Karate

Test Automation Made Simple.

Karate is the only open-source tool to combine API test-automation, mocks, performance-testing and even UI automation into a single, unified framework. The syntax is language-neutral, and easy for even non-programmers. Assertions and HTML reports are built-in, and you can run tests in parallel for speed.

There's also a cross-platform stand-alone executable for teams not comfortable with Java. You don't have to compile code. Just write tests in a simple, readable syntax - carefully designed for HTTP, JSON, GraphQL and XML. And you can mix API and UI test-automation within the same test script.

A Java API also exists for those who prefer to programmatically integrate Karate's rich automation and data-assertion capabilities.

Hello World

For API Testing

If you are familiar with Cucumber / Gherkin, the big difference here is that you don't need to write extra "glue" code or Java "step definitions" !

It is worth pointing out that JSON is a 'first class citizen' of the syntax such that you can express payload and expected data without having to use double-quotes and without having to enclose JSON field names in quotes. There is no need to 'escape' characters like you would have had to in Java or other programming languages.

And you don't need to create additional Java classes for any of the payloads that you need to work with.

Index

Features

• Java knowledge is not required and even non-programmers can write tests
• Scripts are plain-text, require no compilation step or IDE, and teams can collaborate using Git / standard SCM
• Based on the popular Cucumber / Gherkin standard - with IDE support and syntax-coloring options
• Elegant DSL syntax 'natively' supports JSON and XML - including JsonPath and XPath expressions
• Eliminate the need for 'Java Beans' or 'helper code' to represent payloads and HTTP end-points, and dramatically reduce the lines of code needed for a test
• Ideal for testing the highly dynamic responses from GraphQL API-s because of Karate's built-in text-manipulation and JsonPath capabilities
• Tests are super-readable - as scenario data can be expressed in-line, in human-friendly JSON, XML, Cucumber Scenario Outline tables, or a payload builder approach unique to Karate
• Express expected results as readable, well-formed JSON or XML, and assert in a single step that the entire response payload (no matter how complex or deeply nested) - is as expected
• Comprehensive assertion capabilities - and failures clearly report which data element (and path) is not as expected, for easy troubleshooting of even large payloads
• Fully featured debugger that can step backwards and even re-play a step while editing it - a huge time-saver
• Simpler and more powerful alternative to JSON-schema for validating payload structure and format - that even supports cross-field / domain validation logic
• Scripts can call other scripts - which means that you can easily re-use and maintain authentication and 'set up' flows efficiently, across multiple tests
• Embedded JavaScript engine that allows you to build a library of re-usable functions that suit your specific environment or organization
• Re-use of payload-data and user-defined functions across tests is so easy - that it becomes a natural habit for the test-developer
• Built-in support for switching configuration across different environments (e.g. dev, QA, pre-prod)
• Support for data-driven tests and being able to tag or group tests is built-in, no need to rely on an external framework
• Native support for reading YAML and even CSV files - and you can use them for data-driven tests
• Standard Java / Maven project structure, and seamless integration into CI / CD pipelines - and support for JUnit 5
• Option to use as a light-weight stand-alone executable - convenient for teams not comfortable with Java
• Multi-threaded parallel execution, which is a huge time-saver, especially for integration and end-to-end tests
• Built-in test-reports compatible with Cucumber so that you have the option of using third-party (open-source) maven-plugins for even better-looking reports
• Reports include HTTP request and response logs in-line, which makes troubleshooting and debugging easier
• Easily invoke JDK classes, Java libraries, or re-use custom Java code if needed, for ultimate extensibility
• Simple plug-in system for authentication and HTTP header management that will handle any complex, real-world scenario
• Cross-browser Web UI automation so that you can test all layers of your application with the same framework
• [experimental] Android and iOS mobile support via Appium
• Visual Validation via the built-in image comparison capabilities
• Cross platform Desktop Automation that can be mixed into Web Automation flows if needed
• Option to invoke via a Java API, which means that you can easily mix Karate into Java projects or legacy UI-automation suites
• Save significant effort by re-using Karate test-suites as Gatling performance tests that deeply assert that server responses are accurate under load
• Gatling integration can hook into any custom Java code - which means that you can perf-test even non-HTTP protocols such as gRPC
• Built-in distributed-testing capability that works for API, UI and even load-testing - without needing any complex "grid" infrastructure
• API mocks or test-doubles that even maintain CRUD 'state' across multiple calls - enabling TDD for micro-services and Consumer Driven Contracts
• Async support that allows you to seamlessly integrate the handling of custom events or listening to message-queues
• Built-in HTML templating so that you can extend your test-reports into readable specifications
• Comprehensive support for different flavors of HTTP calls:
  • SOAP / XML requests
  • HTTPS / SSL - without needing certificates, key-stores or trust-stores
  • HTTP proxy server support
  • URL-encoded HTML-form data
  • Multi-part file-upload - including multipart/mixed and multipart/related
  • Browser-like cookie handling
  • Full control over HTTP headers, path and query parameters
  • Re-try until condition
  • Websocket support

Real World Examples

A set of real-life examples can be found here: Karate Demos

Comparison with REST-assured

For teams familiar with or currently using REST-assured, this detailed comparison of Karate vs REST-assured - can help you evaluate Karate. Do note that if you prefer a pure Java API - Karate has that covered, and with far more capabilities.

References

• API Testing with Karate - video + demos by Peter Thomas (creator / lead dev of Karate)
• Intro to all features of Karate - video + demos by Peter Thomas (creator / lead dev of Karate)
• Karate entered the ThoughtWorks Tech Radar in 2019 and was upgraded in ranking in May 2020
• マイクロサービスにおけるテスト自動化 with Karate - (Microservices Test Automation with Karate) presentation by Takanori Suzuki
• Writing API Tests with Karate - book by Benjamin Bischoff, Packt Publishing, 2023

Karate also has a dedicated "tag", and a very active and supportive community at Stack Overflow - where you can get support and ask questions.

You can find a lot more references, tutorials and blog-posts at karatelabs.io.

Getting Started

If you are a Java developer - Karate requires at least Java 11 and then either Maven, Gradle, or a Java IDE that embeds either to be installed. Note that Karate works fine on OpenJDK.

If you are new to programming or test-automation, the official IntelliJ plugin is recommended.

If you don't want to use Java, the Karate extension for Visual Studio Code is recommended, and JavaScript, .NET, Ruby and Python programmers will feel right at home.

Both the official Visual Studio Code and IntelliJ plugins support step-through debugging of Karate tests.

Maven

All you need is available in the karate-core artifact. You can run tests with this directly, but teams can choose the JUnit variant (shown below) that pulls in JUnit 5 and slightly improves the in-IDE experience.

<dependency>
    <groupId>com.intuit.karate</groupId>
    <artifactId>karate-junit5</artifactId>
    <version>1.4.0</version>
    <scope>test</scope>
</dependency>

Gradle

Alternatively for Gradle:

    testCompile 'com.intuit.karate:karate-junit5:1.4.0'

Also refer to the wiki for using Karate with Gradle.

Karate Core "Fat JAR"

If you mix Karate into a Maven or Gradle project with many other dependendies, you may run into problems because of dependency conflicts. For example a lot of Java projects directly (or indirectly) depend on Netty or Thymeleaf or ANTLR, etc.

If you face issues such as "class not found", just pull in the karate-core dependency, and use the all classifier in your pom.xml (or build.gradle).

For example when using Maven:

<dependency>
  <groupId>com.intuit.karate</groupId>
  <artifactId>karate-core</artifactId>
  <version>${karate.version}</version>
  <classifier>all</classifier>
  <scope>test</scope>
</dependency>

Note that for very complicated projects you can consider using a Maven profile so that testing-related dependencies don't collide with your development-time dependencies. Of course it is an option to have Karate tests in a separate stand-alone maven project and folder, while still being in the same Git repository.

Quickstart

It may be easier for you to use the Karate Maven archetype to create a skeleton project with one command. You can then skip the next few sections, as the pom.xml, recommended directory structure, sample test and JUnit 5 runners - will be created for you.

If you are behind a corporate proxy, or especially if your local Maven installation has been configured to point to a repository within your local network, the command below may not work. One workaround is to temporarily disable or rename your Maven settings.xml file, and try again.

You can replace the values of com.mycompany and myproject as per your needs.

mvn archetype:generate \
-DarchetypeGroupId=com.intuit.karate \
-DarchetypeArtifactId=karate-archetype \
-DarchetypeVersion=1.4.0 \
-DgroupId=com.mycompany \
-DartifactId=myproject

This will create a folder called myproject (or whatever you set the name to).

IDE Support

Refer to the wiki - IDE Support.

Folder Structure

A Karate test script has the file extension .feature which is the standard followed by Cucumber. You are free to organize your files using regular Java package conventions.

The Maven tradition is to have non-Java source files in a separate src/test/resources folder structure - but we recommend that you keep them side-by-side with your *.java files. When you have a large and complex project, you will end up with a few data files (e.g. *.js, *.json, *.txt) as well and it is much more convenient to see the *.java and *.feature files and all related artifacts in the same place.

This can be easily achieved with the following tweak to your maven <build> section.

<build>
    <testResources>
        <testResource>
            <directory>src/test/java</directory>
            <excludes>
                <exclude>**/*.java</exclude>
            </excludes>
        </testResource>
    </testResources>        
    <plugins>
    ...
    </plugins>
</build>

This is very common in the world of Maven users and keep in mind that these are tests and not production code.

Alternatively, if using Gradle then add the following sourceSets definition

sourceSets {
    test {
        resources {
            srcDir file('src/test/java')
            exclude '**/*.java'
        }
    }
}

With the above in place, you don't have to keep switching between your src/test/java and src/test/resources folders, you can have all your test-code and artifacts under src/test/java and everything will work as expected.

Once you get used to this, you may even start wondering why projects need a src/test/resources folder at all !

Spring Boot Example

Soumendra Daas has created a nice example and guide that you can use as a reference here: hello-karate. This demonstrates a Java Maven + JUnit 5 project set up to test a Spring Boot app.

Naming Conventions

Since these are tests and not production Java code, you don't need to be bound by the com.mycompany.foo.bar convention and the un-necessary explosion of sub-folders that ensues. We suggest that you have a folder hierarchy only one or two levels deep - where the folder names clearly identify which 'resource', 'entity' or API is the web-service under test.

For example:

src/test/java
    |
    +-- karate-config.js
    +-- logback-test.xml
    +-- some-reusable.feature
    +-- some-classpath-function.js
    +-- some-classpath-payload.json
    |
    \-- animals
        |
        +-- AnimalsTest.java
        |
        +-- cats
        |   |
        |   +-- cats-post.feature
        |   +-- cats-get.feature
        |   +-- cat.json
        |   \-- CatsRunner.java
        |
        \-- dogs
            |
            +-- dog-crud.feature
            +-- dog.json
            +-- some-helper-function.js
            \-- DogsRunner.java

Assuming you use JUnit, there are some good reasons for the recommended (best practice) naming convention and choice of file-placement shown above:

• Not using the *Test.java convention for the JUnit classes (e.g. CatsRunner.java) in the cats and dogs folder ensures that these tests will not be picked up when invoking mvn test (for the whole project) from the command line. But you can still invoke these tests from the IDE, which is convenient when in development mode.
• AnimalsTest.java (the only file that follows the *Test.java naming convention) acts as the 'test suite' for the entire project. By default, Karate will load all *.feature files from sub-directories as well. But since some-reusable.feature is above AnimalsTest.java in the folder hierarchy, it will not be picked-up. Which is exactly what we want, because some-reusable.feature is designed to be called only from one of the other test scripts (perhaps with some parameters being passed). You can also use tags to skip files.
• some-classpath-function.js and some-classpath-payload.json are in the 'root' of the Java 'classpath' which means they can be easily read (and re-used) from any test-script by using the classpath: prefix, for e.g: read('classpath:some-classpath-function.js'). Relative paths will also work.

For details on what actually goes into a script or *.feature file, refer to the syntax guide.

JUnit 5

Karate supports JUnit 5 and the advantage is that you can have multiple methods in a test-class. Only 1 import is needed, and instead of a class-level annotation, you use a nice DRY and fluent-api to express which tests and tags you want to use.

Note that the Java class does not need to be public and even the test methods do not need to be public - so tests end up being very concise.

Karate will traverse sub-directories and look for *.feature files. For example if you have the JUnit class in the com.mycompany package, *.feature files in com.mycompany.foo and com.mycompany.bar will also be run. This is one reason why you may want to prefer a 'flat' directory structure as explained above.

Here is an example:

package karate;

import com.intuit.karate.junit5.Karate;

class SampleTest {

    @Karate.Test
    Karate testSample() {
        return Karate.run("sample").relativeTo(getClass());
    }
    
    @Karate.Test
    Karate testTags() {
        return Karate.run("tags").tags("@second").relativeTo(getClass());
    }

    @Karate.Test
    Karate testSystemProperty() {
        return Karate.run("classpath:karate/tags.feature")
                .tags("@second")
                .karateEnv("e2e")
                .systemProperty("foo", "bar");
    }

}

Note that more "builder" methods are available from the Runner.Builder class such as reportDir() etc.

You should be able to right-click and run a single method using your IDE - which should be sufficient when you are in development mode. But to be able to run JUnit 5 tests from the command-line, you need to ensure that the latest version of the maven-surefire-plugin is present in your project pom.xml (within the <build>/<plugins> section):

<plugin>
    <groupId>org.apache.maven.plugins</groupId>
    <artifactId>maven-surefire-plugin</artifactId>
    <version>2.22.2</version>
</plugin>

To run a single test method, for example the testTags() in the example above, you can do this:

mvn test -Dtest=SampleTest#testTags

Also look at how to run tests via the command-line and the parallel runner.

JUnit HTML report

When you use the JUnit runner - after the execution of each feature, an HTML report is output to the target/karate-reports folder and the full path will be printed to the console (see video).

html report: (paste into browser to view)
-----------------------------------------
file:///projects/myproject/target/karate-reports/mypackage.myfeature.html

You can easily select (double-click), copy and paste this file: URL into your browser address bar. This report is useful for troubleshooting and debugging a test because all requests and responses are shown in-line with the steps, along with error messages and the output of print statements. Just re-fresh your browser window if you re-run the test.

