Home

Awesome

JSON Schema Validator <a href="https://www.paypal.me/erosb88"><img style="float: right" width="70" height="35" src="./donate.jpg" title="Support this project by making a donation"></a>

Deprecation notice

This library is currently in maintenance mode and superseded by erosb/json-sKema.

This repository and won't see any new features. It provides solid support for draft-04, draft-06 and draft-07 versions of the JSON Schema specification.

The latest draft 2020-12 is supported only by erosb/json-sKema.

Apache 2.0 License Build Status Coverage Status

This project is an implementation of the JSON Schema Draft v4, Draft v6 and Draft v7 specifications. It uses the org.json API (created by Douglas Crockford) for representing JSON data.

When to use this library?

Lets assume that you already know what JSON Schema is, and you want to utilize it in a Java application to validate JSON data. But - as you may have already discovered - there is also an other Java implementation of the JSON Schema specification. So here are some advices about which one to use:

Maven installation

Add the following dependency to your pom.xml:

<dependency>
	<groupId>com.github.erosb</groupId>
	<artifactId>everit-json-schema</artifactId>
	<version>1.14.4</version>
</dependency>

Note about older versions: versions between 1.6.0 and 1.9.1 can only be found on JitPack with com.github.everit-org.json-schema:org.everit.json.schema coordinates. Versions 1.0.0 ... 1.5.1 are available on Maven Central under org.everit.json:org.everit.json.schema coordinates.

Java6/7 versions

There were a couple of attempts to make the library work on Java 6/7.

A java6 port of version 1.9.2 was developed by @mindbender1 and it is accessible through Maven Central with the following coordinates:

    <dependency>
        <groupId>com.github.erosb</groupId>
        <artifactId>everit-json-schema-jdk6</artifactId>
        <version>1.9.2</version>
    </dependency>

Backports of older versions:

Quickstart

import org.everit.json.schema.Schema;
import org.everit.json.schema.loader.SchemaLoader;
import org.json.JSONObject;
import org.json.JSONTokener;
// ...
try (InputStream inputStream = getClass().getResourceAsStream("/path/to/your/schema.json")) {
  JSONObject rawSchema = new JSONObject(new JSONTokener(inputStream));
  Schema schema = SchemaLoader.load(rawSchema);
  schema.validate(new JSONObject("{\"hello\" : \"world\"}")); // throws a ValidationException if this object is invalid
}

Draft 4, Draft 6 or Draft 7?

JSON Schema has currently 4 major releases, Draft 3, Draft 4, Draft 6 and Draft 7. This library implements the 3 newer ones, you can have a quick look at the differences here and here. Since the two versions have a number of differences - and draft 6 is not backwards-compatible with draft 4 - it is good to know which version will you use.

The best way to denote the JSON Schema version you want to use is to include its meta-schema URL in the document root with the "$schema" key. This is a common notation, facilitated by the library to determine which version should be used.

Quick reference:

If you want to specify the meta-schema version explicitly then you can change the default from Draft 4 to Draft 6 / 7 by configuring the loader this way:

SchemaLoader loader = SchemaLoader.builder()
                .schemaJson(yourSchemaJSON)
                .draftV6Support() // or draftV7Support()
                .build();
Schema schema = loader.load().build();

Investigating failures

Starting from version 1.1.0 the validator collects every schema violations (instead of failing immediately on the first one). Each failure is denoted by a JSON pointer, pointing from the root of the document to the violating part. If more than one schema violations have been detected, then a ValidationException will be thrown at the most common parent elements of the violations, and each separate violations can be obtained using the ValidationException#getCausingExceptions() method.

To demonstrate the above concepts, lets see an example. Lets consider the following schema:

{
	"type" : "object",
	"properties" : {
		"rectangle" : {"$ref" : "#/definitions/Rectangle" }
	},
	"definitions" : {
		"size" : {
			"type" : "number",
			"minimum" : 0
		},
		"Rectangle" : {
			"type" : "object",
			"properties" : {
				"a" : {"$ref" : "#/definitions/size"},
				"b" : {"$ref" : "#/definitions/size"}
			}
		}
	}
}

The following JSON document has only one violation against the schema (since "a" cannot be negative):

{
	"rectangle" : {
		"a" : -5,
		"b" : 5
	}
}

In this case the thrown ValidationException will point to #/rectangle/a and it won't contain sub-exceptions:

try {
  schema.validate(rectangleSingleFailure);
} catch (ValidationException e) {
  // prints #/rectangle/a: -5.0 is not higher or equal to 0
  System.out.println(e.getMessage());
}

Now - to illustrate the way that multiple violations are handled - let's consider the following JSON document, where both the "a" and "b" properties violate the above schema:

{
	"rectangle" : {
		"a" : -5,
		"b" : "asd"
	}
}

In this case the thrown ValidationException will point to #/rectangle, and it has 2 sub-exceptions, pointing to #/rectangle/a and #/rectangle/b :

try {
  schema.validate(rectangleMultipleFailures);
} catch (ValidationException e) {
  System.out.println(e.getMessage());
  e.getCausingExceptions().stream()
      .map(ValidationException::getMessage)
      .forEach(System.out::println);
}

This will print the following output:

#/rectangle: 2 schema violations found
#/rectangle/a: -5.0 is not higher or equal to 0
#/rectangle/b: expected type: Number, found: String

JSON report of the failures

Since version 1.4.0 it is possible to print the ValidationException instances as JSON-formatted failure reports. The ValidationException#toJSON() method returns a JSONObject instance with the following keys:

Please take into account that the complete failure report is a hierarchical tree structure: sub-causes of a cause can be obtained using #getCausingExceptions() .

