mdedetrich / scalajson   1.0.0-M8

BSD 3-clause "New" or "Revised" License GitHub

ScalaJSON - JSON for Scala, currently contains minimal AST

Scala versions: 2.12 2.11 2.10
Scala.js versions: 0.6


Join the chat at Build Status Scala.js

ScalaJSON library, currently provides two minimal implementations of JSON AST, one that is designed for typical use and another that is designed for performance/corner cases.


Built for Scala 2.10.x, 2.11.x, 2.12.x, and 2.13.x milestones.

"org.scala-lang.platform" %% "scalajson" % "1.0.0-M4"

If you are using Scala.js, you need to do

"org.scala-lang.platform" %%% "scalajson" % "1.0.0-M4"

Standard AST

Implementation is in scalajson.ast.JValue


  • Fully immutable (all underlying collections/types used are immutable)
  • constant/effective constant lookup time for scalajson.ast.JArray/scalajson.ast.JObject
  • Adherence to the typical use for the JSON standard.
    • Number representation for scalajson.ast.JNumber is a String which checks if its a valid JSON representation of a number (
      • Equals will properly detect if two numbers are equal, i.e. scalajson.ast.JNumber("34.00") == scalajson.ast.JNumber("34")
      • Hashcode has been designed to provide consistent hash for numbers of unlimited precision.
      • If you construct a JNumber with Float.NaN/Float.PositiveInfinity/Float.NegativeInfinity/Double.NaN/Double.PositiveInfinity/Double.NegativeInfinity it will return a JNull
      • You can construct an unlimited precision number using a string, i.e. JNumber("34324"). Returns an Option[JNumber] (will return None if String isn't a valid number)
        • Note that this doesn't work for Scala 2.10 due to a restriction with how private constructor case classes are handled. For this reason a JNumber.fromString method is provided which compiles on all platforms and scala versions
    • scalajson.ast.JObject is an actual Map[String,JValue]. This means that it doesn't handle duplicate keys for a scalajson.ast.JObject, nor does it handle key ordering.
    • scalajson.ast.JArray is an Vector.
  • Library does not allow invalid JSON in the representation and hence we can guarantee that a scalajson.ast.JValue will always contain a valid structure that can be serialized/rendered into JSON.
    • Note that you can lose precision when using scalajson.ast.JNumber in Scala.js (see Scala.js section for more info).
    • The .copy method of scalajson.ast.JNumber has been overridden to make sure you can't replace the internal String with an incorrect number
      • Will throw a NumberFormatException if you use .copy with an invalid JSON number
  • Due to the above, has properly implemented deep equality for all types of scalajson.ast.JValue

Unsafe AST

Implementation is in scalajson.unsafe.JValue


  • Uses the best performing datastructure's for high performance/low memory usage in construction of a unsafe.JValue
    • scalajson.ast.unsafe.JArray stored as an Array for JVM and js.Array for Scala.js
    • scalajson.ast.unsafe.JObject stored as an Array for JVM and js.Array for Scala.js
    • scalajson.ast.unsafe.JNumber stored as a String
  • Doesn't use Scala's stdlib collection's library
  • Defer all runtime errors. We don't throw errors if you happen to provide Infinity/NaN or other invalid data.
  • Allow duplicate and ordered keys for a scalajson.ast.unsafe.JObject.
  • Allow any length/precision of numbers for scalajson.ast.unsafe.JNumber since its represented as a String
    • Equals/hashcode only checks for String equality, not number equality.
      • Means that scalajson.ast.unsafe.JNumber("34.00") is not equal to scalajson.ast.unsafe.JNumber("34")
  • This means that scalajson.ast.unsafe.JValue can represent everything that can can be considered valid under the official JSON spec, even if its not considered sane (i.e. duplicate keys for a scalajson.ast.unsafe.JObject).
    • Also means it can hold invalid data, due to not doing runtime checks
  • Is referentially transparent in regards to String -> scalajson.ast.unsafe.JValue -> String since scalajson.ast.unsafe.JObject preserves ordering/duplicate keys
  • Implements structural equality for both hashCode and equals. If you need reference equality you can use eq and if you need reference hashCode you can use .value.hashCode. Also note that for deep comparison is used both hashCode and equals.