Dry Run

This will give you the usual HTML report showing what features will be run, including all steps shown (including comments) so that it can be reviewed. Of course the actual time-durations, and logs will be missing, and everything will pass.

The “dry run” report is useful to review the tag "coverage" of what will be run. For example you can get a nice feature “coverage” report, provided you have a rich set of tags. e.g. @smoke @module=one @module=two etc.

The Runner.Builder API has a dryRun() method to switch this on. Note that this mode can be also triggered via the command-line by adding -D or --dryrun to the karate.options.

Command Line

Normally in dev mode, you will use your IDE to run a *.feature file directly or via the companion 'runner' JUnit Java class. When you have a 'runner' class in place, it would be possible to run it from the command-line as well.

Note that the mvn test command only runs test classes that follow the *Test.java naming convention by default. But you can choose a single test to run like this:

mvn test -Dtest=CatsRunner

karate.options

When your Java test "runner" is linked to multiple feature files, which will be the case when you use the recommended parallel runner, you can narrow down your scope to a single feature, scenario or directory via the command-line, useful in dev-mode. Note how even tags to exclude (or include) can be specified:

Note that any Feature or Scenario with the special @ignore tag will be skipped by default.

mvn test "-Dkarate.options=--tags ~@skipme classpath:demo/cats/cats.feature" -Dtest=DemoTestParallel

Multiple feature files (or paths) can be specified, de-limited by the space character. They should be at the end of the karate.options. To run only a single scenario, append the line number on which the scenario is defined, de-limited by :.

mvn test "-Dkarate.options=PathToFeatureFiles/order.feature:12" -Dtest=DemoTestParallel

Since paths are expected at the end of the command-line options - if you want to only over-ride tags, use the = sign to make argument values clear. For example:

mvn test -Dkarate.options='-t=@dev -t=@src' -Dtest=ExamplesTest

Command Line - Gradle

For Gradle, you must extend the test task to allow the karate.options to be passed to the runtime (otherwise they get consumed by Gradle itself). To do that, add the following:

test {
    // pull karate options into the runtime
    systemProperty "karate.options", System.properties.getProperty("karate.options")
    // pull karate env into the runtime
    systemProperty "karate.env", System.properties.getProperty("karate.env")
    // ensure tests are always run
    outputs.upToDateWhen { false }
}

And then the above command in Gradle would look like:

./gradlew test --tests *CatsRunner

or

./gradlew test -Dtest.single=CatsRunner

Test Suites

The recommended way to define and run test-suites and reporting in Karate is to use the parallel runner, described in the next section. The approach in this section is more suited for troubleshooting in dev-mode, using your IDE.

One way to define 'test-suites' in Karate is to have a JUnit class at a level 'above' (in terms of folder hierarchy) all the *.feature files in your project. So if you take the previous folder structure example, you can do this on the command-line:

mvn test "-Dkarate.options=--tags ~@skipme" -Dtest=AnimalsTest

Here, AnimalsTest is the name of the Java class we designated to run the multiple *.feature files that make up your test-suite. There is a neat way to tag your tests and the above example demonstrates how to run all tests except the ones tagged @skipme.

Note that the special, built-in tag @ignore will always be skipped by default, and you don't need to specify ~@ignore anywhere.

You can 'lock down' the fact that you only want to execute the single JUnit class that functions as a test-suite - by using the following maven-surefire-plugin configuration:

<plugin>
    <groupId>org.apache.maven.plugins</groupId>
    <artifactId>maven-surefire-plugin</artifactId>
    <version>${maven.surefire.version}</version>
    <configuration>
        <includes>
            <include>animals/AnimalsTest.java</include>
        </includes>
        <systemProperties>
            <karate.options>--tags @smoke</karate.options>
        </systemProperties>            
    </configuration>
</plugin> 

Note how the karate.options can be specified using the <systemProperties> configuration.

For Gradle, you simply specify the test which is to be include-d:

test {
    include 'animals/AnimalsTest.java'
    // pull karate options into the runtime
    systemProperty "karate.options", System.properties.getProperty("karate.options")
    // pull karate env into the runtime
    systemProperty "karate.env", System.properties.getProperty("karate.env")
    // ensure tests are always run
    outputs.upToDateWhen { false }
}

The big drawback of the approach above is that you cannot run tests in parallel. The recommended approach for Karate reporting in a Continuous Integration set-up is described in the next section which can generate the JUnit XML format that most CI tools can consume. The Cucumber JSON format can be also emitted, which gives you plenty of options for generating pretty reports using third-party maven plugins.

And most importantly - you can run tests in parallel without having to depend on third-party hacks that introduce code-generation and config 'bloat' into your pom.xml or build.gradle.

Parallel Execution

Please note that some user analytics is tracked only when you view the built-in Karate HTML report.

Karate can run tests in parallel, and dramatically cut down execution time. This is a 'core' feature and does not depend on JUnit, Maven or Gradle.

• You can easily "choose" features and tags to run and compose test-suites in a very flexible manner.
• You can use the returned Results object to check if any scenarios failed, and to even summarize the errors
• JUnit XML reports can be generated in the "reportDir" path you specify, and you can easily configure your CI to look for these files after a build (for e.g. in **/*.xml or **/karate-reports/*.xml). Note that you have to call the outputJunitXml(true) method on the Runner "builder".
• Cucumber JSON reports can be generated, except that the extension will be .json instead of .xml. Note that you have to call the outputCucumberJson(true) method on the Runner "builder".
• HTML reports can be disabled by calling outputHtmlReport(false).
• The Runner.path() "builder" method in karate-core is how you refer to the package you want to execute, and all feature files within sub-directories will be picked up
• Runner.path() takes multiple string parameters, so you can refer to multiple packages or even individual *.feature files and easily "compose" a test-suite
  • e.g. Runner.path("classpath:animals", "classpath:some/other/package.feature")
• To choose tags, call the tags() API, note that by default, any *.feature file tagged with the special (built-in) tag: @ignore will be skipped. You can also specify tags on the command-line. The tags() method also takes multiple arguments, for e.g.
  • this is an "AND" operation: tags("@customer", "@smoke")
  • and this is an "OR" operation: tags("@customer,@smoke")
• There is an optional reportDir() method if you want to customize the directory to which the HTML, XML and JSON files will be output, it defaults to target/karate-reports
• If you want to dynamically and programmatically determine the tags and features to be included - the API also accepts List<String> as the path() and tags() methods arguments
• parallel() has to be the last method called, and you pass the number of parallel threads needed. It returns a Results object that has all the information you need - such as the number of passed or failed tests.

JUnit 5 Parallel Execution

The example below assumes that JUnit 5 is available on the classpath, and uses the @Test annotation and the assertEquals() method.

But if you really want, you could use the Runner and Results API directly in any Java class, and even a "main" method.

Use the karate-template project if you want to get an example as part of a working, "skeleton" project.

import com.intuit.karate.Results;
import com.intuit.karate.Runner;
import static org.junit.jupiter.api.Assertions.*;
import org.junit.jupiter.api.Test;

class TestParallel {

    @Test
    void testParallel() {
        Results results = Runner.path("classpath:animals").tags("~@skipme").parallel(5);
        assertEquals(0, results.getFailCount(), results.getErrorMessages());
    }

}

Parallel Stats

For convenience, some stats are logged to the console when execution completes, which should look something like this:

======================================================
elapsed:   2.35 | threads:    5 | thread time: 4.98 
features:    54 | ignored:   25 | efficiency: 0.42
scenarios:  145 | passed:   145 | failed: 0
======================================================

The parallel runner will always run Feature-s in parallel. Karate will also run Scenario-s in parallel by default. So if you have a Feature with multiple Scenario-s in it - they will execute in parallel, and even each Examples row in a Scenario Outline will do so !

A karate-timeline.html file will also be saved to the report output directory mentioned above (target/karate-reports by default) - which is useful for visually verifying or troubleshooting the effectiveness of the test-run (see video).

@parallel=false

In rare cases you may want to suppress the default of Scenario-s executing in parallel and the special tag @parallel=false can be used. If you place it above the Feature keyword, it will apply to all Scenario-s. And if you just want one or two Scenario-s to NOT run in parallel, you can place this tag above only those Scenario-s. See example.

Note that forcing Scenario-s to run in a particular sequence is an anti-pattern, and should be avoided as far as possible.

Test Reports

As mentioned above, most CI tools would be able to process the JUnit XML output of the parallel runner and determine the status of the build as well as generate reports.

The Karate Demo has a working example of the recommended parallel-runner set up. It also details how a third-party library can be easily used to generate some very nice-looking reports, from the JSON output of the parallel runner.

For example, here below is an actual report generated by the cucumber-reporting open-source library.

Another example for a popular Maven reporting plugin that is compatible with Karate JSON is Cluecumber.

The demo also features code-coverage using Jacoco, and some tips for even non-Java back-ends. Some third-party report-server solutions integrate with Karate such as ReportPortal.io.

Logging

This is optional, and Karate will work without the logging config in place, but the default console logging may be too verbose for your needs.

Karate uses LOGBack which looks for a file called logback-test.xml on the 'classpath'.

In rare cases, e.g. if you are using Karate to create a Java application, LOGBack will look for logback.xml

Here is a sample logback-test.xml for you to get started.

<?xml version="1.0" encoding="UTF-8"?>
<configuration>
 
    <appender name="STDOUT" class="ch.qos.logback.core.ConsoleAppender">
        <encoder>
            <pattern>%d{HH:mm:ss.SSS} [%thread] %-5level %logger{36} - %msg%n</pattern>
        </encoder>
    </appender>
  
    <appender name="FILE" class="ch.qos.logback.core.FileAppender">
        <file>target/karate.log</file>
        <encoder>
            <pattern>%d{HH:mm:ss.SSS} [%thread] %-5level %logger{36} - %msg%n</pattern>
        </encoder>
    </appender>    
   
    <logger name="com.intuit.karate" level="DEBUG"/>
   
    <root level="info">
        <appender-ref ref="STDOUT" />
        <appender-ref ref="FILE" />
    </root>
  
</configuration>

You can change the com.intuit.karate logger level to INFO to reduce the amount of logging. When the level is DEBUG the entire request and response payloads are logged. If you use the above config, logs will be captured in target/karate.log.

If you want to keep the level as DEBUG (for HTML reports) but suppress logging to the console, you can comment out the STDOUT "root" appender-ref:

  <root level="warn">
      <!-- <appender-ref ref="STDOUT" /> -->
      <appender-ref ref="FILE" />
  </root>

Or another option is to use a ThresholdFilter, so you still see critical logs on the console:

  <appender name="STDOUT" class="ch.qos.logback.core.ConsoleAppender">
      <filter class="ch.qos.logback.classic.filter.ThresholdFilter">
          <level>WARN</level>
      </filter>
      <encoder>
          <pattern>%d{HH:mm:ss.SSS} [%thread] %-5level %logger{36} - %msg%n</pattern>
      </encoder>
  </appender>

If you want to exclude the logs from your CI/CD pipeline but keep them in the execution of your users in their locals you can configure your logback using Janino. In such cases it might be desirable to have your tests using karate.logger.debug('your additional info') instead of the print keyword so you can keep logs in your pipeline in INFO.

For suppressing sensitive information such as secrets and passwords from the log and reports, see Log Masking and Report Verbosity.

Configuration

You can skip this section and jump straight to the Syntax Guide if you are in a hurry to get started with Karate. Things will work even if the karate-config.js file is not present.

Classpath

The 'classpath' is a Java concept and is where some configuration files such as the one for logging are expected to be by default. If you use the Maven <test-resources> tweak described earlier (recommended), the 'root' of the classpath will be in the src/test/java folder, or else would be src/test/resources.

karate-config.js

The only 'rule' is that on start-up Karate expects a file called karate-config.js to exist on the 'classpath' and contain a JavaScript function. The function is expected to return a JSON object and all keys and values in that JSON object will be made available as script variables.

And that's all there is to Karate configuration ! You can easily get the value of the current 'environment' or 'profile', and then set up 'global' variables using some simple JavaScript. Here is an example:

function fn() {   
  var env = karate.env; // get java system property 'karate.env'
  karate.log('karate.env system property was:', env);
  if (!env) {
    env = 'dev'; // a custom 'intelligent' default
  }
  var config = { // base config JSON
    appId: 'my.app.id',
    appSecret: 'my.secret',
    someUrlBase: 'https://some-host.com/v1/auth/',
    anotherUrlBase: 'https://another-host.com/v1/'
  };
  if (env == 'stage') {
    // over-ride only those that need to be
    config.someUrlBase = 'https://stage-host/v1/auth';
  } else if (env == 'e2e') {
    config.someUrlBase = 'https://e2e-host/v1/auth';
  }
  // don't waste time waiting for a connection or if servers don't respond within 5 seconds
  karate.configure('connectTimeout', 5000);
  karate.configure('readTimeout', 5000);
  return config;
}

Here above, you see the karate.log(), karate.env and karate.configure() "helpers" being used. Note that the karate-config.js is re-processed for every Scenario and in rare cases, you may want to initialize (e.g. auth tokens) only once for all of your tests. This can be achieved using karate.callSingle().

A common requirement is to pass dynamic parameter values via the command line, and you can use the karate.properties['some.name'] syntax for getting a system property passed via JVM options in the form -Dsome.name=foo. Refer to the section on dynamic port numbers for an example.

You can even retrieve operating-system environment variables via Java interop as follows: var systemPath = java.lang.System.getenv('PATH');

This decision to use JavaScript for config is influenced by years of experience with the set-up of complicated test-suites and fighting with Maven profiles, Maven resource-filtering and the XML-soup that somehow gets summoned by the Maven AntRun plugin.

Karate's approach frees you from Maven, is far more expressive, allows you to eyeball all environments in one place, and is still a plain-text file. If you want, you could even create nested chunks of JSON that 'name-space' your config variables.