ValidationListeners - Tracking the validation process

ValidationListeners can serve the purpose of resolving ambiguity about how does an instance JSON match (or does not match) against a schema. You can attach a ValidationListener implementation to the validator to receive event notifications about intermediate success/failure results.

Example:

import org.everit.json.schema.Validator;
...
Validator validator = Validator.builder()
	.withListener(new YourValidationListenerImplementation())
	.build();
validator.performValidation(schema, input);

The currently supported events:

See the javadoc of the org.everit.json.schema.event.ValidationListener interface for more details. The particular event classes also have proper #toJSON() and #toString() implementations so you can print them in an easily parse-able format.

Early failure mode

By default the validation error reporting in collecting mode (see the "Investigating failures" chapter). That is convenient for having a detailed error report, but under some circumstances it is more appropriate to stop the validation when a failure is found without checking the rest of the JSON document. To toggle this fast-failing validation mode

Example:

import org.everit.json.schema.Validator;
...
Validator validator = Validator.builder()
	.failEarly()
	.build();
validator.performValidation(schema, input);

Note: the Validator class is immutable and thread-safe, so you don't have to create a new one for each validation, it is enough to configure it only once.

Lenient mode

In some cases, when validating numbers or booleans, it makes sense to accept string values that are parseable as such primitives, because any successive processing will also automatically parse these literals into proper numeric and logical values. Also, non-string primitive values are trivial to convert to strings, so why not to permit any json primitives as strings?

For example, let's take this schema:

{
    "properties": {
        "booleanProp": {
            "type": "boolean"
        },
        "integerProp": {
            "type": "integer"
        },
        "nullProp": {
            "type": "null"
        },
        "numberProp": {
            "type": "number"
        },
        "stringProp": {
          "type": "string"
        }
    }
}

The following JSON document fails to validate, although all of the strings could easily be converted into appropriate values:

{
  "numberProp": "12.34",
  "integerProp": "12",
  "booleanProp": "true",
  "nullProp": "null",
  "stringProp": 12.34
}

In this case, if you want the above instance to pass the validation against the schema, you need to use the lenient primitive validation configuration turned on. Example:

import org.everit.json.schema.*;
...
Validator validator = Validator.builder()
	.primitiveValidationStrategry(PrimitiveValidationStrategy.LENIENT)
	.build();
validator.performValidation(schema, input);

Note: in lenient parsing mode, all 22 possible boolean literals will be accepted as logical values.

Default values

The JSON Schema specification defines the "default" keyword for denoting default values, though it doesn't explicitly state how it should affect the validation process. By default this library doesn't set the default values, but if you need this feature, you can turn it on by the SchemaLoaderBuilder#useDefaults(boolean) method, before loading the schema:

{
  "properties": {
    "prop": {
      "type": "number",
      "default": 1
    }
  }
}

JSONObject input = new JSONObject("{}");
System.out.println(input.get("prop")); // prints null
Schema schema = SchemaLoader.builder()
	.useDefaults(true)
	.schemaJson(rawSchema)
	.build()
	.load().build();
schema.validate(input);
System.out.println(input.get("prop")); // prints 1

If there are some properties missing from input which have "default" values in the schema, then they will be set by the validator during validation.

RegExp Implementations

For supporting the "regex" keyword of JSON Schema the library offers two possible implementations:

While the RE2J library provides significantly better performance than java.util.regex, it is not completely compatible with the syntax supported by java.util or ECMA 262. So RE2J is recommended if you are concerned about performance and its limitations are acceptable.

The RE2J implementation can be activated with the SchemaLoaderBuilder#regexpFactory() call:

SchemaLoader loader = SchemaLoader.builder()
    .regexpFactory(new RE2JRegexpFactory())
    // ...
    .build();

Notes:

readOnly and writeOnly context

The library supports the readOnly and writeOnly keywords which first appeared in Draft 7. If you want to utilize this feature, then before validation you need to tell the validator if the validation happens in read or write context. Example:

schema.json:

{
   "properties": {
     "id": {
       "type": "number",
       "readOnly": true
     }
   }  
}

Validation code snippet:


Validator validator = Validator.builder()
                .readWriteContext(ReadWriteContext.WRITE)
                .build();

validator.performValidation(schema, new JSONObject("{\"id\":42}"));

In this case we told the validator that the validation happens in WRITE context, and in the input JSON object the "id" property appears, which is marked as "readOnly" in the schema, therefore this call will throw a ValidationException.

