A quick tour of JSON libraries in Scala

Update (18.11.2015): added spray-json-shapeless library
Update (06.11.15): added circe library

Some time ago I wrote a post on relational database access in Scala since I was looking for a library and there were many of them available, making it hard to make a choice. It turns out that the situation is similar if not worse when it comes to JSON libraries in Scala.

There are just plenty of them. You have no idea. (me neither until I wrote this post)

The following is an attempt to provide a quick overview at how a subset of the libraries I found does a few of the most common things one would likely need to do in regard to JSON:

  • parsing it from a raw string
  • browsing the AST
  • building an AST
  • mapping to a case class

There are of course plenty more valid use-cases but this post is not going to cover those.

Let’s get started!

Scala JSON libraries

play-json

The Play Framework comes with a JSON library that covers most of the common use-cases one would encounter when building web applications:

Parsing raw JSON

 
 
1
2
3
4
5
6
7
8
9
scala> import play.api.libs.json._
import play.api.libs.json._
 
scala> val rawJson = """{"hello": "world", "age": 42}"""
rawJson: String = {"hello": "world", "age": 42}
 
scala> Json.parse(rawJson)
res0: play.api.libs.json.JsValue = {"hello":"world","age":42}
 

Browsing the AST

 
 
1
2
3
scala> (res0 \ "hello").as[String]
res1: String = world
 

Building a JSON AST

 
 
1
2
3
scala> Json.obj("hello" -> "world", "age" -> 42)
res2: play.api.libs.json.JsObject = {"hello":"world","age":42}
 

Mapping to a case class

 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
scala> case class Model(hello: String, age: Int)
defined class Model
 
scala> implicit val modelFormat = Json.format[Model]
modelFormat: play.api.libs.json.OFormat[Model] = play.api.libs.json.OFormat$$anon$1@81116d
 
scala> Json.fromJson[Model](res0)
res3: play.api.libs.json.JsResult[Model] = JsSuccess(Model(world,42),)
 
scala> res3.get
res4: Model = Model(world,42)
 
scala> Json.toJson(Model("bar", 23))
 
res5: play.api.libs.json.JsValue = {"hello":"bar","age":23}
 

lift-json

lift-json is the JSON library of the Lift framework. It is one of the oldest one out there if I’m not mistaken.

Parsing raw JSON

 
 
1
2
3
4
5
6
7
8
9
scala> import net.liftweb.json._
import net.liftweb.json._
 
scala> val rawJson = """{"hello": "world", "age": 42, "nested": { "deeper": { "treasure": true }}}"""
rawJson: String = {"hello": "world", "age": 42, "nested": { "deeper": { "treasure": true }}}
 
scala> parse(rawJson)
res0: net.liftweb.json.JValue = JObject(List(JField(hello,JString(world)), JField(age,JInt(42)), JField(nested,JObject(List(JField(deeper,JObject(List(JField(treasure,JBool(true))))))))))
 

Browsing the AST

 
 
1
2
3
scala> res0 \ "nested" \ "deeper" \ "treasure"
res1: net.liftweb.json.JsonAST.JValue = JBool(true)
 

Building a JSON AST

 
 
1
2
3
4
5
6
scala> import net.liftweb.json.JsonDSL._
import net.liftweb.json.JsonDSL._
 
scala> ("hello" -> "world") ~ ("age" -> 42)
res2: net.liftweb.json.JsonAST.JObject = JObject(List(JField(hello,JString(world)), JField(age,JInt(42))))
 

Mapping to a case class

 
 
1
2
3
4
5
6
7
8
9
10
11
12
object LiftJsonExample {
  def main(args: Array[String]): Unit = {
    import net.liftweb.json._
    implicit val formats = DefaultFormats
 
    case class Model(hello: String, age: Int)
    val rawJson = """{"hello": "world", "age": 42}"""
 
    println(parse(rawJson).extract[Model])
  }
}
 

spray-json

spray rols its own JSON library that focuses on working with case classes:

Parsing raw JSON

 
 
1
2
3
4
5
6
7
8
9
10
11
12
scala> import spray.json._
import spray.json._
 
scala> import DefaultJsonProtocol._
import DefaultJsonProtocol._
 
scala> val rawJson = """{"hello": "world", "age": 42}"""
rawJson: String = {"hello": "world", "age": 42}
 
scala> rawJson.parseJson
res0: spray.json.JsValue = {"hello":"world","age":42}
 

Browsing the AST

No can do?