One way to appreciate Karate's approach is to think over what it takes to add a new environment-dependent variable (e.g. a password) into a test. In typical frameworks it could mean changing multiple properties files, maven profiles and placeholders, and maybe even threading the value via a dependency-injection framework - before you can even access the value within your test.

This approach is indeed slightly more complicated than traditional *.properties files - but you need this complexity. Keep in mind that these are tests (not production code) and this config is going to be maintained more by the dev or QE team instead of the 'ops' or operations team.

And there is no more worrying about Maven profiles and whether the 'right' *.properties file has been copied to the proper place.

Switching the Environment

There is only one thing you need to do to switch the environment - which is to set a Java system property.

By default, the value of karate.env when you access it within karate-config.js - would be null.

The recipe for doing this when running Maven from the command line is:

mvn test -DargLine="-Dkarate.env=e2e"

Or in Gradle:

./gradlew test -Dkarate.env=e2e

You can refer to the documentation of the Maven Surefire Plugin for alternate ways of achieving this, but the argLine approach is the simplest and should be more than sufficient for your Continuous Integration or test-automation needs.

Here's a reminder that running any single JUnit test via Maven can be done by:

mvn test -Dtest=CatsRunner

Where CatsRunner is the JUnit class name (in any package) you wish to run.

Karate is flexible, you can easily over-write config variables within the Java or JUnit "runner" - which is very convenient when in dev-mode or rapid-prototyping.

System.setProperty("karate.env", "pre-prod");

But the recommended way is to use the karateEnv(name, value) or systemProperty(name, value) API on the parallel-runner.

For advanced users, note that tags and the karate.env environment-switch can be "linked" using the special environment tags.

Environment Specific Config

When your project gets complex, you can have separate karate-config-<env>.js files that will be processed for that specific value of karate.env. This is especially useful when you want to maintain passwords, secrets or even URL-s specific for your local dev environment.

Make sure you configure your source code management system (e.g. Git) to ignore karate-config-*.js if needed.

There should always be karate-config.js in the "root" folder, even if you don't have any "common" config. In such cases, the function can do nothing or return an empty JSON. Learn more.

Here are the rules Karate uses on bootstrap (before every Scenario or Examples row in a Scenario Outline):

• if the system-property karate.config.dir was set, Karate will look in this folder for karate-config.js - and if found, will process it
• else if karate-config.js was not found in the above location (or karate.config.dir was not set), classpath:karate-config.js would be processed (this is the default / common case)
• if the karate.env system property was set
  • if karate.config.dir was set, Karate will also look for file:<karate.config.dir>/karate-config-<env>.js
  • else (if the karate.config.dir was not set), Karate will look for classpath:karate-config-<env>.js
• if the over-ride karate-config-<env>.js exists, it will be processed, and the configuration (JSON entries) returned by this function will over-ride any set by karate-config.js

Refer to the karate demo for an example.

karate-base.js

Advanced users who build frameworks on top of Karate have the option to supply a karate-base.js file that Karate will look for on the classpath:. This is useful when you ship a JAR file containing re-usable features and JavaScript / Java code and want to 'default' a few variables that teams can 'inherit' from. So an additional rule in the above flow of 'rules' (before the first step) is as follows:

• if classpath:karate-base.js exists - Karate will process this as a configuration source before anything else

Syntax Guide

Script Structure

Karate scripts are technically in 'Gherkin' format - but all you need to grok as someone who needs to test web-services are the three sections: Feature, Background and Scenario. There can be multiple Scenario-s in a *.feature file, and at least one should be present. The Background is optional.

Variables set using def in the Background will be re-set before every Scenario. If you are looking for a way to do something only once per Feature, take a look at callonce. On the other hand, if you are expecting a variable in the Background to be modified by one Scenario so that later ones can see the updated value - that is not how you should think of them, and you should combine your 'flow' into one scenario. Keep in mind that you should be able to comment-out a Scenario or skip some via tags without impacting any others. Note that the parallel runner will run Scenario-s in parallel, which means they can run in any order. If you are looking for ways to do something only once per feature or across all your tests, see Hooks.

Lines that start with a # are comments.

Feature: brief description of what is being tested
    more lines of description if needed.

Background:
  # this section is optional !
  # steps here are executed before each Scenario in this file
  # variables defined here will be 'global' to all scenarios
  # and will be re-initialized before every scenario

Scenario: brief description of this scenario
  # steps for this scenario

Scenario: a different scenario
  # steps for this other scenario

There is also a variant of Scenario called Scenario Outline along with Examples, useful for data-driven tests.

Given-When-Then

The business of web-services testing requires access to low-level aspects such as HTTP headers, URL-paths, query-parameters, complex JSON or XML payloads and response-codes. And Karate gives you control over these aspects with the small set of keywords focused on HTTP such as url, path, param, etc.

Karate does not attempt to have tests be in "natural language" like how Cucumber tests are traditionally expected to be. That said, the syntax is very concise, and the convention of every step having to start with either Given, And, When or Then, makes things very readable. You end up with a decent approximation of BDD even though web-services by nature are "headless", without a UI, and not really human-friendly.

Cucumber vs Karate

Karate was based on Cucumber-JVM until version 0.8.0 but the parser and engine were re-written from scratch in 0.9.0 onwards. So we use the same Gherkin syntax - but the similarity ends there.

If you are familiar with Cucumber (JVM), you may be wondering if you need to write step-definitions. The answer is no.

Karate's approach is that all the step-definitions you need in order to work with HTTP, JSON and XML have been already implemented. And since you can easily extend Karate using JavaScript, there is no need to compile Java code any more.

The following table summarizes some key differences between Cucumber and Karate.

One nice thing about the design of the Gherkin syntax is that script-steps are treated the same no matter whether they start with the keyword Given, And, When or Then. What this means is that you are free to use whatever makes sense for you. You could even have all the steps start with When and Karate won't care.

In fact Gherkin supports the catch-all symbol '*' - instead of forcing you to use Given, When or Then. This is perfect for those cases where it really doesn't make sense - for example the Background section or when you use the def or set syntax. When eyeballing a test-script, think of the * as a 'bullet-point'.

You can read more about the Given-When-Then convention at the Cucumber reference documentation. Since Karate uses Gherkin, you can also employ data-driven techniques such as expressing data-tables in test scripts. Another good thing that Karate inherits is the nice IDE support for Cucumber that IntelliJ and Eclipse have. So you can do things like right-click and run a *.feature file (or scenario) without needing to use a JUnit runner.

For a detailed discussion on BDD and how Karate relates to Cucumber, please refer to this blog-post: Yes, Karate is not true BDD. It is the opinion of the author of Karate that true BDD is un-necessary over-kill for API testing, and this is explained more in this answer on Stack Overflow.

With the formalities out of the way, let's dive straight into the syntax.

Setting and Using Variables

def

Set a named variable

# assigning a string value:
Given def myVar = 'world'

# using a variable
Then print myVar

# assigning a number (you can use '*' instead of Given / When / Then)
* def myNum = 5

Note that def will over-write any variable that was using the same name earlier. Keep in mind that the start-up configuration routine could have already initialized some variables before the script even started. For details of scope and visibility of variables, see Script Structure.

Note that url and request are not allowed as variable names. This is just to reduce confusion for users new to Karate who tend to do * def request = {} and expect the request body or similarly, the url to be set.

The examples above are simple, but a variety of expression 'shapes' are supported on the right hand side of the = symbol. The section on Karate Expressions goes into the details.

assert

Assert if an expression evaluates to true

Once defined, you can refer to a variable by name. Expressions are evaluated using the embedded JavaScript engine. The assert keyword can be used to assert that an expression returns a boolean value.

Given def color = 'red '
And def num = 5
Then assert color + num == 'red 5'

Everything to the right of the assert keyword will be evaluated as a single expression.

Something worth mentioning here is that you would hardly need to use assert in your test scripts. Instead you would typically use the match keyword, that is designed for performing powerful assertions against JSON and XML response payloads.

print

Log to the console

You can use print to log variables to the console in the middle of a script. For convenience, you can have multiple expressions separated by commas, so this is the recommended pattern:

* print 'the value of a is:', a

Similar to assert, the expressions on the right-hand-side of a print have to be valid JavaScript. JsonPath and Karate expressions are not supported.

If you use commas (instead of concatenating strings using +), Karate will 'pretty-print' variables, which is what you typically want when dealing with JSON or XML.

* def myJson = { foo: 'bar', baz: [1, 2, 3] }
* print 'the value of myJson is:', myJson

Which results in the following output:

20:29:11.290 [main] INFO  com.intuit.karate - [print] the value of myJson is: {
  "foo": "bar",
  "baz": [
    1,
    2,
    3
  ]
}

Since XML is represented internally as a JSON-like or map-like object, if you perform string concatenation when printing, you will not see XML - which can be confusing at first. Use the comma-delimited form (see above) or the JS helper (see below).

The built-in karate object is explained in detail later, but for now, note that this is also injected into print (and even assert) statements, and it has a helpful pretty method, that takes a JSON argument and a prettyXml method that deals with XML. So you could have also done something like:

* print 'the value of myJson is:\n' + karate.pretty(myJson)

Also refer to the configure keyword on how to switch on pretty-printing of all HTTP requests and responses.

'Native' data types

Native data types mean that you can insert them into a script without having to worry about enclosing them in strings and then having to 'escape' double-quotes all over the place. They seamlessly fit 'in-line' within your test script.

JSON

Note that the parser is 'lenient' so that you don't have to enclose all keys in double-quotes.

* def cat = { name: 'Billie', scores: [2, 5] }
* assert cat.scores[1] == 5

Some characters such as the hyphen - are not permitted in 'lenient' JSON keys (because they are interpreted by the JS engine as a 'minus sign'). In such cases, you have to use string quotes: { 'Content-Type': 'application/json' }

When asserting for expected values in JSON or XML, always prefer using match instead of assert. Match failure messages are much more descriptive and useful, and you get the power of embedded expressions and fuzzy matching.

* def cats = [{ name: 'Billie' }, { name: 'Bob' }]
* match cats[1] == { name: 'Bob' }

Karate's native support for JSON means that you can assign parts of a JSON instance into another variable, which is useful when dealing with complex response payloads.

* def first = cats[0]
* match first == { name: 'Billie' }

For manipulating or updating JSON (or XML) using path expressions, refer to the set keyword.

XML

Given def cat = <cat><name>Billie</name><scores><score>2</score><score>5</score></scores></cat>
# sadly, xpath list indexes start from 1
Then match cat/cat/scores/score[2] == '5'
# but karate allows you to traverse xml like json !!
Then match cat.cat.scores.score[1] == 5

Embedded Expressions

Karate has a very useful payload 'templating' approach. Variables can be referred to within JSON, for example:

Given def user = { name: 'john', age: 21 }
And def lang = 'en'
When def session = { name: '#(user.name)', locale: '#(lang)', sessionUser: '#(user)'  }

So the rule is - if a string value within a JSON (or XML) object declaration is enclosed between #( and ) - it will be evaluated as a JavaScript expression. And any variables which are alive in the context can be used in this expression. Here's how it works for XML:

Given def user = <user><name>john</name></user>
And def lang = 'en'
When def session = <session><locale>#(lang)</locale><sessionUser>#(user)</sessionUser></session>

This comes in useful in some cases - and avoids needing to use the set keyword or JavaScript functions to manipulate JSON. So you get the best of both worlds: the elegance of JSON to express complex nested data - while at the same time being able to dynamically plug values (that could even be other JSON or XML 'trees') into a 'template'.

Note that embedded expressions will be evaluated even when you read() from a JSON or XML file. This is super-useful for re-use and data-driven tests.

A few special built-in variables such as $ (which is a reference to the JSON root) - can be mixed into JSON embedded expressions.

A special case of embedded expressions can remove a JSON key (or XML element / attribute) if the expression evaluates to null.

Rules for Embedded Expressions

• They work only within JSON or XML
• and when on the Right Hand Side of a
  • def
  • match
  • configure
• and when you read() a JSON or XML file
• the expression has to start with #( and end with )

Because of the last rule above, note that string-concatenation may not work quite the way you expect:

# wrong !
* def foo = { bar: 'hello #(name)' }
# right !
* def foo = { bar: '#("hello " + name)' }

Observe how you can achieve string concatenation if you really want, because any valid JavaScript expression can be stuffed within an embedded expression. You could always do this in two steps:

* def temp = 'hello ' + name
* def foo = { bar: '#(temp)' }

As a convenience, embedded expressions are supported on the Right Hand Side of a match statement even for "quoted string" literals:

* def foo = 'a1'
* match foo == '#("a" + 1)'

And do note that in Karate 1.0 onwards, ES6 string-interpolation within "backticks" is supported:

* param filter = `ORDER_DATE:"${todaysDate}"`

Enclosed JavaScript

An alternative to embedded expressions (for JSON only) is to enclose the entire payload within parentheses - which tells Karate to evaluate it as pure JavaScript. This can be a lot simpler than embedded expressions in many cases, and JavaScript programmers will feel right at home.

The example below shows the difference between embedded expressions and enclosed JavaScript:

When def user = { name: 'john', age: 21 }
And def lang = 'en'

* def embedded = { name: '#(user.name)', locale: '#(lang)', sessionUser: '#(user)' }
* def enclosed = ({ name: user.name, locale: lang, sessionUser: user })
* match embedded == enclosed

So how would you choose between the two approaches to create JSON ? Embedded expressions are useful when you have complex JSON read from files, because you can auto-replace (or even remove) data-elements with values dynamically evaluated from variables. And the JSON will still be 'well-formed', and editable in your IDE or text-editor. Embedded expressions also make more sense in validation and schema-like short-cut situations. It can also be argued that the # symbol is easy to spot when eyeballing your test scripts - which makes things more readable and clear.