Format validators

Starting from version 1.2.0 the library supports the "format" keyword (which is an optional part of the specification).

The supported formats vary depending on the schema spec version you use (since the standard formats were introduced in different versions on the validation specification).

Here is a compatibility table of the supported standard formats:

Draft 4Draft 6Draft 7
date-time:white_check_mark::white_check_mark::white_check_mark:
email:white_check_mark::white_check_mark::white_check_mark:
hostname:white_check_mark::white_check_mark::white_check_mark:
ipv4:white_check_mark::white_check_mark::white_check_mark:
ipv6:white_check_mark::white_check_mark::white_check_mark:
uri:white_check_mark::white_check_mark::white_check_mark:
uri-reference:white_check_mark::white_check_mark:
uri-template:white_check_mark::white_check_mark:
json-pointer:white_check_mark::white_check_mark:
date:white_check_mark:
time:white_check_mark:
regex:white_check_mark:
relative-json-pointer:white_check_mark:

The library also supports adding custom format validators. To use a custom validator basically you have to

Example

Lets assume the task is to create a custom validator which accepts strings with an even number of characters.

The custom FormatValidator will look something like this:

public class EvenCharNumValidator implements FormatValidator {

  @Override
  public Optional<String> validate(final String subject) {
    if (subject.length() % 2 == 0) {
      return Optional.empty();
    } else {
      return Optional.of(String.format("the length of string [%s] is odd", subject));
    }
  }

}

To bind the EvenCharNumValidator to a "format" value (for example "evenlength") you have to bind a validator instance to the keyword in the schema loader configuration:

JSONObject rawSchema = new JSONObject(new JSONTokener(inputStream));
SchemaLoader schemaLoader = SchemaLoader.builder()
	.schemaJson(rawSchema) // rawSchema is the JSON representation of the schema utilizing the "evenlength" non-standard format
	.addFormatValidator("evenlength", new EvenCharNumValidator()) // the EvenCharNumValidator gets bound to the "evenlength" keyword
	.build();
Schema schema = schemaLoader.load().build(); // the schema is created using the above created configuration
schema.validate(jsonDocument);  // the document validation happens here

$ref resolution

In a JSON Schema document it is possible to use relative URIs to refer previously defined types. Such references are expressed using the "$ref" and "$id" keywords. While the specification describes resolution scope alteration and dereferencing in detail, it doesn't explain the expected behavior when the first occurring "$ref" or "$id" is a relative URI.

In the case of this implementation it is possible to explicitly define an absolute URI serving as the base URI (resolution scope) using the appropriate builder method:

SchemaLoader schemaLoader = SchemaLoader.builder()
        .schemaJson(jsonSchema)
        .resolutionScope("http://example.org/") // setting the default resolution scope
        .build();

Loading from the classpath

As your schemas grow you will want to split that up into multiple source files and wire them with "$ref" references. If you want to store the schemas on the classpath (instead of eg. serving them through HTTP) then the recommended way is to use the classpath: protocol to make the schemas reference each other. To make the classpath: protocol work:

SchemaLoader schemaLoader = SchemaLoader.builder()
        .schemaClient(SchemaClient.classPathAwareClient())
        .schemaJson(jsonSchema)
        .resolutionScope("classpath://my/schemas/directory/") // setting the default resolution scope
        .build();

Given this configuration, the following references will be properly resolved in jsonSchema:

{
    "properties": {
        "sameDir": { "$ref": "sameDirSchema.json" },
        "absPath": { "$ref": "classpath://somewhere/else/otherschema.json" },
        "httpPath": { "$ref": "http://example.org/http-works-as-usual" },
    }
}

and sameDirSchema.json will be looked for in /my/schemas/directory/sameDirSchema.json on the classpath.

Registering schemas by URI

Sometimes it is useful to work with preloaded schemas, to which we assign an arbitary URI (maybe an uuid) instead of loading the schema through a URL. This can be done by assigning the schemas to a URI with the #registerSchemaByURI() method of the schema loader. Example:

SchemaLoader schemaLoader = SchemaLoader.builder()
        .registerSchemaByURI(new URI("urn:uuid:a773c7a2-1a13-4f6a-a70d-694befe0ce63"), aJSONObject)
        .registerSchemaByURI(new URI("http://example.org"), otherJSONObject)
        .schemaJson(jsonSchema)
        .resolutionScope("classpath://my/schemas/directory/")
        .build();

Notes:

Excluding dependencies

Some of the dependencies can be excluded from the library, and it still remains usable, with some limitations:

Javadoc

By library version:

The generated javadoc of versions 1.0.0 - 1.5.1 is available at javadoc.io

For the versions in between (1.6.0 - 1.9.1) it isn't published anywhere.