Building a JSON AST

No can do?

Mapping to a case class

 
 
1
2
3
4
5
6
7
8
9
scala> case class Model(hello: String, age: Int)
defined class Model
 
scala> implicit val modelFormat = jsonFormat2(Model)
modelFormat: spray.json.RootJsonFormat[Model] = spray.json.ProductFormatsInstances$$anon$2@7bc880f8
 
scala> res1.convertTo[Model]
res4: Model = Model(world,42)
 

spray-json-shapeless

spray-json-shapeless derives spray-json’s JsonFormat-s automatically using shapeless (no need to define an implicit JsonFormat

Mapping to a case class

 
 
1
2
3
4
5
6
7
8
9
10
11
12
scala> import spray.json._
import spray.json._
 
scala> import fommil.sjs.FamilyFormats._
import fommil.sjs.FamilyFormats._
 
scala> case class Model(hello: String, age: Int)
defined class Model
 
scala> Model("hello", 42).toJson
res0: spray.json.JsValue = {"hello":"hello","age":42}
 

This is quite useful, it removes the boilerplate formats hanging around

json4s

json4s is a bit like slf4j in the sense that it tries to unite all kind of rogue libraries serving the same purpose by providing a common interface. But not every library uses it, which means that chances are high that your project will contain json4s in addition to another (few) Scala JSON libraries. Hopefully it will one day succeed with its slogan “One AST to rule them all”.

json4s has its roots in lift-json so this will look familiar:

Parsing raw JSON

 
 
1
2
3
4
5
6
7
8
9
10
11
12
scala> import org.json4s._
import org.json4s._
 
scala> import org.json4s.native.JsonMethods._
import org.json4s.native.JsonMethods._
 
scala> val rawJson = """{"hello": "world", "age": 42}"""
rawJson: String = {"hello": "world", "age": 42}
 
scala> parse(rawJson)
res0: org.json4s.JValue = JObject(List((hello,JString(world)), (age,JInt(42))))
 

Browsing the AST

 
 
1
2
3
scala> res0 \ "hello"
res1: org.json4s.JValue = JString(world)
 

Building a JSON AST tree

 
 
1
2
3
scala> ("hello" -> "world") ~ ("age" -> 42)
res2: org.json4s.JsonAST.JObject = JObject(List((hello,JString(world)), (age,JInt(42))))
 

Mapping to a case class

 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
object Json4sExample {
  def main(args: Array[String]): Unit = {
    import org.json4s._
    import org.json4s.native.JsonMethods._
    implicit val formats = DefaultFormats
 
    case class Model(hello: String, age: Int)
    val rawJson = """{"hello": "world", "age": 42}"""
 
    println(parse(rawJson).extract[Model])
  }
}
 

argonaut

Argonaut promotes “purely functional JSON in Scala”. It uses scalaz under the hood and

Parsing raw JSON

 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
import scalaz._, Scalaz._
import scalaz._
import Scalaz._
 
scala> import argonaut._, Argonaut._
import argonaut._
import Argonaut._
 
scala> val rawJson = """{"hello": "world", "age": 42, "nested": { "deeper": { "treasure": true }}}"""
rawJson: String = {"hello": "world", "age": 42, "nested": { "deeper": { "treasure": true }}}
 
scala> rawJson.parseOption
res0: Option[argonaut.Json] = Some({"hello":"world","age":42,"nested":{"deeper":{"treasure":true}}})
 

Browsing the AST

There are several mechanisms available, let’s use a lense. Those are funky:

 
 