Multi-Line Expressions

The keywords def, set, match, request and eval take multi-line input as the last argument. This is useful when you want to express a one-off lengthy snippet of text in-line, without having to split it out into a separate file. Note how triple-quotes (""") are used to enclose content. Here are some examples:

# instead of:
* def cat = <cat><name>Billie</name><scores><score>2</score><score>5</score></scores></cat>

# this is more readable:
* def cat = 
  """
  <cat>
      <name>Billie</name>
      <scores>
          <score>2</score>
          <score>5</score>
      </scores>
  </cat>
  """
# example of a request payload in-line
Given request 
  """ 
  <?xml version='1.0' encoding='UTF-8'?>
  <S:Envelope xmlns:S="http://schemas.xmlsoap.org/soap/envelope/">
  <S:Body>
  <ns2:QueryUsageBalance xmlns:ns2="http://www.mycompany.com/usage/V1">
      <ns2:UsageBalance>
          <ns2:LicenseId>12341234</ns2:LicenseId>
      </ns2:UsageBalance>
  </ns2:QueryUsageBalance>
  </S:Body>
  </S:Envelope>
  """

# example of a payload assertion in-line
Then match response ==
  """
  { id: { domain: "DOM", type: "entityId", value: "#ignore" },
    created: { on: "#ignore" }, 
    lastUpdated: { on: "#ignore" },
    entityState: "ACTIVE"
  }
  """

table

A simple way to create JSON Arrays

Now that we have seen how JSON is a 'native' data type that Karate understands, there is a very nice way to create JSON using the support for expressing data-tables.

* table cats
  | name   | age |
  | 'Bob'  | 2   |
  | 'Wild' | 4   |
  | 'Nyan' | 3   |

* match cats == [{name: 'Bob', age: 2}, {name: 'Wild', age: 4}, {name: 'Nyan', age: 3}]

The match keyword is explained later, but it should be clear right away how convenient the table keyword is. JSON can be combined with the ability to call other *.feature files to achieve dynamic data-driven testing in Karate.

Notice that in the above example, string values within the table need to be enclosed in quotes. Otherwise they would be evaluated as expressions - which does come in useful for some dynamic data-driven situations:

* def one = 'hello'
* def two = { baz: 'world' }
* table json
  | foo     | bar            |
  | one     | { baz: 1 }     |
  | two.baz | ['baz', 'ban'] |
* match json == [{ foo: 'hello', bar: { baz: 1 } }, { foo: 'world', bar: ['baz', 'ban'] }]

Yes, you can even nest chunks of JSON in tables, and things work as you would expect.

Empty cells or expressions that evaluate to null will result in the key being omitted from the JSON. To force a null value, wrap it in parentheses:

* def one = { baz: null }
* table json
  | foo     | bar    |
  | 'hello' |        |
  | one.baz | (null) |
  | 'world' | null   |
* match json == [{ foo: 'hello' }, { bar: null }, { foo: 'world' }]

An alternate way to create data is using the set multiple syntax. It is actually a 'transpose' of the table approach, and can be very convenient when there are a large number of keys per row or if the nesting is complex. Here is an example of what is possible:

* set search
  | path       | 0        | 1      | 2       |
  | name.first | 'John'   | 'Jane' |         |
  | name.last  | 'Smith'  | 'Doe'  | 'Waldo' |
  | age        | 20       |        |         |

* match search[0] == { name: { first: 'John', last: 'Smith' }, age: 20 }
* match search[1] == { name: { first: 'Jane', last: 'Doe' } }
* match search[2] == { name: { last: 'Waldo' } }

text

Don't parse, treat as raw text

Not something you would commonly use, but in some cases you need to disable Karate's default behavior of attempting to parse anything that looks like JSON (or XML) when using multi-line / string expressions. This is especially relevant when manipulating GraphQL queries - because although they look suspiciously like JSON, they are not, and tend to confuse Karate's internals. And as shown in the example below, having text 'in-line' is useful especially when you use the Scenario Outline: and Examples: for data-driven tests involving place-holder substitutions in strings.

Scenario Outline:
  # note the 'text' keyword instead of 'def'
  * text query =
    """
    {
      hero(name: "<name>") {
        height
        mass
      }
    }
    """
  Given path 'graphql'
  And request { query: '#(query)' }
  And header Accept = 'application/json'
  When method post
  Then status 200

  Examples:
    | name  |
    | John  |
    | Smith | 

Note that if you did not need to inject Examples: into 'placeholders' enclosed within < and >, reading from a file with the extension *.txt may have been sufficient.

For placeholder-substitution, the replace keyword can be used instead, but with the advantage that the text can be read from a file or dynamically created.

Karate is a great fit for testing GraphQL because of how easy it is to deal with dynamic and deeply nested JSON responses. Refer to this example for more details: graphql.feature.

replace

Text Placeholder Replacement

Modifying existing JSON and XML is natively supported by Karate via the set keyword, and replace is primarily intended for dealing with raw strings. But when you deal with complex, nested JSON (or XML) - it may be easier in some cases to use replace, especially when you want to substitute multiple placeholders with one value, and when you don't need array manipulation. Since replace auto-converts the result to a string, make sure you perform type conversion back to JSON (or XML) if applicable.

Karate provides an elegant 'native-like' experience for placeholder substitution within strings or text content. This is useful in any situation where you need to concatenate dynamic string fragments to form content such as GraphQL or SQL.

The placeholder format defaults to angle-brackets, for example: <replaceMe>. Here is how to replace one placeholder at a time:

* def text = 'hello <foo> world'
* replace text.foo = 'bar'
* match text == 'hello bar world'

Karate makes it really easy to substitute multiple placeholders in a single, readable step as follows:

* def text = 'hello <one> world <two> bye'

* replace text
  | token | value   |
  | one   | 'cruel' |
  | two   | 'good'  |

* match text == 'hello cruel world good bye'

Note how strings have to be enclosed in quotes. This is so that you can mix expressions into text replacements as shown below. This example also shows how you can use a custom placeholder format instead of the default:

* def text = 'hello <one> world ${two} bye'
* def first = 'cruel'
* def json = { second: 'good' }

* replace text
    | token  | value       |
    | one    | first       |
    | ${two} | json.second |

* match text == 'hello cruel world good bye'

Refer to this file for a detailed example: replace.feature

YAML Files

For those who may prefer YAML as a simpler way to represent data, Karate allows you to read YAML content from a file - and it will be auto-converted into JSON.

# yaml from a file (the extension matters), and the data-type of 'bar' would be JSON
* def bar = read('data.yaml')

yaml

A very rare need is to be able to convert a string which happens to be in YAML form into JSON, and this can be done via the yaml type cast keyword. For example - if a response data element or downloaded file is YAML and you need to use the data in subsequent steps. Also see type conversion.

* text foo =
  """
  name: John
  input:
    id: 1
    subType: 
      name: Smith
      deleted: false
  """
# yaml to json type conversion  
* yaml foo = foo
* match foo ==
  """
  {
    name: 'John',
    input: { 
      id: 1,
      subType: { name: 'Smith', deleted: false }    
    }
  }
  """

CSV Files

Karate can read *.csv files and will auto-convert them to JSON. A header row is always expected. See the section on reading files - and also this example dynamic-csv.feature, which shows off the convenience of dynamic Scenario Outline-s.

In rare cases you may want to use a csv-file as-is and not auto-convert it to JSON. A good example is when you want to use a CSV file as the request-body for a file-upload. You could get by by renaming the file-extension to say *.txt but an alternative is to use the karate.readAsString() API.

csv

Just like yaml, you may occasionally need to convert a string which happens to be in CSV form into JSON, and this can be done via the csv keyword.

* text foo =
    """
    name,type
    Billie,LOL
    Bob,Wild
    """
* csv bar = foo
* match bar == [{ name: 'Billie', type: 'LOL' }, { name: 'Bob', type: 'Wild' }]

JavaScript Functions

JavaScript Functions are also 'native'. And yes, functions can take arguments.

Standard JavaScript syntax rules apply, but the right-hand-side should begin with the function keyword if declared in-line. When using stand-alone *.js files, you can have a comment before the function keyword, and you can use fn as the function name, so that your IDE does not complain about JavaScript syntax errors, e.g. function fn(x){ return x + 1 }

* def greeter = function(title, name) { return 'hello ' + title + ' ' + name }
* assert greeter('Mr.', 'Bob') == 'hello Mr. Bob'

When JavaScript executes in Karate, the built-in karate object provides some commonly used utility functions. And with Karate expressions, you can "dive into" JavaScript without needing to define a function - and conditional logic is a good example.

Java Interop

For more complex functions you are better off using the multi-line 'doc-string' approach. This example actually calls into existing Java code, and being able to do this opens up a whole lot of possibilities. The JavaScript interpreter will try to convert types across Java and JavaScript as smartly as possible. For e.g. JSON objects become Java Map-s, JSON arrays become Java List-s, and Java Bean properties are accessible (and update-able) using 'dot notation' e.g. 'object.name'

* def dateStringToLong =
  """
  function(s) {
    var SimpleDateFormat = Java.type('java.text.SimpleDateFormat');
    var sdf = new SimpleDateFormat("yyyy-MM-dd'T'HH:mm:ss.SSSZ");
    return sdf.parse(s).time; // '.getTime()' would also have worked instead of '.time'
  } 
  """
* assert dateStringToLong("2016-12-24T03:39:21.081+0000") == 1482550761081

More examples of Java interop and how to invoke custom code can be found in the section on Calling Java.

The call keyword provides an alternate way of calling JavaScript functions that have only one argument. The argument can be provided after the function name, without parentheses, which makes things slightly more readable (and less cluttered) especially when the solitary argument is JSON.

* def timeLong = call dateStringToLong '2016-12-24T03:39:21.081+0000'
* assert timeLong == 1482550761081

# a better example, with a JSON argument
* def greeter = function(name){ return 'Hello ' + name.first + ' ' + name.last + '!' }
* def greeting = call greeter { first: 'John', last: 'Smith' }

Reading Files

Karate makes re-use of payload data, utility-functions and even other test-scripts as easy as possible. Teams typically define complicated JSON (or XML) payloads in a file and then re-use this in multiple scripts. Keywords such as set and remove allow you to to 'tweak' payload-data to fit the scenario under test. You can imagine how this greatly simplifies setting up tests for boundary conditions. And such re-use makes it easier to re-factor tests when needed, which is great for maintainability.

Note that the set (multiple) keyword can build complex, nested JSON (or XML) from scratch in a data-driven manner, and you may not even need to read from files for many situations. Test data can be within the main flow itself, which makes scripts highly readable.

Reading files is achieved using the built-in JavaScript function called read(). By default, the file is expected to be in the same folder (package) and side-by-side with the *.feature file. But you can prefix the name with classpath: in which case the 'root' folder would be src/test/java (assuming you are using the recommended folder structure).

Prefer classpath: when a file is expected to be heavily re-used all across your project. And yes, relative paths will work.

# json
* def someJson = read('some-json.json')
* def moreJson = read('classpath:more-json.json')

# xml
* def someXml = read('../common/my-xml.xml')

# import yaml (will be converted to json)
* def jsonFromYaml = read('some-data.yaml')

# csv (will be converted to json)
* def jsonFromCsv = read('some-data.csv')

# string
* def someString = read('classpath:messages.txt')

# javascript (will be evaluated)
* def someValue = read('some-js-code.js')

# if the js file evaluates to a function, it can be re-used later using the 'call' keyword (or invoked just like normal js)
* def someFunction = read('classpath:some-reusable-code.js')
* def someCallResult = call someFunction
* def sameCallResult = someFunction()

# the following short-cut is also allowed
* def someCallResult = call read('some-js-code.js')

You can also re-use other *.feature files from test-scripts:

# perfect for all those common authentication or 'set up' flows
* def result = call read('classpath:some-reusable-steps.feature')

When a called feature depends on some side-by-side resources such as JSON or JS files, you can use the this: prefix to ensure that relative paths work correctly - because by default Karate calculates relative paths from the "root" feature or the top-most "caller".

* def data = read('this:payload.json')

If a file does not end in .json, .xml, .yaml, .js, .csv or .txt, it is treated as a stream - which is typically what you would need for multipart file uploads.

* def someStream = read('some-pdf.pdf')

The .graphql and .gql extensions are also recognized (for GraphQL) but are handled the same way as .txt and treated as a string.

For JSON and XML files, Karate will evaluate any embedded expressions on load. This enables more concise tests, and the file can be re-usable in multiple, data-driven tests.

Since it is internally implemented as a JavaScript function, you can mix calls to read() freely wherever JavaScript expressions are allowed:

* def someBigString = read('first.txt') + read('second.txt')

Tip: you can even use JS expressions to dynamically choose a file based on some condition: * def someConfig = read('my-config-' + someVariable + '.json'). Refer to conditional logic for more ideas.

And a very common need would be to use a file as the request body:

Given request read('some-big-payload.json')

Or in a match:

And match response == read('expected-response-payload.json')

The rarely used file: prefix is also supported. You could use it for 'hard-coded' absolute paths in dev mode, but is obviously not recommended for CI test-suites. A good example of where you may need this is if you programmatically write a file to the target folder, and then you can read it like this:

* def payload = read('file:target/large.xml')

Path Prefixes

To summarize the possible prefixes:

Take a look at the Karate Demos for real-life examples of how you can use files for validating HTTP responses, like this one: read-files.feature.

Read File As String

In some rare cases where you don't want to auto-convert JSON, XML, YAML or CSV, and just get the raw string content (without having to re-name the file to end with .txt) - you can use the karate.readAsString() API. Here is an example of using a CSV file as the request-body:

Given path 'upload'
And header Content-Type = 'text/csv'
And request karate.readAsString('classpath:my.csv')
When method post
Then status 202

Compare Image

Karate provides a flexible way to compare two images to determine if they are the same or similar. This is especially useful when capturing screenshots during tests and comparing against baseline images that are known to be correct.

A stand-alone example can be found here: examples/image-comparison along with a video explanation.

Below is a simple example that will compare a baseline image to a more recent latest image. An image comparison UI will also be embedded into the Karate HTML report with detailed information about any differences between the two images.

