon encounters, both to and from JSON. You achieve this result by passing an
implementation of the FieldRenamer
interface to your Klaxon
object:
val renamer = object: FieldRenamer {
override fun toJson(fieldName: String) = FieldRenamer.camelToUnderscores(fieldName)
override fun fromJson(fieldName: String) = FieldRenamer.underscoreToCamel(fieldName)
}
val klaxon = Klaxon().fieldRenamer(renamer)
Klaxon will do its best to initialize the objects with what it found in the JSON document but you can take control of this mapping yourself by defining type converters.
The converter interface is as follows:
interface Converter {
fun canConvert(cls: Class<*>) : Boolean
fun toJson(value: Any): String
fun fromJson(jv: JsonValue) : Any
}
You define a class that implements this interface and implement the logic that converts your class to and from JSON.
For example, suppose you receive a JSON document with a field that can either be a 0
or a 1
and you want to
convert that field into your own type that's initialized with a boolean:
class BooleanHolder(val flag: Boolean)
val myConverter = object: Converter {
override fun canConvert(cls: Class<*>)
= cls == BooleanHolder::class.java
override fun toJson(value: Any): String
= """{"flag" : "${if ((value as BooleanHolder).flag == true) 1 else 0}"""
override fun fromJson(jv: JsonValue)
= BooleanHolder(jv.objInt("flag") != 0)
}
Next, you declare your converter to your Klaxon
object before parsing:
val result = Klaxon()
.converter(myConverter)
.parse<BooleanHolder>("""
{ "flag" : 1 }
""")
assert(result.flag)
The Converter
type passes you an instance of the JsonValue
class.
This class is a container of a Json value. It
is guaranteed to contain one and exactly one of either a number, a string, a character, a JsonObject
or a JsonArray
.
If one of these fields is set, the others are guaranteed to be null
. Inspect that value in your converter to make
sure that the value you are expecting is present, otherwise, you can cast a KlaxonException
to report the invalid
JSON that you just found.
It's sometimes useful to be able to specify a type conversion for a specific field without that conversion applying to all types of your document (for example, your JSON document might contain various dates of different formats). You can use field conversion types for this kind of situation.
Such fields are specified by your own annotation, which you need to specify as targetting a FIELD
:
@Target(AnnotationTarget.FIELD)
annotation class KlaxonDate
Next, annotate the field that requires this specific handling in the constructor of your class. Do note that such
a constructor needs to be annotated with @JvmOverloads
:
class WithDate @JvmOverloads constructor(
@KlaxonDate
val date: LocalDateTime
)
Define your type converter (which has the same type as the converters defined previously). In this case, we
are converting a String
from JSON into a LocalDateTime
:
val dateConverter = object: Converter {
override fun canConvert(cls: Class<*>)
= cls == LocalDateTime::class.java
override fun fromJson(jv: JsonValue) =
if (jv.string != null) {
LocalDateTime.parse(jv.string, DateTimeFormatter.ofPattern("yyyy-MM-dd HH:mm"))
} else {
throw KlaxonException("Couldn't parse date: ${jv.string}")
}
override fun toJson(o: Any)
= """ { "date" : $o } """
}
Finally, declare the association between that converter and your annotation in your Klaxon
object before parsing:
val result = Klaxon()
.fieldConverter(KlaxonDate::class, dateConverter)
.parse<WithDate>("""
{
"theDate": "2017-05-10 16:30"
}
""")
assert(result?.date == LocalDateTime.of(2017, 5, 10, 16, 30))
You can instruct Klaxon to dynamically ignore properties with the PropertyStrategy
interface:
interface PropertyStrategy {
/**
* @return true if this property should be mapped.
*/
fun accept(property: KProperty<*>): Boolean
}
This is a dynamic version of @Json(ignored = true)
, which you can register with your Klaxon
instance with the function propertyStrategy()
:
val ps = object: PropertyStrategy {
override fun accept(property: KProperty<*>) = property.name != "something"
}
val klaxon = Klaxon().propertyStrategy(ps)
You can define multiple PropertyStrategy
instances, and in such a case, they all need to return true
for a property to be included.
JSON documents sometimes contain dynamic payloads whose type can vary. Klaxon supports two different use cases for
polymorphism: polymorphic classes and polymorphic fields. Klaxon gives you control on polymorphism with the
annotation @TypeFor
, which can be placed either on a class or on a field.