1
2
3
4
5
6
scala> val ohMyLens = jObjectPL >=> jsonObjectPL("nested") >=> jObjectPL >=> jsonObjectPL("deeper") >=> jObjectPL >=> jsonObjectPL("treasure") >=> jBoolPL
ohMyLens: scalaz.PLensFamily[argonaut.Json,argonaut.Json,Boolean,Boolean] = scalaz.PLensFamilyFunctions$$anon$2@8c894ab
 
scala> ohMyLens.get(res0.get)
res1: Option[Boolean] = Some(true)
 

Building a JSON AST tree

 
 
1
2
3
scala> ("hello", jString("world")) ->: ("age", jNumber(42)) ->: jEmptyObject
res2: argonaut.Json = {"age":42,"hello":"world"}
 

Mapping to a case class

 
 
1
2
3
4
5
6
7
8
9
scala> case class Model(hello: String, age: Int)
defined class Model
 
scala> implicit def ModelCodecJson: CodecJson[Model] = casecodec2(Model.apply, Model.unapply)("hello", "age")
ModelCodecJson: argonaut.CodecJson[Model]
 
scala> rawJson.decodeOption[Model]
res3: Option[Model] = Some(Model(world,42))
 

circe

Circe is a fork of Argonaut that uses cats instead of scalaz and uses shapeless to generate codecs.

Parsing raw JSON

 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
scala> import io.circe._, io.circe.generic.auto._, io.circe.parse._, io.circe.syntax._      
import io.circe._        
import io.circe.generic.auto._      
import io.circe.parse._      
import io.circe.syntax._        
scala> val rawJson = """{"hello": "world", "age": 42, "nested": { "deeper": { "treasure": true }}}"""        
rawJson: String = {"hello": "world", "age": 42, "nested": { "deeper": { "treasure": true }}}        
scala> parse(rawJson).getOrElse(Json.empty)      
res0: io.circe.Json =        
{        
"hello" : "world",      
"age" : 42,        
"nested" : {        
"deeper" : {        
"treasure" : true      
}      
}      
}
 

Browsing the AST

So far there’s only support for cursors which let you move around the JSON tree and do changes if you would like, not for Lenses yet:

 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
scala> val cursor = res0.cursor      
cursor: io.circe.Cursor =        
CJson({      
"hello" : "world",      
"age" : 42,        
"nested" : {        
"deeper" : {        
"treasure" : true      
}      
}      
})      
scala> for {        
| nested <- cursor.downField("nested")      
| deeper <- nested.downField("deeper")      
| treasure <- deeper.downField("treasure")      
| } yield treasure.as[Boolean]      
res1: Option[io.circe.Decoder.Result[Boolean]] = Some(Right(true))
 

Mapping to a case class hierarchy

I didn’t get this to work with the example below. The example in the documentation works though so here is that one:

 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
scala> sealed trait Foo      
defined trait Foo        
scala> case class Bar(xs: List[String]) extends Foo      
defined class Bar        
scala> case class Qux(i: Int, d: Option[Double]) extends Foo        
defined class Qux        
scala> val foo: Foo = Qux(13, Some(14.0))        
foo: Foo = Qux(13,Some(14.0))        
scala> foo.asJson.noSpaces      
res0: String = {"Qux":{"d":14.0,"i":13}}        
scala> decode[Foo](foo.asJson.spaces4)      
res1: cats.data.Xor[io.circe.Error,Foo] = Right(Qux(13,Some(14.0)))
 

sphere-json

The sphere-json library focuses on providing an easy way to get de/serializers for entire families of case classes. This is really useful when working with any kind of protocol-related system. I am using it in a CQRS system where commands and events are travelling back and forth between the server and a Javascript UI. Instead of having to define a codec for each one of my case classes, I simply have them extend a common abstract class and let the library take care of the rest. Watch for yourselves:

Mapping to a case class hierarchy

 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
import io.sphere.json.generic._
import io.sphere.json._
 
object Messages {
 
  sealed abstract class Message
  case class Hello(hello: String) extends Message
  case class Age(age: Int) extends Message
 
}
object SphereJsonExample {
  import Messages._
 
  val rawHello = """{ "hello": "world", "type": "Hello" }"""
  val rawAge = """{ "age": 42, "type": "Age" }"""
 
  implicit val allYourJson = deriveJSON[Message]
 
  def magic(json: String) = fromJSON[Message](json).fold(
    fail => println("Oh noes: " + fail),
    hi => println(hi)
  )
 
  def main(args: Array[String]): Unit = {
    magic(rawHello)
    magic(rawAge)
  }
 
}
 

jawn

jawn is a library that focuses on speed. It defines a lightweight AST and is compatible with all kind of other ASTs that we have seen here.