* compareImage { baseline: 'screenshots/login.png', latest: '/tmp/login.png' }

You can also compare images using Karate path prefixes (e.g. classpath:, this:, file:) or byte arrays:

* def latestImgBytes = karate.readAsBytes('login.png')
* compareImage { baseline: 'classpath:screenshots/login.png', latest: '#(latestImgBytes)' }

You may configure the following image comparison options using the configure action:

* configure imageComparison = { /* image comparison options ... */ }

Image comparison configuration options:

Examples:

# use only 'ssim' (structural similarity) engine
* configure imageComparison = { engine: 'ssim' }

# always use both 'resemble' and 'ssim' engines but only evaluate the lowest mismatch percentage against our `failureThreshold`
* configure imageComparison = { engine: 'resemble,ssim' }

# prefer 'resemble' and fallback to 'ssim' engine only if the resemble mismatch percentage is >= `failureThreshold`
* configure imageComparison = { engine: 'resemble|ssim' }

# only consider the comparison as failed when 2% or more pixels are different from the baseline
* configure imageComparison = { failureThreshold: 2 }

# consider image comparisons that fail due to too many mismatched pixels as passed (especially useful when you are first starting without any baseline images)
* configure imageComparison = { mismatchShouldPass: true }

# custom JS function called in Karate HTML image comparison UI when the user clicks the `Rebase` button
* text onShowRebaseFn =
"""
function (config, downloadLatestFn) {
  // trigger download of latest image with custom file name
  downloadLatestFn('custom_latest.png')
  return 'this text will be displayed to the user when they click the rebase button'
}
"""
* configure imageComparison = { onShowRebase: '#(onShowRebaseFn)' }

# custom JS function called in Karate HTML image comparison UI when the user clicks the `Show config` button
* text onShowConfigFn =
"""
function (customConfigJson, config) {
  return 'this text will be displayed above the image comparison config\n' + customConfigJson
}
"""
* configure imageComparison = { onShowConfig: '#(onShowConfigFn)' }

# don't embed the image comparison UI when the latest image is the same / similar to the baseline (e.g. to save space and speed up report loading)
* configure imageComparison = { hideUiOnSuccess: true }

Image comparison engines can also be customized:

* compareImage { baseline: 'baseline.png', latest: 'latest.png', options: { /* engine options ... */ } }

Image comparison configuration options:

Examples:

# ignore areas of an image (e.g. to avoid constant failures due to loading animations)
* def boxes =
"""
[{
  top: 483,
  left: 1085,
  bottom: 893,
  right: 1496
}]
"""
* compareImage { baseline: 'baseline.png', latest: 'latest.png', options: { ignoredBoxes: #(boxes) } }

#############################
### Resemble-only options ###
#############################

# zero-tolerance for color shifts
* compareImage { baseline: 'baseline.png', latest: 'latest.png', options: { ignore: 'nothing' } }

# ignore all purple areas
* def purple =
"""
{
  r: 190,
  g: 0,
  b: 255
}
"""
* compareImage { baseline: 'baseline.png', latest: 'latest.png', options: { ignoreAreasColoredWith: '#(purple)' } }

# compare images as grayscale
* compareImage { baseline: 'baseline.png', latest: 'latest.png', options: { ignoreColors: true } }

# attempt to detect and ignore antialiasing
* compareImage { baseline: 'baseline.png', latest: 'latest.png', options: { ignoreAntialiasing: true } }

# customize color / brightness tolerances
* def customTolerances =
"""
{
  red: 4,
  green: 4,
  blue: 4,
  alpha: 4,
  minBrightness: 4,
  maxBrightness: 250
}
* compareImage { baseline: 'baseline.png', latest: 'latest.png', options: { tolerances: '#(customTolerances)' } }

#########################
### SSIM-only options ###
#########################

# switch to `fast` SSIM algorithm
* compareImage { baseline: 'baseline.png', latest: 'latest.png', options: { ssim: 'FAST' } }

# switch to `original` grayscale SSIM algorithm
* compareImage { baseline: 'baseline.png', latest: 'latest.png', options: { rgb2grayVersion: 'ORIGINAL' } }

# update SSIM stability constants
* compareImage { baseline: 'baseline.png', latest: 'latest.png', options: { k1: 0, k2: 0 } }

# update SSIM window size
* compareImage { baseline: 'baseline.png', latest: 'latest.png', options: { windowSize: 3 } }

# update SSIM bit depth
* compareImage { baseline: 'baseline.png', latest: 'latest.png', options: { bitDepth: 16 } }

# update SSIM max size
* compareImage { baseline: 'baseline.png', latest: 'latest.png', options: { maxSize: 512 } }

Type Conversion

Best practice is to stick to using only def unless there is a very good reason to do otherwise.

Internally, Karate will auto-convert JSON (and even XML) to Java Map objects. And JSON arrays would become Java List-s. But you will never need to worry about this internal data-representation most of the time.

In some rare cases, for e.g. if you acquired a string from some external source, or if you generated JSON (or XML) by concatenating text or using replace, you may want to convert a string to JSON and vice-versa. You can even perform a conversion from XML to JSON if you want.

One example of when you may want to convert JSON (or XML) to a string is when you are passing a payload to custom code via Java interop. Do note that when passing JSON, the default Map and List representations should suffice for most needs (see example), and using them would avoid un-necessary string-conversion.

So you have the following type markers you can use instead of def (or the rarely used text). The first four below are best explained in this example file: type-conv.feature.

• string - convert JSON or any other data-type (except XML) to a string
• json - convert XML, a map-like or list-like object, a string, or even a Java object into JSON
• xml - convert JSON, a map-like object, a string, or even a Java object into XML
• xmlstring - specifically for converting the map-like Karate internal representation of XML into a string
• csv - convert a CSV string into JSON, see csv
• yaml - convert a YAML string into JSON, see yaml
• bytes - convert to a byte-array, useful for binary payloads or comparisons, see example
• copy - to clone a given payload variable reference (JSON, XML, Map or List), refer: copy

The csv and yaml types can be initialized in-line using the "triple quote" or "docstring" multi-line approach as shown here.

If you want to 'pretty print' a JSON or XML value with indenting, refer to the documentation of the print keyword.

Floats and Integers

While converting a number to a string is easy (just concatenate an empty string e.g. myInt + ''), in some rare cases, you may need to convert a string to a number. You can do this by multiplying by 1 or using the built-in JavaScript parseInt() function:

* def foo = '10'
* string json = { bar: '#(1 * foo)' }
* match json == '{"bar":10.0}'

* string json = { bar: '#(parseInt(foo))' }
* match json == '{"bar":10.0}'

As per the JSON spec, all numeric values are treated as doubles, so for integers - it really doesn't matter if there is a decimal point or not. In fact it may be a good idea to slip doubles instead of integers into some of your tests ! Anyway, there are times when you may want to force integers (perhaps for cosmetic reasons) and you can easily do so using the 'double-tilde' short-cut: '~~'.

* def foo = '10'
* string json = { bar: '#(~~foo)' }
* match json == '{"bar":10}'

# JS math can introduce a decimal point in some cases
* def foo = 100
* string json = { bar: '#(foo * 0.1)' }
* match json == '{"bar":10.0}'

# but you can easily coerce to an integer if needed
* string json = { bar: '#(~~(foo * 0.1))' }
* match json == '{"bar":10}'

Large Numbers

Sometimes when dealing with very large numbers, the JS engine may mangle the number into scientific notation:

* def big = 123123123123
* string json = { num: '#(big)' }
* match json == '{"num":1.23123123123E11}'

This can be easily solved by using java.math.BigDecimal:

* def big = new java.math.BigDecimal(123123123123)
* string json = { num: '#(big)' }
* match json == '{"num":123123123123}'

doc

Karate has a built-in HTML templating engine that can be used to insert additional custom HTML into the test-reports. Here is an example:

* url 'https://jsonplaceholder.typicode.com/users'
* method get
* doc { read: 'users.html' }

Any Karate variable will be available to the template, which is users.html in this example.

<table class="table table-striped">
  <thead>
    <tr>
      <th>ID</th>
      <th>Name</th>
      <th>E-Mail</th>
    </tr>
  </thead>
  <tbody>
    <tr th:each="user: response">
      <td th:text="user.id"></td>
      <td th:text="user.name"></td>
      <td th:text="user.email"></td>
    </tr>
  </tbody>
</table>

You can see what the result looks like here.

Since templates can be loaded using the classpath: prefix, you can even re-use templates across your projects via Java JAR files.

Karate Expressions

Before we get to the HTTP keywords, it is worth doing a recap of the various 'shapes' that the right-hand-side of an assignment statement can take:

Core Keywords

They are url, path, request, method and status.

These are essential HTTP operations, they focus on setting one (un-named or 'key-less') value at a time and therefore don't need an = sign in the syntax.

url

Given url 'https://myhost.com/v1/cats'

Within a Scenario, a URL remains constant until you use the url keyword again, so this is a good place to set-up the 'non-changing' parts of your REST URL-s.

A URL can take expressions, so the approach below is legal. And yes, variables can come from global config.

Given url 'https://' + e2eHostName + '/v1/api'

JavaScript enthusiasts may prefer variable interpolation using backticks:

* url `https://${e2eHostName}/v1/api`

If you are trying to build dynamic URLs including query-string parameters in the form: http://myhost/some/path?foo=bar&search=true - please refer to the param keyword.

configure url

When you call other features, the url will be "reset". But if you want the url to persist, you can do this:

Feature:

Scenario:
* configure url = 'https://httpbin.org'
* path 'anything'
* method get
* call read('@called')

@ignore @called
Scenario:
* path 'anything'
* method get

Note how in the "called" Scenario you could omit the url. It is easy to change the url anytime by using the keyword. Note that you can use variables to set up the url any time you need to.

path

REST-style path parameters. Can be expressions that will be evaluated. Comma delimited values are supported which can be more convenient, and takes care of URL-encoding and appending '/' between path segments as needed.

Given path 'documents', documentId, 'download'

# or you can do the same on multiple lines if you wish
Given path 'documents'
And path documentId
And path 'download'

Note that the path 'resets' after any HTTP request is made but not the url. The Hello World is a great example of 'REST-ful' use of the url when the test focuses on a single REST 'resource'. Look at how the path did not need to be specified for the second HTTP get call since /cats is part of the url.

Important: If you attempt to build a URL in the form ?myparam=value by using path the ? will get encoded into %3F. Use either the param keyword, e.g.: * param myparam = 'value' or url: * url 'http://example.com/v1?myparam'

Because Karate strips trailing slashes if part of a path parameter, if you want to append a forward-slash to the end of the URL in the final HTTP request - make sure that the last path is a single '/'. For example, if your path has to be documents/ (and not just documents), use:

Given path 'documents', '/'

request

In-line JSON:

Given request { name: 'Billie', type: 'LOL' }

In-line XML:

And request <cat><name>Billie</name><type>Ceiling</type></cat>

From a file in the same package. Use the classpath: prefix to load from the classpath instead.

Given request read('my-json.json')

You could always use a variable:

And request myVariable

In most cases you won't need to set the Content-Type header as Karate will automatically do the right thing depending on the data-type of the request.

Defining the request is mandatory if you are using an HTTP method that expects a body such as post. If you really need to have an empty body, you can use an empty string as shown below, and you can force the right Content-Type header by using the header keyword.

Given request ''
And header Content-Type = 'text/html'

Sending a file as the entire binary request body is easy (note that multipart is different):

Given path 'upload'
And request read('my-image.jpg')
When method put
Then status 200

method

The HTTP verb - get, post, put, delete, patch, options, head, connect, trace.

Lower-case is fine.

When method post

It is worth internalizing that during test-execution, it is upon the method keyword that the actual HTTP request is issued. Which suggests that the step should be in the When form, for example: When method post. And steps that follow should logically be in the Then form. Also make sure that you complete the set up of things like url, param, header, configure etc. before you fire the method.

# set headers or params (if any) BEFORE the method step
Given header Accept = 'application/json'
When method get
# the step that immediately follows the above would typically be:
Then status 200

Although rarely needed, variable references or expressions are also supported:

* def putOrPost = (someVariable == 'dev' ? 'put' : 'post')
* method putOrPost

status

This is a shortcut to assert the HTTP response code.

Then status 200

And this assertion will cause the test to fail if the HTTP response code is something else.

See also responseStatus if you want to do some complex assertions against the HTTP status code.

Keywords that set key-value pairs

They are param, header, cookie, form field and multipart field.

The syntax will include a '=' sign between the key and the value. The key should not be within quotes.

To make dynamic data-driven testing easier, the following keywords also exist: params, headers, cookies and form fields. They use JSON to build the relevant parts of the HTTP request.

param

Setting query-string parameters:

Given param someKey = 'hello'
And param anotherKey = someVariable

The above would result in a URL like: http://myhost/mypath?someKey=hello&anotherKey=foo. Note that the ? and & will be automatically inserted.

Multi-value params are also supported:

* param myParam = ['foo', 'bar']

For convenience, a null value will be ignored. You can also use JSON to set multiple query-parameters in one-line using params and this is especially useful for dynamic data-driven testing.

header

You can use functions or expressions:

Given header Authorization = myAuthFunction()
And header transaction-id = 'test-' + myIdString

It is worth repeating that in most cases you won't need to set the Content-Type header as Karate will automatically do the right thing depending on the data-type of the request.

Because of how easy it is to set HTTP headers, Karate does not provide any special keywords for things like the Accept header. You simply do something like this:

Given path 'some/path'
And request { some: 'data' }
And header Accept = 'application/json'
When method post
Then status 200

A common need is to send the same header(s) for every request, and configure headers (with JSON) is how you can set this up once for all subsequent requests. And if you do this within a Background: section, it would apply to all Scenario: sections within the *.feature file.

* configure headers = { 'Content-Type': 'application/xml' }

Note that Content-Type had to be enclosed in quotes in the JSON above because the "-" (hyphen character) would cause problems otherwise. Also note that "; charset=UTF-8" would be appended to the Content-Type header that Karate sends by default, and in some rare cases, you may need to suppress this behavior completely. You can do so by setting the charset to null via the configure keyword:

* configure charset = null

If you need headers to be dynamically generated for each HTTP request, use a JavaScript function with configure headers instead of JSON.

Multi-value headers (though rarely used in the wild) are also supported:

* header myHeader = ['foo', 'bar']

Also look at the headers keyword which uses JSON and makes some kinds of dynamic data-driven testing easier.

cookie

Setting a cookie:

Given cookie foo = 'bar'

You also have the option of setting multiple cookies in one-step using the cookies keyword.

