Jackson jr

General

Category
Free
Tag
JSON
License
Apache License, Version 2.0
Registered
Jul 5, 2014
Favorites
1
Link
https://github.com/FasterXML/jackson-jr
See also
kraph
GSON
ig-json-parser
ModelMapper
RapidJSON

Additional

Language
Java
Version
rc4b (May 22, 2014)
Created
Sep 7, 2013
Updated
Apr 12, 2018
Owner
FasterXML, LLC (FasterXML)
Contributors
brharrington
cowtowncoder
tatu-at-salesforce
prb
komamitsu
valery1707
lomigmegard
csessafitbit
8
Activity
Badge
Generate
Download
Source code

Advertising

Overview

Jackson jr is a compact alternative to full Jackson Databind component. It implements a subset of functionality, for example for cases where:

  1. Size of jar matters (jackson-jr size is about 95kB)
  2. Startup time matters (jackson-jr has very low initialization overhead)

In addition to basic datatypes (core JDK types like Lists, Maps, wrapper types), package supports reading and writing of standard Java Beans (implementation that mimics standard JDK Bean Introspection): that is, subset of POJOs that define setters/getters (starting with Jackson-jr 2.8) you can alternatively use public fields).

Jackson jr also adds composer implementation that can be used to construct JSON output with builder-style API, but without necessarily having to build an in-memory representation: instead, it can directly use streaming-api for direct output. It is also possible to build actual in-memory JSON String or byte[] representation, if that is preferable.

Jackson jr artifact itself is currently about 95 kB in size, and only depends on Jackson Streaming API package. Combined size, for "all" jar, is about 400 kB (of which streaming API is about 300 kB), for use cases where a single jar is preferred over more modular approach. Finally, use of jar minimizers like ProGuard can bring the jar size down even further, by renaming and removing debug information.

License

Good old Apache License.

Packaging

Project is composed of multiple Maven sub-modules, each corresponding to a jar:

  • jr-objects contains the "core" databinding implementation, and is commonly the only dependency to use
    • Depends on jackson-core for low-level reading/writing
  • jr-stree contains a simple TreeCodec implementation, with which it is possible to read JSON as TreeNodes (see more below)
  • jr-retrofit2 contains jackson-jr - based handlers for Retrofit 2 library
    • Depends on jackson-jr and Retrofit API jars, and indirectly on jackson-core
  • jr-all creates an "uber-jar" that contains individual modules along with all their dependencies:
    • jr-objects classes as-is, without relocating
    • jr-stree classes as-is, without relocating
    • Jackson streaming ( jackson-core) contents relocated ("shaded"), for private use by jackson-jr
    • Does NOT contain jr-retrofit2 component

If you are not sure which package to use, the answer is usually jr-objects, and build system (maven, gradle) will fetch the dependency needed. jr-all jar is only used if the single-jar deployment (self-contained, no external dependencies) is needed.

Status

Usage

Reading/writing Simple Objects, Beans, List/arrays thereof

Functionality of this package is contained in Java package com.fasterxml.jackson.jr.ob.

All functionality is accessed through main JSON Object; you can either used singleton JSON.std, or construct individual objects -- either way, JSON instances are ALWAYS immutable and hence thread-safe.

We can start by reading JSON

String INPUT = "{\"a\":[1,2,{\"b\":true},3],\"c\":3}";
Object ob = JSON.std.anyFrom(INPUT);
// or
Map<String,Object> map = JSON.std.mapFrom(INPUT);
// or
MyBean bean = JSON.std.beanFrom(MyBean.class, INPUT);

from any of the usual input sources ( InputStream, Reader, String or byte[] that contains JSON, URL, JsonParser); and can write same Objects as JSON:

String json = JSON.std.asString(map);
JSON.std.write(ob, new File("/tmp/stuff.json");
// and with indentation; but skip writing of null properties
byte[] bytes = JSON.std
    .with(Feature.PRETTY_PRINT_OUTPUT)
    .without(Feature.WRITE_NULL_PROPERTIES)
    .asBytes(bean);

and may also read Lists and arrays of simple and Bean types:

List<MyType> beans = JSON.std.listOfFrom(MyType.class, INPUT);

(writing of Lists and arrays works without addition effort: just pass List/array as-is)

Writing with composers

An alternative method exists for writing: "fluent" style output can be used as follows:

String json = JSON.std
  .with(JSON.Feature.PRETTY_PRINT_OUTPUT)
  .composeString()
  .startObject()
    .put("a", 1)
    .startArrayField("arr")
      .add(1).add(2).add(3)
    .end()
    .startObjectField("ob")
      .put("x", 3)
      .put("y", 4)
      .startArrayField("args").add("none").end()
    .end()
    .put("last", true)
  .end()
  .finish();

would produce (since pretty-printing is enabled)

{
  "a" : 1,
  "arr" : [1,2,3],
  "ob" : {
    "x" : 3,
    "y" : 4,
    "args" : ["none"]
  },
  "last" : true
}

Reading/writing JSON Trees

Jackson jr allows pluggable "tree models", and also provides one implementation, jr-stree. Usage for jr-stree is by configuring JSON with codec, and then using treeFrom and write methods like so:

JSON json = JSON.std.with(new JacksonJrsTreeCodec());
TreeNode root = json.treeFrom("{\"value\" : [1, 2, 3]}");
assertTrue(root.isObject());
TreeNode array = root.get("value");
assertTrue(array.isArray());
JrsNumber n = (JrsNumber) array.get(1);
assertEquals(2, n.getValue().intValue());

String json = json.asString(root);

Note that jr-stree implementation is a small minimalistic implementation with immutable nodes. It is most useful for simple reading use cases.

It is however possible to write your own TreeCodec implementations that integrate seamlessly, and in future other tree models may be offered as part of jackson-jr, or via other libraries.

Designing your Beans

To support readability and writability of your own types, your Java objects must either:

  • Implement Bean style accesors (getters for accessing data to write and/or setter for binding JSON data into objects), and define no-argument (default) constructor, OR
  • Define single-argument constructor if binding from JSON String (single-String argument) or JSON integer number (single- long or Long argument)

Note that although getters and setters need to be public (since JDK Bean Introspection does not find any other methods), constructors may have any access right, including private.

Also: starting with version 2.8, public fields may also be used (although their discovery may be disabled using JSON.Feature.USE_FIELDS) as an alternative: this is useful when limiting number of otherwise useless "getter" and "setter" methods.

Customizing behavior with Features

There are many customizable features you can use with JSON object; see Full List of Features for details. But usage itself is via fluent methods like so:

String json = JSON.std
  .with(JSON.Feature.PRETTY_PRINT_OUTPUT)
  .without(JSON.Feature.FAIL_ON_DUPLICATE_MAP_KEYS)
  .asString(...);

Get it!

You can use Maven dependency like:

<dependency>
  <groupId>com.fasterxml.jackson.jr</groupId>
  <artifactId>jackson-jr-objects</artifactId>
  <version>2.9.0</version>
</dependency>

and then you can also download jars via Central Maven repository.

Or you can also clone the project and build it locally with mvn clean install.

Alternatively if you want a single jar deployment, you can use jackson-jr-all jar which embeds jackson-core (repackaged using Shade plug-in, so as not to conflict with "vanilla" jackson-core):

http://repo1.maven.org/maven2/com/fasterxml/jackson/jr/jackson-jr-all/

Performance

Initial performance testing using JVM Serializers benchmark suggests that it is almost as fast as full Jackson databind -- additional overhead for tests is 5-10% for both serialization and deserialization. So performance is practically identical.

In fact, when only handling Lists and Maps style content, speed jackson-jr speed fully matches jackson-databind performance (Bean/POJO case is where full databinding's extensive optimizations help more). So performance should be adequate, and choice should be more based on functionality, convenience and deployment factors.

About the only thing missing is that there is no equivalent to Afterburner, which can further speed up databind by 20-30%, for most performance-sensitive systems.