Parsing raw JSON

 
 
1
2
3
4
5
6
7
8
9
10
11
12
scala> import jawn._
import jawn._
 
scala> import jawn.ast._
import jawn.ast._
 
scala> val rawJson = """{"hello": "world", "age": 42}"""
rawJson: String = {"hello": "world", "age": 42}
 
scala> jawn.ast.JParser.parseFromString(rawJson).get
res0: jawn.ast.JValue = {"age":42,"hello":"world"}
 

rapture

rapture’s json library is the ultimate Scala JSON library. It doesn’t really do anything with JSON itself, instead, it abstracts over the following JSON libraries (which it callsbackends):

  • Argonaut
  • Jackson
  • Jawn
  • JSON4S
  • Lift
  • Play
  • Scala standard library JSON
  • Spray

Parsing raw JSON

 
 
1
2
3
4
5
6
7
8
scala> import rapture.json._
 
scala> import rapture.json.jsonBackends.play._
import rapture.json.jsonBackends.play._
 
scala> Json.parse(rawJson)
res2: rapture.json.Json = {"hello":"world","age":42}
 

Browsing the AST

Rapture is using Scala’s Dynamic trait, which makes this fun:

 
 
1
2
3
scala> res2.hello.as[String]
res3: String = world
 

Building a JSON AST tree

 
 
1
2
3
scala> json"""{ "hello": "world", "age": 42}"""
res6: rapture.json.Json = {"hello":"world","age":42}
 

The Scala standard library

Up until recently I did not know that Scala had a JSON utility in its standard library. But here it is!

Parsing raw JSON

 
 
1
2
3
4
5
6
7
8
9
10
scala> import scala.util.parsing.json._
import scala.util.parsing.json._
 
scala> val rawJson = """{"hello": "world", "age": 42}"""
rawJson: String = {"hello": "world", "age": 42}
 
scala> JSON.parseFull(rawJson)
warning: there was one deprecation warning; re-run with -deprecation for details
res0: Option[Any] = Some(Map(hello -> world, age -> 42.0))
 

Browsing the “AST”

 
 
1
2
3
scala> res0.get.asInstanceOf[Map[String, Any]]("hello").asInstanceOf[String]
res4: String = world
 

Even more!

I am forgetting a ton of libraries here I am sure. That, and I am tired and my glass ofUigeadail is getting empty.

So let me mention a few more:

Great, now which one to pick?

Honestly I don’t have any good advice here. These days I am sticking to play-json and sphere-json, but that’s for no particular reason other than these libraries being there and doing what I need to do (parse and write JSON, traverse and some rare times transform the AST, bind to case classes, make it possible to support custom Java types). If play-json had support for hierarchies out of the box I would probably not have even looked for anything else.

Because for all the joy there seems to be in implementing JSON libraries in Scala, one thing has to be said: JSON de/serialization is boring. It’s this annoying thing that you have to do in order to get your application to talk to another computerized system, period.

I have never met a developer who told me how much pleasure they derived from turning classes into JSON strings and back. That’s just not a thing.

I have, however, met more than one developer that has run into trouble getting library X to cover one of the simple use-cases outlined above. Believe me, there is nothing more frustrating than having to spend time on the annoying task of setting up JSON de/serialization in order to do the boring thing of tossing strings back and forth the network. That’s time you will never get back.

Good night, and good luck.

原文

A quick tour of JSON libraries in Scala的更多相关文章

  1. Avro schemas are defined with JSON . This facilitates implementation in languages that already have JSON libraries.

    https://avro.apache.org/docs/current/ Introduction Apache Avro™ is a data serialization system. Avro ...