A polymorphic class is a class whose actual type is defined by one of its own properties. Consider this JSON:
[
{ "type": "rectangle", "width": 100, "height": 50 },
{ "type": "circle", "radius": 20}
]
The content of the field type
determines the class that needs to be instantiated. You would model this as
follows with Klaxon:
@TypeFor(field = "type", adapter = ShapeTypeAdapter::class)
open class Shape(val type: String)
data class Rectangle(val width: Int, val height: Int): Shape("rectangle")
data class Circle(val radius: Int): Shape("circle")
The type adapter is as follows:
class ShapeTypeAdapter: TypeAdapter<Shape> {
override fun classFor(type: Any): KClass<out Shape> = when(type as String) {
"rectangle" -> Rectangle::class
"circle" -> Circle::class
else -> throw IllegalArgumentException("Unknown type: $type")
}
}
Klaxon also allows a field to determine the class to be instantiated for another field. Consider the following JSON document:
[
{ "type": 1, "shape": { "width": 100, "height": 50 } },
{ "type": 2, "shape": { "radius": 20} }
]
This is an array of polymorphic objects. The type
field is a discriminant which determines the type of the field
shape
: if its value is 1
, the shape is a rectangle, if 2
, it's a circle.
To parse this document with Klaxon, we first model these classes with a hierarchy:
open class Shape
data class Rectangle(val width: Int, val height: Int): Shape()
data class Circle(val radius: Int): Shape()
We then define the class that the objects of this array are instances of:
class Data (
@TypeFor(field = "shape", adapter = ShapeTypeAdapter::class)
val type: Integer,
val shape: Shape
)
Notice the @TypeFor
annotation, which tells Klaxon which field this value is a discriminant for, and also provides
a class that will translate these integer values into the correct class:
class ShapeTypeAdapter: TypeAdapter<Shape> {
override fun classFor(type: Any): KClass<out Shape> = when(type as Int) {
1 -> Rectangle::class
2 -> Circle::class
else -> throw IllegalArgumentException("Unknown type: $type")
}
}
With this code in place, you can now parse the provided JSON document above the regular way and the the following tests will pass:
val shapes = Klaxon().parseArray<Data>(json)
assertThat(shapes!![0].shape as Rectangle).isEqualTo(Rectangle(100, 50))
assertThat(shapes[1].shape as Circle).isEqualTo(Circle(20))
The streaming API is useful in a few scenarios:
- When your JSON document is very large and reading it all in memory might cause issues.
- When you want your code to react as soon as JSON values are being read, without waiting for the entire document to be parsed.
This second point is especially important to make mobile apps as responsive as possible and make them less reliant on network speed.
Note: the streaming API requires that each value in the document be handled by the reader. If you are simply
looking to extract a single value the PathMatcher API
may be a better fit.
As opposed to conventional JSON libraries, Klaxon doesn't supply a JsonWriter
class to create JSON documents since
this need is already covered by the json()
function, documented in the Advanced DSL section.
Streaming JSON is performed with the JsonReader
class. Here is an example:
val objectString = """{
"name" : "Joe",
"age" : 23,
"flag" : true,
"array" : [1, 3],
"obj1" : { "a" : 1, "b" : 2 }
}"""
JsonReader(StringReader(objectString)).use { reader ->
reader.beginObject() {
var name: String? = null
var age: Int? = null
var flag: Boolean? = null
var array: List<Any> = arrayListOf<Any>()
var obj1: JsonObject? = null
while (reader.hasNext()) {
val readName = reader.nextName()
when (readName) {
"name" -> name = reader.nextString()
"age" -> age = reader.nextInt()
"flag" -> flag = reader.nextBoolean()
"array" -> array = reader.nextArray()
"obj1" -> obj1 = reader.nextObject()
else -> Assert.fail("Unexpected name: $readName")
}
}
}
}
There are two special functions to be aware of: beginObject()
and beginArray()
. Use these functions
when you are about to read an object or an array from your JSON stream. These functions will make sure
that the stream is correctly positioned (open brace or open bracket) and once you are done consuming
the content of that entity, the functions will make sure that your object is correctly closed (closing brace
or closing bracket). Note that these functions accept a closure as an argument, so there are no closeObject()/closeArray()
functions.
It is possible to mix both the object binding and streaming API's, so you can benefit from the best of both worlds.