Note that any cookies returned in the HTTP response would be automatically set for any future requests. This mechanism works by calling configure cookies behind the scenes and if you need to stop auto-adding cookies for future requests, just do this:

* configure cookies = null

Also refer to the built-in variable responseCookies for how you can access and perform assertions on cookie data values.

form field

HTML form fields would be URL-encoded when the HTTP request is submitted (by the method step). You would typically use these to simulate a user sign-in and then grab a security token from the response.

Note that the Content-Type header will be automatically set to: application/x-www-form-urlencoded. You just need to do a normal POST (or GET).

For example:

Given path 'login'
And form field username = 'john'
And form field password = 'secret'
When method post
Then status 200
And def authToken = response.token

A good example of the use of form field for a typical sign-in flow is this OAuth 2 demo: oauth2.feature.

Multi-values are supported the way you would expect (e.g. for simulating check-boxes and multi-selects):

* form field selected = ['apple', 'orange']

You can also dynamically set multiple fields in one step using the form fields keyword.

multipart field

Use this for building multipart named (form) field requests. This is typically combined with multipart file as shown below.

Multiple fields can be set in one step using multipart fields.

multipart file

Given multipart file myFile = { read: 'test.pdf', filename: 'upload-name.pdf', contentType: 'application/pdf' }
And multipart field message = 'hello world'
When method post
Then status 200

It is important to note that myFile above is the "field name" within the multipart/form-data request payload. This roughly corresponds to a cURL argument of -F @myFile=test.pdf.

multipart file uploads can be tricky, and hard to get right. If you get stuck and ask a question on Stack Overflow, make sure you provide a cURL command that works - or else it would be very difficult for anyone to troubleshoot what you could be doing wrong. Also see this thread.

Also note that multipart file takes a JSON argument so that you can easily set the filename and the contentType (mime-type) in one step.

• read: the name of a file, and the classpath: prefix also is allowed. mandatory unless value is used, see below.
• value: alternative to read in rare cases where something like a JSON or XML file is being uploaded and you want to create it dynamically.
• filename: optional, if not specified there will be no filename attribute in Content-Disposition
• contentType: optional, will default to application/octet-stream

When 'multipart' content is involved, the Content-Type header of the HTTP request defaults to multipart/form-data. You can over-ride it by using the header keyword before the method step. Look at multipart entity for an example.

Also refer to this demo example for a working example of multipart file uploads: upload.feature.

You can also dynamically set multiple files in one step using multipart files.

multipart entity

This is technically not in the key-value form: multipart field name = 'foo', but logically belongs here in the documentation.

Use this for multipart content items that don't have field-names. Here below is an example that also demonstrates using the multipart/related content-type.

Given path 'v2', 'documents'
And multipart entity read('foo.json')
And multipart field image = read('bar.jpg')
And header Content-Type = 'multipart/related'
When method post 
Then status 201

Multi-Param Keywords

Keywords that set multiple key-value pairs in one step

params, headers, cookies, form fields, multipart fields and multipart files take a single JSON argument (which can be in-line or a variable reference), and this enables certain types of dynamic data-driven testing, especially because any JSON key with a null value will be ignored. Here is a good example in the demos: dynamic-params.feature

params

* params { searchBy: 'client', active: true, someList: [1, 2, 3] }

See also param.

headers

* def someData = { Authorization: 'sometoken', tx_id: '1234', extraTokens: ['abc', 'def'] }
* headers someData

See also header.

cookies

* cookies { someKey: 'someValue', foo: 'bar' }

See also cookie.

form fields

* def credentials = { username: '#(user.name)', password: 'secret', projects: ['one', 'two'] }
* form fields credentials

See also form field.

multipart fields

And multipart fields { message: 'hello world', json: { foo: 'bar' } }

See also multipart field.

multipart files

The single JSON argument needs to be in the form { field1: { read: 'file1.ext' }, field2: { read: 'file2.ext' } } where each nested JSON is in the form expected by multipart file

* def json = {}
* set json.myFile1 = { read: 'test1.pdf', filename: 'upload-name1.pdf', contentType: 'application/pdf' }
# if you have dynamic keys you can do this
* def key = 'myFile2'
* json[key] = { read: 'test2.pdf', filename: 'upload-name2.pdf', contentType: 'application/pdf' }
And multipart files json

For an example, refer: upload-multiple-files.feature.

SOAP

Since a SOAP request needs special handling, this is the only case where the method step is not used to actually fire the request to the server.

soap action

The name of the SOAP action specified is used as the 'SOAPAction' header. Here is an example which also demonstrates how you could assert for expected values in the response XML.

Given request read('soap-request.xml')
When soap action 'QueryUsageBalance'
Then status 200
And match response /Envelope/Body/QueryUsageBalanceResponse/Result/Error/Code == 'DAT_USAGE_1003'
And match response /Envelope/Body/QueryUsageBalanceResponse == read('expected-response.xml')

Refer to the demos for an example: soap.feature.

More examples are available that showcase various ways of parameter-izing and dynamically manipulating SOAP requests in a data-driven fashion. Karate is quite flexible, and provides multiple options for you to evolve patterns that fit your environment, as you can see here: xml.feature.

retry until

Karate has built-in support for re-trying an HTTP request until a certain condition has been met. The default setting for the max retry-attempts is 3 with a poll interval of 3000 milliseconds (3 seconds). If needed, this can be changed by using configure - any time during a test, or set globally via karate-config.js

* configure retry = { count: 10, interval: 5000 }

The retry keyword is designed to extend the existing method syntax (and should appear before a method step) like so:

Given url demoBaseUrl
And path 'greeting'
And retry until response.id > 3
When method get
Then status 200

Any JavaScript expression that uses any variable in scope can be placed after the "retry until" part. So you can refer to the response, responseStatus or even responseHeaders if needed. For example:

Given url demoBaseUrl
And path 'greeting'
And retry until responseStatus == 200 && response.id > 3
When method get

Note that it has to be a pure JavaScript expression - which means that match syntax such as contains will not work. But you can easily achieve any complex logic by using the JS API.

Refer to polling.feature for an example, and also see the alternative way to achieve polling.

configure

Managing Headers, SSL, Timeouts and HTTP Proxy

You can adjust configuration settings for the HTTP client used by Karate using this keyword. The syntax is similar to def but instead of a named variable, you update configuration. Here are the configuration keys supported:

Examples:

# pretty print the response payload
* configure logPrettyResponse = true

# enable ssl (and no certificate is required)
* configure ssl = true

# enable ssl and force the algorithm to TLSv1.2
* configure ssl = 'TLSv1.2'

# time-out if the response is not received within 10 seconds (after the connection is established)
* configure readTimeout = 10000

# set the uri of the http proxy server to use
* configure proxy = 'http://my.proxy.host:8080'

# proxy which needs authentication
* configure proxy = { uri: 'http://my.proxy.host:8080', username: 'john', password: 'secret' }

configure globally

If you need to set any of these "globally" you can easily do so using the karate object in karate-config.js - for e.g:

  karate.configure('ssl', true);
  karate.configure('readTimeout', 5000);

In rare cases where you need to add nested non-JSON data to the configure value, you have to play by the rules that apply within karate-config.js. Here is an example of performing a configure driver step in JavaScript:

  var LM = Java.type('com.mycompany.MyHttpLogModifier');
  var driverConfig = { type:'chromedriver', start: false, webDriverUrl:'https://user:[email protected]/wd/hub' };
  driverConfig.httpConfig = karate.toMap({ logModifier: LM.INSTANCE });
  karate.configure('driver', driverConfig);

Report Verbosity

By default, Karate will add logs to the report output so that HTTP requests and responses appear in-line in the HTML reports. There may be cases where you want to suppress this to make the reports "lighter" and easier to read.

The configure key here is report and it takes a JSON value. For example:

* configure report = { showLog: true, showAllSteps: false }

You can 'reset' default settings by using the following short-cut:

# reset to defaults
* configure report = true

Since you can use configure any time within a test, you have control over which requests or steps you want to show / hide. This can be convenient if a particular call results in a huge response payload.

The following short-cut is also supported which will disable all logs:

* configure report = false

@report=false

When you use a re-usable feature that has commonly used utilities, you may want to hide this completely from the HTML reports. The special tag @report=false can be used, and it can even be used only for a single Scenario:

@ignore @report=false
Feature:

Scenario:
# some re-usable steps

Log Masking

In cases where you want to "mask" values which are sensitive from a security point of view from the output files, logs and HTML reports, you can implement the HttpLogModifier and tell Karate to use it via the configure keyword. Here is an example of an implementation. For performance reasons, you can implement enableForUri() so that this "activates" only for some URL patterns.

Instantiating a Java class and using this in a test is easy (see example):

# if this was in karate-config.js, it would apply "globally"
* def LM = Java.type('demo.headers.DemoLogModifier')
* configure logModifier = new LM()

Or globally in karate-config.js

var LM = Java.type('demo.headers.DemoLogModifier');
karate.configure('logModifier', new LM());

Since karate-config.js is processed for every Scenario, you can use a singleton instead of calling new every time. Something like this:

var LM = Java.type('demo.headers.DemoLogModifier');
karate.configure('logModifier', LM.INSTANCE);

System Properties for SSL and HTTP proxy

For HTTPS / SSL, you can also specify a custom certificate or trust store by setting Java system properties. And similarly - for specifying the HTTP proxy.

X509 Certificate Authentication

Also referred to as "mutual auth" - if your API requires that clients present an X509 certificate for authentication, Karate supports this via JSON as the configure ssl value. The following parameters are supported:

Example:

# enable X509 certificate authentication with PKCS12 file 'certstore.pfx' and password 'certpassword'
* configure ssl = { keyStore: 'classpath:certstore.pfx', keyStorePassword: 'certpassword', keyStoreType: 'pkcs12' }

# trust all server certificates, in the feature file
* configure ssl = { trustAll: true }

// trust all server certificates, global configuration in 'karate-config.js'
karate.configure('ssl', { trustAll: true });

For end-to-end examples in the Karate demos, look at the files in this folder.

Payload Assertions

Prepare, Mutate, Assert.

Now it should be clear how Karate makes it easy to express JSON or XML. If you read from a file, the advantage is that multiple scripts can re-use the same data.

Once you have a JSON or XML object, Karate provides multiple ways to manipulate, extract or transform data. And you can easily assert that the data is as expected by comparing it with another JSON or XML object.

match

Payload Assertions / Smart Comparison

The match operation is smart because white-space does not matter, and the order of keys (or data elements) does not matter. Karate is even able to ignore fields you choose - which is very useful when you want to handle server-side dynamically generated fields such as UUID-s, time-stamps, security-tokens and the like.

The match syntax involves a double-equals sign '==' to represent a comparison (and not an assignment '=').

Since match and set go well together, they are both introduced in the examples in the section below.

set

Game, set and match - Karate !

JS for JSON

Before you consider the set keyword - note that for simple JSON update operations, you can use eval - especially useful when the path you are trying to mutate is dynamic. Since the eval keyword can be omitted when operating on variables using JavaScript, this leads to very concise code:

* def myJson = { a: '1' }
* myJson.b = 2
* match myJson == { a: '1', b: 2 }

Refer to eval for more / advanced examples.

Manipulating Data

Setting values on JSON documents is simple using the set keyword.

* def myJson = { foo: 'bar' }
* set myJson.foo = 'world'
* match myJson == { foo: 'world' }

# add new keys.  you can use pure JsonPath expressions (notice how this is different from the above)
* set myJson $.hey = 'ho'
* match myJson == { foo: 'world', hey: 'ho' }

# and even append to json arrays (or create them automatically)
* set myJson.zee[0] = 5
* match myJson == { foo: 'world', hey: 'ho', zee: [5] }

# omit the array index to append
* set myJson.zee[] = 6
* match myJson == { foo: 'world', hey: 'ho', zee: [5, 6] }

# nested json ? no problem
* set myJson.cat = { name: 'Billie' }
* match myJson == { foo: 'world', hey: 'ho', zee: [5, 6], cat: { name: 'Billie' } }

# and for match - the order of keys does not matter
* match myJson == { cat: { name: 'Billie' }, hey: 'ho', foo: 'world', zee: [5, 6] }

# you can ignore fields marked with '#ignore'
* match myJson == { cat: '#ignore', hey: 'ho', foo: 'world', zee: [5, 6] }

XML and XPath works just like you'd expect.

* def cat = <cat><name>Billie</name></cat>
* set cat /cat/name = 'Jean'
* match cat / == <cat><name>Jean</name></cat>

# you can even set whole fragments of xml
* def xml = <foo><bar>baz</bar></foo>
* set xml/foo/bar = <hello>world</hello>
* match xml == <foo><bar><hello>world</hello></bar></foo>

Refer to the section on XPath Functions for examples of advanced XPath usage.

match and variables

In case you were wondering, variables (and even expressions) are supported on the right-hand-side. So you can compare 2 JSON (or XML) payloads if you wanted to:

* def foo = { hello: 'world', baz: 'ban' }
* def bar = { baz: 'ban', hello: 'world' }
* match foo == bar

If you are wondering about the finer details of the match syntax, the Left-Hand-Side has to be either a

• variable name - e.g. foo
• a 'named' JsonPath or XPath expression - e.g. foo[0].bar or foo[*].bar
  • note that this cannot be "dynamic" (with in-line variables) so use an extra step if needed
• any valid function or method call - e.g. foo.bar() or foo.bar('hello').baz
• or anything wrapped in parentheses which will be evaluated as JavaScript - e.g. (foo + bar) or (42) - and in this case, variables can be used

And the right-hand-side can be any valid Karate expression. Refer to the section on JsonPath short-cuts for a deeper understanding of 'named' JsonPath expressions in Karate.

match != (not equals)

The 'not equals' operator != works as you would expect:

* def test = { foo: 'bar' }
* match test != { foo: 'baz' }

You typically will never need to use the != (not-equals) operator ! Use it sparingly, and only for string, number or simple payload comparisons.

set multiple

Karate has an elegant way to set multiple keys (via path expressions) in one step. For convenience, non-existent keys (or array elements) will be created automatically. You can find more JSON examples here: js-arrays.feature.