  2. 开发工具-scala处理json格式利器-json4s

    1.为什么是json4s 从json4s的官方描述 At this moment there are at least 6 json libraries for scala, not counting ...

  3. scala读取解析json文件

    import scala.util.parsing.json.JSON._ import scala.io.Source object ScalaJsonParse { def main(args: ...

  4. scala解析json —— json4s 解析json方法汇总

    使用json4s的框架,包括spark,flink 1.org.json4s 引入pom的方法 对于本地支持,引入以下依赖项添加到pom中 <dependency> <groupId ...

  5. Awesome Go精选的Go框架，库和软件的精选清单.A curated list of awesome Go frameworks, libraries and software

    Awesome Go      financial support to Awesome Go A curated list of awesome Go frameworks, libraries a ...

  6. scala vs java 相同点和差异

    本贴是我摘抄自国外网站,用作备忘,也作为分享! Similarities between Scala and Java Following are some of the major similari ...

  7. scala攻略--简介

    在个人学习scala的过程中,产生了写一系列随笔的想法,这些随笔包括:翻译自官网.其他英文网站的文章以及自己的心得体会,本文章作为这个系列中的第一个. 由于本人能力所限,以及对scala还处于初级阶段 ...

  8. spark2.1操作json（save/read）

    建筑物配置信息: case class BuildingConfig(buildingid: String, building_height: Long, gridcount: Long, gis_d ...

  9. 数据序列化导读(3)[JSON v.s. YAML]

    前面两节介绍了JSON和YAML,本文则对下面的文章做一个中英文对照翻译. Comparison between JSON and YAML for data serialization用于数据序列化 ...

随机推荐

  1. web2py学习之getting start环境搭建

    一般如果做一个工程,可能需要ide,需要好的工具,web2py自包含了一个基于web的开发工具,但是并不算很好的编辑器 第二个可以使用的是pycharm,利用pycharm可以创建web2py的web ...

  2. CSS里的pointer-events属性

    现代浏览器里CSS的职责范围和JavaScript的越来越模糊分不清.比如CSS里-webkit-touch-callout属性在iOS里能禁止当用户点击时弹出气泡框.而本文要说的pointer-ev ...

  3. Android学习系列(42)--Android Studio实战技巧

    使用android studio开发项目的一些问题,功能和技巧. 1. 环境 Mac OSX 10.9.5 + Android Studio 0.8.9 2. gradle项目加载超慢 这是因为gra ...

  4. Oracle常用命令

    查看数据库锁的情况 SELECT object_name, machine, s.sid, s.serial#  FROM gv$locked_object l, dba_objects o, gv$ ...

  5. POJ 3349 HASH

    题目链接:http://poj.org/problem?id=3349 题意:你可能听说话世界上没有两片相同的雪花,我们定义一个雪花有6个瓣,如果存在有2个雪花相同[雪花是环形的,所以相同可以是旋转过 ...

  6. SU sunmo命令学习

  7. android测试点汇总

    Android的功能测试点 安装\卸载 App具体功能点 联网(默认的联网方式是什么?Wifi orSim卡?网络切换是否有相应的提示说明?飞行模式) 程序进入输入功能时,是否正常弹出键盘;键盘是否遮 ...

  8. 我的c++学习(6)默认参数和内联函数

    默认参数 一般情况下,函数调用时实参个数应与形参相同,但为了更方便地使用函数,C++也允许定义具有默认参数的函数,这种函数调用时实参个数可以与形参不相同.“默认参数”指在定义或声明函数时为形参指定默认 ...

  9. jQuery跨域

    其实jQuery跨域很简单很简单,你记住格式就好,跨域的原理请参考 <jsonp跨域> jQuery跨域代码: $.ajax({ url:'https://suggest.taobao.c ...

  10. HDU4871 Shortest-path tree（最短路径树 + 树的点分治）

    题目大概要先求一张边有权的图的根为1的最短路径树,要满足根到各点路径序列的字典序最小:然后求这棵最短路径树包含k个结点的最长路径的长度和个数. 首先先构造出这棵字典序最小的最短路径树..好吧,我太傻逼 ...