For example, suppose your JSON document contains an array with thousands of elements in them, each of these elements being an object in your code base. You can use the streaming API to consume the array one element at a time and then use the object binding API to easily map these elements directly to one of your objects:
data class Person(val name: String, val age: Int)
val array = """[
{ "name": "Joe", "age": 23 },
{ "name": "Jill", "age": 35 }
]"""
fun streamingArray() {
val klaxon = Klaxon()
JsonReader(StringReader(array)).use { reader ->
val result = arrayListOf<Person>()
reader.beginArray {
while (reader.hasNext()) {
val person = klaxon.parse<Person1>(reader)
result.add(person)
}
}
}
}
The JSON Path specification defines how to locate elements inside a JSON document. Klaxon allows you to define path matchers that can match specific elements in your document and receive a callback each time a matching element is found.
Consider the following docoument:
{
"library": {
"books": [
{
"author": "Herman Melville",
"title": "Moby Dick"
},
{
"author": "Jules Vernes",
"title": "L'île mystérieuse"
}
]
}
}
According to the JSON Path spec, the two authors have the following JSON paths:
$.library.books[0].author
$.library.books[1].author
We'll define a PathMatcher
that uses a regular expression to filter only the elements we want:
val pathMatcher = object : PathMatcher {
override fun pathMatches(path: String) = Pattern.matches(".*library.*books.*author.*", path)
override fun onMatch(path: String, value: Any) {
println("Adding $path = $value")
}
}
Klaxon()
.pathMatcher(pathMatcher)
.parseJsonObject(document)
Output:
Adding $.library.books[0].author = Herman Melville
Adding $.library.books[1].author = Jules Vernes
Two notes:
- Klaxon doesn't support the JSON Path expression language, only the element location specification.
- This API is streaming: your path observers will be notified as soon as a matching element has been detected and its value completely parsed.
Values parsed from a valid JSON file can be of the following type:
- Int
- Long
- BigInteger
- String
- Double
- Boolean
- JsonObject
- JsonArray
JsonObject
behaves like a Map
while JsonArray
behaves like a List
. Once you have parsed a file, you should cast it to the type that you expect. For example, consider this simple file called object.json
:
{
"firstName" : "Cedric",
"lastName" : "Beust"
}
Since this is a JSON object, we parse it as follows:
fun parse(name: String) : Any? {
val cls = Parser::class.java
return cls.getResourceAsStream(name)?.let { inputStream ->
return Parser.default().parse(inputStream)
}
}
// ...
val obj = parse("/object.json") as JsonObject
Parse from String value :
val parser: Parser = Parser.default()
val stringBuilder: StringBuilder = StringBuilder("{\"name\":\"Cedric Beust\", \"age\":23}")
val json: JsonObject = parser.parse(stringBuilder) as JsonObject
println("Name : ${json.string("name")}, Age : ${json.int("age")}")
Result :
Name : Cedric Beust, Age : 23
You can also access the JSON content as a file, or any other resource you can get an InputStream
from.
Let's query these values:
val firstName = obj.string("firstName")
val lastName = obj.string("lastName")
println("Name: $firstName $lastName")
// Prints: Name: Cedric Beust
JsonObject
implements the following methods:
fun int(fieldName: String) : Int?
fun long(fieldName: String) : Long?
fun bigInt(fieldName: String) : BigInteger?
fun string(fieldName: String) : String?
fun double(fieldName: String) : Double?
fun boolean(fieldName: String) : Boolean?
fun obj(fieldName: String) : JsonObject?
fun <T> array(thisType: T, fieldName: String) : JsonArray<T>?