* def cat = { name: '' }

* set cat
  | path   | value |
  | name   | 'Bob' |
  | age    | 5     |

* match cat == { name: 'Bob', age: 5 }

One extra convenience for JSON is that if the variable itself (which was cat in the above example) does not exist, it will be created automatically. You can even create (or modify existing) JSON arrays by using multiple columns.

* set foo
  | path | 0     | 1     |
  | bar  | 'baz' | 'ban' |

* match foo == [{ bar: 'baz' }, { bar: 'ban' }]

If you have to set a bunch of deeply nested keys, you can move the parent path to the top, next to the set keyword and save a lot of typing ! Note that this is not supported for "arrays" like above, and you can have only one value column.

* set foo.bar
  | path   | value |
  | one    | 1     |
  | two[0] | 2     |
  | two[1] | 3     |

* match foo == { bar: { one: 1, two: [2, 3] } }

The same concept applies to XML and you can build complicated payloads from scratch in just a few, extremely readable lines. The value column can take expressions, even XML chunks. You can find more examples here: xml.feature.

* set search /acc:getAccountByPhoneNumber
  | path                        | value |
  | acc:phone/@foo              | 'bar' |
  | acc:phone/acc:number[1]     | 1234  |
  | acc:phone/acc:number[2]     | 5678  |     
  | acc:phoneNumberSearchOption | 'all' |

* match search ==
  """
  <acc:getAccountByPhoneNumber>
      <acc:phone foo="bar">
          <acc:number>1234</acc:number>
          <acc:number>5678</acc:number>
      </acc:phone>
      <acc:phoneNumberSearchOption>all</acc:phoneNumberSearchOption>        
  </acc:getAccountByPhoneNumber>
  """

remove

This is like the opposite of set if you need to remove keys or data elements from JSON or XML instances. You can even remove JSON array elements by index.

* def json = { foo: 'world', hey: 'ho', zee: [1, 2, 3] }
* remove json.hey
* match json == { foo: 'world', zee: [1, 2, 3] }
* remove json $.zee[1]
* match json == { foo: 'world', zee: [1, 3] }

remove works for XML elements as well:

* def xml = <foo><bar><hello>world</hello></bar></foo>
* remove xml/foo/bar/hello
* match xml == <foo><bar/></foo>
* remove xml /foo/bar
* match xml == <foo/>

Also take a look at how a special case of embedded-expressions can remove key-value pairs from a JSON (or XML) payload: Remove if Null.

See also delete, below.

delete

For JSON, you can also use the JS delete operator via eval, useful when the path you are trying to mutate is dynamic.

* def key = 'a'
* def foo = { a: 1 }
* eval delete foo[key]

As a convenience, you can omit the eval:

* delete foo[key]

Fuzzy Matching

Ignore or Validate

When expressing expected results (in JSON or XML) you can mark some fields to be ignored when the match (comparison) is performed. You can even use a regular-expression so that instead of checking for equality, Karate will just validate that the actual value conforms to the expected pattern.

This means that even when you have dynamic server-side generated values such as UUID-s and time-stamps appearing in the response, you can still assert that the full-payload matched in one step.

* def cat = { name: 'Billie', type: 'LOL', id: 'a9f7a56b-8d5c-455c-9d13-808461d17b91' }
* match cat == { name: '#ignore', type: '#regex [A-Z]{3}', id: '#uuid' }
# this will fail
# * match cat == { name: '#ignore', type: '#regex .{2}', id: '#uuid' }	

Note that regex escaping has to be done with a double back-slash - for e.g: '#regex a\\.dot' will match 'a.dot'

The supported markers are the following:

Note that #present and #notpresent only make sense when you are matching within a JSON or XML context or using a JsonPath or XPath on the left-hand-side.

* def json = { foo: 'bar' }
* match json == { foo: '#present' }
* match json.nope == '#notpresent'

The rest can also be used even in 'primitive' data matches like so:

* match foo == '#string'
# convenient (and recommended) way to check for array length
* match bar == '#[2]'

Optional Fields

If two cross-hatch # symbols are used as the prefix (for example: ##number), it means that the key is optional or that the value can be null.

* def foo = { bar: 'baz' }
* match foo == { bar: '#string', ban: '##string' }

Remove If Null

A very useful behavior when you combine the optional marker with an embedded expression is as follows: if the embedded expression evaluates to null - the JSON key (or XML element or attribute) will be deleted from the payload (the equivalent of remove).

* def data = { a: 'hello', b: null, c: null }
* def json = { foo: '#(data.a)', bar: '#(data.b)', baz: '##(data.c)' }
* match json == { foo: 'hello', bar: null }

If you are just trying to pre-define schema snippets to use in a fuzzy-match, you can use enclosed Javascript to suppress the default behavior of replacing placeholders. For example:

* def dogSchema = { id: '#string', color: '#string' }
# here we enclose in round-brackets to preserve the optional embedded expression
# so that it can be used later in a "match"
* def schema = ({ id: '#string', name: '#string', dog: '##(dogSchema)' })

* def response1 = { id: '123', name: 'foo' }
* match response1 == schema

And if you need to suppress placeholder substitution for read(), but still need a JSON snippet, you can do this. Note how we read as a string, but "cast" to JSON:

* json schema = karate.readAsString('schema.json')

If you want to use the triple-quote / multi-line way of defining JSON or if you have to use XML - you can use text and "cast" to JSON or XML as a second step - before using in a match:

* text schema =
"""
<root>
  <a>#string</a>
  <b>##(subSchema)</b>
</root>
"""
* xml schema = schema

#null and #notpresent

Karate's match is strict, and the case where a JSON key exists but has a null value (#null) is considered different from the case where the key is not present at all (#notpresent) in the payload.

But note that ##null can be used to represent a convention that many teams adopt, which is that keys with null values are stripped from the JSON payload. In other words, { a: 1, b: null } is considered 'equal' to { a: 1 } and { a: 1, b: '##null' } will match both cases.

These examples (all exact matches) can make things more clear:

* def foo = { }
* match foo == { a: '##null' }
* match foo == { a: '##notnull' }
* match foo == { a: '#notpresent' }
* match foo == { a: '#ignore' }

* def foo = { a: null }
* match foo == { a: '#null' }    
* match foo == { a: '##null' }
* match foo == { a: '#present' }
* match foo == { a: '#ignore' }

* def foo = { a: 1 }
* match foo == { a: '#notnull' }
* match foo == { a: '##notnull' }
* match foo == { a: '#present' }
* match foo == { a: '#ignore' }

Note that you can alternatively use JsonPath on the left-hand-side:

* def foo = { a: 1 }
* match foo.a == '#present'
* match foo.nope == '#notpresent'

But of course it is preferable to match whole objects in one step as far as possible.

'Self' Validation Expressions

The special 'predicate' marker #? EXPR in the table above is an interesting one. It is best explained via examples. Any valid JavaScript expression that evaluates to a Truthy or Falsy value is expected after the #?.

Observe how the value of the field being validated (or 'self') is injected into the 'underscore' expression variable: '_'

* def date = { month: 3 }
* match date == { month: '#? _ > 0 && _ < 13' }

What is even more interesting is that expressions can refer to variables:

* def date = { month: 3 }
* def min = 1
* def max = 12
* match date == { month: '#? _ >= min && _ <= max' }

And functions work as well ! You can imagine how you could evolve a nice set of utilities that validate all your domain objects.

* def date = { month: 3 }
* def isValidMonth = function(m) { return m >= 1 && m <= 12 }
* match date == { month: '#? isValidMonth(_)' }

Especially since strings can be easily coerced to numbers (and vice-versa) in Javascript, you can combine built-in validators with the self-validation 'predicate' form like this: '#number? _ > 0'

# given this invalid input (string instead of number)
* def date = { month: '3' }
# this will pass
* match date == { month: '#? _ > 0' }
# but this 'combined form' will fail, which is what we want
# * match date == { month: '#number? _ > 0' }

Referring to the JSON root

You can actually refer to any JsonPath on the document via $ and perform cross-field or conditional validations ! This example uses contains and the #? 'predicate' syntax, and situations where this comes in useful will be apparent when we discuss match each.

Given def temperature = { celsius: 100, fahrenheit: 212 }
Then match temperature == { celsius: '#number', fahrenheit: '#? _ == $.celsius * 1.8 + 32' }
# when validation logic is an 'equality' check, an embedded expression works better
Then match temperature contains { fahrenheit: '#($.celsius * 1.8 + 32)' }

match text or binary

# when the response is plain-text
Then match response == 'Health Check OK'
And match response != 'Error'

# when the response is binary (byte-array)
Then match responseBytes == read('test.pdf')

# incidentally, match and assert behave exactly the same way for strings
* def hello = 'Hello World!'
* match hello == 'Hello World!'
* assert hello == 'Hello World!'

Checking if a string is contained within another string is a very common need and match (name) contains works just like you'd expect:

* def hello = 'Hello World!'
* match hello contains 'World'
* match hello !contains 'blah'

For case-insensitive string comparisons, see how to create custom utilities or karate.lowerCase(). And for dealing with binary content - see bytes.

match header

Since asserting against header values in the response is a common task - match header has a special meaning. It short-cuts to the pre-defined variable responseHeaders and reduces some complexity - because strictly, HTTP headers are a 'multi-valued map' or a 'map of lists' - the Java-speak equivalent being Map<String, List<String>>. And since header names are case-insensitive - it ignores the case when finding the header to match.

# so after a http request
Then match header Content-Type == 'application/json'
# 'contains' works as well
Then match header Content-Type contains 'application'

Note the extra convenience where you don't have to enclose the LHS key in quotes.

You can always directly access the variable called responseHeaders if you wanted to do more checks, but you typically won't need to.

match and XML

All the fuzzy matching markers will work in XML as well. Here are some examples:

  * def xml = <root><hello>world</hello><foo>bar</foo></root>
  * match xml == <root><hello>world</hello><foo>#ignore</foo></root>
  * def xml = <root><hello foo="bar">world</hello></root>
  * match xml == <root><hello foo="#ignore">world</hello></root>

Refer to this file for a comprehensive set of XML examples: xml.feature.

Matching Sub-Sets of JSON Keys and Arrays

match contains

JSON Keys

In some cases where the response JSON is wildly dynamic, you may want to only check for the existence of some keys. And match (name) contains is how you can do so:

* def foo = { bar: 1, baz: 'hello', ban: 'world' }

* match foo contains { bar: 1 }
* match foo contains { baz: 'hello' }
* match foo contains { bar:1, baz: 'hello' }
# this will fail
# * match foo == { bar:1, baz: 'hello' }

Note that match contains will not "recurse" any nested JSON chunks so use match contains deep instead.

Also note that match contains any is possible for JSON objects as well as JSON arrays.

(not) !contains

It is sometimes useful to be able to check if a key-value-pair does not exist. This is possible by prefixing contains with a ! (with no space in between).

* def foo = { bar: 1, baz: 'hello', ban: 'world' }
* match foo !contains { bar: 2 }
* match foo !contains { huh: '#notnull' }

Here's a reminder that the #notpresent marker can be mixed into an equality match (==) to assert that some keys exist and at the same time ensure that some keys do not exist:

* def foo = { a: 1 }
* match foo == { a: '#number', b: '#notpresent' }

# if b can be present (optional) but should always be null
* match foo == { a: '#number', b: '##null' }

The ! (not) operator is especially useful for contains and JSON arrays.

* def foo = [1, 2, 3]
* match foo !contains 4
* match foo !contains [5, 6]

JSON Arrays

This is a good time to deep-dive into JsonPath, which is perfect for slicing and dicing JSON into manageable chunks. It is worth taking a few minutes to go through the documentation and examples here: JsonPath Examples.

Here are some example assertions performed while scraping a list of child elements out of the JSON below. Observe how you can match the result of a JsonPath expression with your expected data.

Given def cat = 
  """
  {
    name: 'Billie',
    kittens: [
      { id: 23, name: 'Bob' },
      { id: 42, name: 'Wild' }
    ]
  }
  """
# normal 'equality' match. note the wildcard '*' in the JsonPath (returns an array)
Then match cat.kittens[*].id == [23, 42]

# when inspecting a json array, 'contains' just checks if the expected items exist
# and the size and order of the actual array does not matter
Then match cat.kittens[*].id contains 23
Then match cat.kittens[*].id contains [42]
Then match cat.kittens[*].id contains [23, 42]
Then match cat.kittens[*].id contains [42, 23]

# the .. operator is great because it matches nodes at any depth in the JSON "tree"
Then match cat..name == ['Billie', 'Bob', 'Wild']

# and yes, you can assert against nested objects within JSON arrays !
Then match cat.kittens contains [{ id: 42, name: 'Wild' }, { id: 23, name: 'Bob' }]

# ... and even ignore fields at the same time !
Then match cat.kittens contains { id: 42, name: '#string' }

It is worth mentioning that to do the equivalent of the last line in Java, you would typically have to traverse 2 Java Objects, one of which is within a list, and you would have to check for nulls as well.

When you use Karate, all your data assertions can be done in pure JSON and without needing a thick forest of companion Java objects. And when you read your JSON objects from (re-usable) files, even complex response payload assertions can be accomplished in just a single line of Karate-script.

Refer to this case study for how dramatic the reduction of lines of code can be.

match contains only

For those cases where you need to assert that all array elements are present but in any order you can do this:

* def data = { foo: [1, 2, 3] }
* match data.foo contains 1
* match data.foo contains [2]
* match data.foo contains [3, 2]
* match data.foo contains only [3, 2, 1]
* match data.foo contains only [2, 3, 1]
# this will fail
# * match data.foo contains only [2, 3]

match contains any

To assert that any of the given array elements are present.

* def data = { foo: [1, 2, 3] }
* match data.foo contains any [9, 2, 8]

And this happens to work as expected for JSON object keys as well:

* def data = { a: 1, b: 'x' }
* match data contains any { b: 'x', c: true }

match contains deep

This modifies the behavior of match contains so that nested lists or objects are processed for a "deep contains" match instead of a "deep equals" one which is the default. This is convenient for complex nested payloads where you are sure that you only want to check for some values in the various "trees" of data.