Conversion between scalajson.JValue and scalajson.ast.unsafe.JValue

Any scalajson.ast.JValue implements a conversion to scalajson.ast.unsafe.JValue with a toUnsafe method and vice versa with a toStandard method. These conversion methods have been written to be as fast as possible.

There are some peculiarities when converting between the two AST's. When converting a scalajson.ast.unsafe.JNumber to a scalajson.ast.JNumber, it is possible for this to fail at runtime (since the internal representation of scalajson.ast.unsafe.JNumber is a String and it doesn't have a runtime check). It is up to the caller on how to handle this error (and when), a runtime check is deliberately avoided on our end for performance reasons.

Converting from a scalajson.ast.JObject to a scalajson.ast.unsafe.JObject will produce an scalajson.ast.unsafe.JObject with an undefined ordering for its internal Array/js.Array representation. This is because a Map has no predefined ordering. If you wish to provide ordering, you will either need to write your own custom conversion to handle this case. Duplicate keys will also be removed for the same reason in an undefined manner.

Do note that according to the JSON spec, whether to order keys for a JObject is not specified. Also note that Map disregards ordering for equality, however Array/js.Array equality takes ordering into account.

Number conversions

ScalaJSON JNumber provides conversions to various number types with the following conventions

  • toInt: Safe conversion to Int which accounts for values such as 1.0 and 100.00e-2 (which both evaluate to 1). Also safely detects over/underflow.
  • toLong: Safe conversion to Long which accounts for values such as 1.0 and 100.00e-2 (which both evaluate to 1). Also safely detects over/underflow.
  • toDouble: Converts to a Double assuming the same semantics of Double (i.e. precision loss is expected).
  • toFloat: Converts to a Float assuming the same semantics of Float (i.e. precision loss is expected).
  • toBigInt: Converts to a BigInt which accounts for values such as 1.0 and 100.00e-2 (which evaluates to 1). Can construct a BigIntfor as much as memory as the system has (if your system runs out of memory this is considered undefined behaviour).
  • toBigDecimal: Converts to a BigDecimal with all of the caveats of BigDecimal construction. The BigDecimal is constructed with MathContext.UNLIMITED precision.

With the .toFloat and .toDouble methods, if you don't want any loss in precision, its advisable to convert to BigDecimal first and then work from there, i.e. when working with Decimal/Float, its implied that you will have loss of precision.

Remember that in all cases if these methods are not applicable, you can always use the .value field to get the original string representation of the number.


ScalaJSON also provides support for Scala.js. The usage of Scala.js mirrors the usage of Scala on the JVM however Scala.js also implements a .toJsAny method which allows you to convert any scalajson.ast.JValue/scalajson.ast.unsafe.JValue to a Javascript value in Scala.js.

Note that, since a scalajson.ast.JNumber/scalajson.ast.unsafe.JNumber is unlimited precision (represented internally as a String), calls to .toJsAny can lose precision on the underlying number (numbers in Javascript are represented as double precision floating point number). You can use the .value method on a scalajson.ast.JNumber/scalajson.ast.unsafe.JNumber to get the raw string value as a solution to this problem.

Further, toFloat on JNumber (see Number Conversions ) can have different semantics on Scala.js, depending on whether you have strict-floats enabled in your application. Please see the Scala.js semantics page for more information.


scalajson.JNumber uses jNumberRegex to validate whether a number is a valid JSON number. One can use jNumberRegex explicitly if you want to use the validation that is used by scalajson.JNumber (for example, you may want to validate proper numbers before creating a scalajson.unsafe.JNumber).

import scalajson.jNumberRegex

"3535353" match {
  case jNumberRegex(_ *) => true
  case _ => false

Code of Conduct

ScalaJSON uses the Scala Code of Conduct for all communication and discussion. This includes both GitHub, Gitter chat and other more direct lines of communication such as email.

Code formatting

The project is formatted using scalafmt. Please run scalafmt in SBT before committing any changes