JsonArray
implements the same methods, except that they return JsonArray
s of the same type. This allows you to easily fetch collections of fields or even sub-objects. For example, consider the following:
[
{
"name" : "John",
"age" : 20
},
{
"name" : "Amy",
"age" : 25
},
{
"name" : "Jessica",
"age" : 38
}
]
We can easily collect all the ages as follows:
val array = parse("/e.json") as JsonArray<JsonObject>
val ages = array.long("age")
println("Ages: $ages")
// Prints: Ages: JsonArray(value=[20, 25, 38])
Since a JsonArray
behaves like a List
, we can apply closures on them, such as filter
:
val oldPeople = array.filter {
it.long("age")!! > 30
}
println("Old people: $oldPeople")
// Prints: Old people: [JsonObject(map={age=38, name=Jessica})]
Let's look at a more complex example:
[
{
"first": "Dale",
"last": "Cooper",
"schoolResults" : {
"scores": [
{ "name": "math", "grade" : 90 },
{ "name": "physics", "grade" : 50 },
{ "name": "history", "grade" : 85 }
],
"location" : "Berkeley"
}
},
{
"first": "Kara",
"last": "Thrace",
"schoolResults" : {
"scores": [
{ "name": "math", "grade" : 75 },
{ "name": "physics", "grade" : 90 },
{ "name": "history", "grade" : 55 }
],
"location" : "Stanford"
}
},
{
"first": "Jack",
"last": "Harkness",
"schoolResults" : {
"scores": [
{ "name": "math", "grade" : 40 },
{ "name": "physics", "grade" : 82 },
{ "name": "history", "grade" : 60 }
],
"location" : "Berkeley"
}
}
]
Let's chain a few operations, for example, finding the last names of all the people who studied in Berkeley:
println("=== Everyone who studied in Berkeley:")
val berkeley = array.filter {
it.obj("schoolResults")?.string("location") == "Berkeley"
}.map {
it.string("last")
}
println("$berkeley")
// Prints:
// === Everyone who studied in Berkeley:
// [Cooper, Harkness]
All the grades over 75:
println("=== All grades bigger than 75")
val result = array.flatMap {
it.obj("schoolResults")
?.array<JsonObject>("scores")?.filter {
it.long("grade")!! > 75
}!!
}
println("Result: $result")
// Prints:
// === All grades bigger than 75
// Result: [JsonObject(map={name=math, grade=90}), JsonObject(map={name=history, grade=85}), JsonObject(map={name=physics, grade=90}), JsonObject(map={name=physics, grade=82})]
Note the use of flatMap
which transforms an initial result of a list of lists into a single list. If you use map
, you will get a list of three lists:
// Using map instead of flatMap
// Prints:
// Result: [[JsonObject(map={name=math, grade=90}), JsonObject(map={name=history, grade=85})], [JsonObject(map={name=physics, grade=90})], [JsonObject( map={name=physics, grade=82})]]
You can convert any JsonObject
to a valid JSON string by calling toJsonString()
on it. If you want to get a pretty-printed
version of that string, call toJsonString(true)
You can easily create JSON objects with Klaxon's DSL. There are two different variants of that DSL: declarative and imperative.
The declarative DSL uses maps and pairs (with the to
operator) to declare the associations between your keys and your values:
val obj = json {
"color" to "red",
"age" to 23
}
The declarative syntax limits you to only having values in your object, so if you need to use arbitrary pieces of code inside your DSL object, you can use the imperative syntax instead. This syntax doesn't use pairs but lambdas, and you use the function put()
to define your fields:
val obj = json {
repeat(3) {
put("field$it", it * 2)
}
}
Output:
{
"fields": {
"field0": 0,
"field1": 2,
"field2": 4
}
}
If we have the following JSON
{
"users" : [
{
"email" : "user@is.here"
},
{
"email" : "spammer@there.us"
}
]
}
We can find all emails by
(parse("my.json") as JsonObject).lookup<String?>("users.email")
There are two parser implementations with official support. The first is written in Kotlin and is accessed by calling
Parser.default()
.
The second is a parser implemented using the FasterXML Jackson mapper. This parser has been found to take 1/2 the time of the default Parser on large JSON payloads.
The Jackson mapper can be found at the coordinates
com.beust:klaxon-jackson:[version]
. To use this parser, call the extension Parser.jackson()
.
The Kotlin based Parser is implemented as a mutable state machine supported by a simplistic State
monad,
making the main loop very simple:
val stateMachine = StateMachine()
val lexer = Lexer(inputStream)
var world = World(Status.INIT)
do {
val token = lexer.nextToken()
world = stateMachine.next(world, token)
} while (token.tokenType != Type.EOF)
Here are a few common errors and how to resolve them.
NoSuchMethodException: <init>
You might see the following exception:
Caused by: java.lang.NoSuchMethodException: BindingAdapterTest$personMappingTest$Person.<init>()
at java.lang.Class.getConstructor0(Class.java:3082)
at java.lang.Class.newInstance(Class.java:412)
This is typically caused by your object class being defined inside a function (which makes its constructor require an additional parameter that Klaxon doesn't know how to fill).
Solution: move that class definition outside of the function.