Here is an example:

Scenario: recurse nested json
  * def original = { a: 1, b: 2, c: 3, d: { a: 1, b: 2 } }
  * def expected = { a: 1, c: 3, d: { b: 2 } }
  * match original contains deep expected

Scenario: recurse nested array
  * def original = { a: 1, arr: [ { b: 2, c: 3 }, { b: 3, c: 4 } ] }
  * def expected = { a: 1, arr: [ { b: 2 }, { c: 4 } ] }
  * match original contains deep expected

the NOT operator e.g. !contains deep is not yet supported, please contribute code if you can.

match contains only deep

This is exactly like match == but the order of arrays does not matter. All arrays no matter the "depth" will be checked in this way.

* def response = { foo: [ 'a', 'b' ] }
* match response contains only deep { foo: [ 'b', 'a' ] }

Validate every element in a JSON array

match each

The match keyword can be made to iterate over all elements in a JSON array using the each modifier. Here's how it works:

* def data = { foo: [{ bar: 1, baz: 'a' }, { bar: 2, baz: 'b' }, { bar: 3, baz: 'c' }]}

* match each data.foo == { bar: '#number', baz: '#string' }

# and you can use 'contains' the way you'd expect
* match each data.foo contains { bar: '#number' }
* match each data.foo contains { bar: '#? _ != 4' }

# some more examples of validation macros
* match each data.foo contains { baz: "#? _ != 'z'" }
* def isAbc = function(x) { return x == 'a' || x == 'b' || x == 'c' }
* match each data.foo contains { baz: '#? isAbc(_)' }

# this is also possible, see the subtle difference from the above
* def isXabc = function(x) { return x.baz == 'a' || x.baz == 'b' || x.baz == 'c' }
* match each data.foo == '#? isXabc(_)'

Here is a contrived example that uses match each, contains and the #? 'predicate' marker to validate that the value of totalPrice is always equal to the roomPrice of the first item in the roomInformation array.

Given def json =
  """
  {
    "hotels": [
      { "roomInformation": [{ "roomPrice": 618.4 }], "totalPrice": 618.4  },
      { "roomInformation": [{ "roomPrice": 679.79}], "totalPrice": 679.79 }
    ]
  }
  """
Then match each json.hotels contains { totalPrice: '#? _ == _$.roomInformation[0].roomPrice' }
# when validation logic is an 'equality' check, an embedded expression works better
Then match each json.hotels contains { totalPrice: '#(_$.roomInformation[0].roomPrice)' }

Referring to self

While $ always refers to the JSON 'root', note the use of _$ above to represent the 'current' node of a match each iteration. Here is a recap of symbols that can be used in JSON embedded expressions:

There is a shortcut for match each explained in the next section that can be quite useful, especially for 'in-line' schema-like validations.

match each contains deep

match each can be combined with contains deep so that for each JSON object a “deep contains” match is performed within nested lists or objects.

This is useful for testing payloads with JSON arrays whose members have a few essential keys that you wish to validate.

  Given def response = 
  """
  [
    {
      "a": 1,
      "arr": [
          {
              "b": 2,
              "c": 3
          }
      ]
    },
    {
      "a": 1,
      "arr": [
          {
              "b": 2,
              "c": 3
          },
          {
              "b": 4,
              "c": 5
          }
      ]
    }
  ]
  """
  Then match each response contains deep { a: 1, arr: [ { b: 2 } ] }

Schema Validation

Karate provides a far more simpler and more powerful way than JSON-schema to validate the structure of a given payload. You can even mix domain and conditional validations and perform all assertions in a single step.

But first, a special short-cut for array validation needs to be introduced:

* def foo = ['bar', 'baz']

# should be an array
* match foo == '#[]'

# should be an array of size 2
* match foo == '#[2]'

# should be an array of strings with size 2
* match foo == '#[2] #string'

# each array element should have a 'length' property with value 3
* match foo == '#[]? _.length == 3'

# should be an array of strings each of length 3
* match foo == '#[] #string? _.length == 3'

# should be null or an array of strings
* match foo == '##[] #string'

This 'in-line' short-cut for validating JSON arrays is similar to how match each works. So now, complex payloads (that include arrays) can easily be validated in one step by combining validation markers like so:

* def oddSchema = { price: '#string', status: '#? _ < 3', ck: '##number', name: '#regex[0-9X]' }
* def isValidTime = read('time-validator.js')
When method get
Then match response ==
  """
  { 
    id: '#regex[0-9]+',
    count: '#number',
    odd: '#(oddSchema)',
    data: { 
      countryId: '#number', 
      countryName: '#string', 
      leagueName: '##string', 
      status: '#number? _ >= 0', 
      sportName: '#string',
      time: '#? isValidTime(_)'
    },
    odds: '#[] oddSchema'  
  }
  """

Especially note the re-use of the oddSchema both as an embedded-expression and as an array validation (on the last line).

And you can perform conditional / cross-field validations and even business-logic validations at the same time.

# optional (can be null) and if present should be an array of size greater than zero
* match $.odds == '##[_ > 0]'

# should be an array of size equal to $.count
* match $.odds == '#[$.count]'

# use a predicate function to validate each array element
* def isValidOdd = function(o){ return o.name.length == 1 }
* match $.odds == '#[]? isValidOdd(_)'

Refer to this for the complete example: schema-like.feature

And there is another example in the karate-demos: schema.feature where you can compare Karate's approach with an actual JSON-schema example. You can also find a nice visual comparison and explanation here.

contains short-cuts

Especially when payloads are complex (or highly dynamic), it may be more practical to use contains semantics. Karate has the following short-cut symbols designed to be mixed into embedded expressions:

For completeness, == and != also belong in the above list.

Here'a table of the alternative 'in-line' forms compared with the 'standard' form. Note that all the short-cut forms on the right-side of the table resolve to 'equality' (==) matches, which enables them to be 'in-lined' into a full (single-step) payload match, using embedded expressions.

A very useful capability is to be able to check that an array contains an object that contains the provided sub-set of keys instead of having to specify the complete JSON - which can get really cumbersome for large objects. This turns out to be very useful in practice, and this particular match jsonArray contains '#(^partialObject)' form has no 'in-line' equivalent (see the third-from-last row above).

The last row in the table is a little different from the rest, and this short-cut form is the recommended way to validate the length of a JSON array. As a rule of thumb, prefer match over assert, because match failure messages are more detailed and descriptive.

In real-life tests, these are very useful when the order of items in arrays returned from the server are not guaranteed. You can easily assert that all expected elements are present, even in nested parts of your JSON - while doing a match on the full payload.

* def cat = 
  """
  {
    name: 'Billie',
    kittens: [
      { id: 23, name: 'Bob' },
      { id: 42, name: 'Wild' }
    ]
  }
  """
* def expected = [{ id: 42, name: 'Wild' }, { id: 23, name: 'Bob' }]
* match cat == { name: 'Billie', kittens: '#(^^expected)' }

There's a lot going on in the last line above ! It validates the entire payload in one step and checks if the kittens array contains all the expected items but in any order.

get

By now, it should be clear that JsonPath can be very useful for extracting JSON 'trees' out of a given object. The get keyword allows you to save the results of a JsonPath expression for later use - which is especially useful for dynamic data-driven testing.

* def cat = 
  """
  {
    name: 'Billie',
    kittens: [
      { id: 23, name: 'Bob' },
      { id: 42, name: 'Wild' }
    ]
  }
  """
* def kitnums = get cat.kittens[*].id
* match kitnums == [23, 42]
* def kitnames = get cat $.kittens[*].name
* match kitnames == ['Bob', 'Wild']

get short-cut

The 'short cut' $variableName form is also supported. Refer to JsonPath short-cuts for a detailed explanation. So the above could be re-written as follows:

* def kitnums = $cat.kittens[*].id
* match kitnums == [23, 42]
* def kitnames = $cat.kittens[*].name
* match kitnames == ['Bob', 'Wild']

It is worth repeating that the above can be condensed into 2 lines. Note that since only JsonPath is expected on the left-hand-side of the == sign of a match statement, you don't need to prefix the variable reference with $:

* match cat.kittens[*].id == [23, 42]
* match cat.kittens[*].name == ['Bob', 'Wild']

# if you prefer using 'pure' JsonPath, you can do this
* match cat $.kittens[*].id == [23, 42]
* match cat $.kittens[*].name == ['Bob', 'Wild']

get plus index

A convenience that the get syntax supports (but not the $ short-cut form) is to return a single element if the right-hand-side evaluates to a list-like result (e.g. a JSON array). This is useful because the moment you use a wildcard [*] or search filter in JsonPath (see the next section), you get an array back - even though typically you would only be interested in the first item.

* def actual = 23

# so instead of this
* def kitnums = get cat.kittens[*].id
* match actual == kitnums[0]

# you can do this in one line
* match actual == get[0] cat.kittens[*].id

JsonPath filters

JsonPath filter expressions are very useful for extracting elements that meet some filter criteria out of arrays.

* def cat = 
  """
  {
    name: 'Billie',
    kittens: [
      { id: 23, name: 'Bob' },
      { id: 42, name: 'Wild' }
    ]
  }
  """
# find single kitten where id == 23
* def bob = get[0] cat.kittens[?(@.id==23)]
* match bob.name == 'Bob'

# using the karate object if the expression is dynamic
* def temp = karate.jsonPath(cat, "$.kittens[?(@.name=='" + bob.name + "')]")
* match temp[0] == bob

# or alternatively
* def temp = karate.jsonPath(cat, "$.kittens[?(@.name=='" + bob.name + "')]")[0]
* match temp == bob

You usually won't need this, but the second-last line above shows how the karate object can be used to evaluate JsonPath if the filter expression depends on a variable. If you find yourself struggling to write dynamic JsonPath filters, look at karate.filter() as an alternative, described just below.

JSON Transforms

Karate supports the following functional-style operations via the JS API - karate.map(), karate.filter() and karate.forEach(). They can be very useful in some situations. A good example is when you have the expected data available as ready-made JSON but it is in a different "shape" from the actual data or HTTP response. There is also a karate.mapWithKey() for a common need - which is to convert an array of primitives into an array of objects, which is the form that data driven features expect.

The Graal JS engine that Karate uses supports the full ES6 spec, which means that JSON variables are first-class JS objects, and arrays can be directly looped over or manipulated using map(), filter() and forEach(). And JS "arrow functions" are supported, which makes code much more concise.

A few more useful "transforms" are to select a sub-set of key-value pairs using karate.filterKeys(), merging 2 or more JSON-s using karate.merge() and combining 2 or more arrays (or objects) into a single array using karate.append(). And karate.appendTo() is for updating an existing variable (the equivalent of array.push() in JavaScript), which is especially useful in the body of a karate.forEach().

You can also sort arrays of arbitrary JSON using karate.sort(). Simple arrays of strings or numbers can be stripped of duplicates using karate.distinct(). All JS "native" array operations can be used, such as someName.reverse().

Note that a single JS function is sufficient to transform a given JSON object into a completely new one, and you can use complex conditional logic if needed.

Scenario: karate map operation
    * def fun = function(x){ return x * x }
    * def list = [1, 2, 3]
    * def res = karate.map(list, fun)
    * match res == [1, 4, 9]

Scenario: js style map operation
    * def list = [1, 2, 3]
    * def res = list.map(list, x => x * x)
    * match res == [1, 4, 9]

Scenario: convert an array into a different shape
    * def before = [{ foo: 1 }, { foo: 2 }, { foo: 3 }]
    * def fun = function(x){ return { bar: x.foo } }
    * def after = karate.map(before, fun)
    * match after == [{ bar: 1 }, { bar: 2 }, { bar: 3 }]

Scenario: convert array of primitives into array of objects
    * def list = [ 'Bob', 'Wild', 'Nyan' ]
    * def data = karate.mapWithKey(list, 'name')
    * match data == [{ name: 'Bob' }, { name: 'Wild' }, { name: 'Nyan' }]

Scenario: karate filter operation
    * def fun = function(x){ return x % 2 == 0 }
    * def list = [1, 2, 3, 4]
    * def res = karate.filter(list, fun)
    * match res == [2, 4]

Scenario: js style filter operation
    * def list = [1, 2, 3, 4]
    * def res = list.filter(list, x => x % 2 == 0)
    * match res == [2, 4]    

Scenario: karate.forEach() works even on object key-values, not just arrays
    * def keys = []
    * def vals = []
    * def idxs = []
    * def fun = 
    """
    function(x, y, i) { 
      karate.appendTo(keys, x); 
      karate.appendTo(vals, y); 
      karate.appendTo(idxs, i); 
    }
    """
    * def map = { a: 2, b: 4, c: 6 }
    * karate.forEach(map, fun)
    * match keys == ['a', 'b', 'c']
    * match vals == [2, 4, 6]
    * match idxs == [0, 1, 2]

Scenario: filterKeys
    * def schema = { a: '#string', b: '#number', c: '#boolean' }
    * def response = { a: 'x', c: true }
    # very useful for validating a response against a schema "super-set"
    * match response == karate.filterKeys(schema, response)
    * match karate.filterKeys(response, 'b', 'c') == { c: true }
    * match karate.filterKeys(response, ['a', 'b']) == { a: 'x' }

Scenario: merge
    * def foo = { a: 1 }
    * def bar = karate.merge(foo, { b: 2 })
    * match bar == { a: 1, b: 2 }

Scenario: append
    * def foo = [{ a: 1 }]
    * def bar = karate.append(foo, { b: 2 })
    * match bar == [{ a: 1 }, { b: 2 }]

Scenario: sort
    * def foo = [{a: { b: 3 }}, {a: { b: 1 }}, {a: { b: 2 }}]
    * def fun = function(x){ return x.a.b }
    * def bar = karate.sort(foo, fun)
    * match bar == [{a: { b: 1 }}, {a: { b: 2 }}, {a: { b: 3 }}]
    * match bar.reverse() == [{a: { b: 3 }}, {a: { b: 2 }}, {a: { b: 1 }}]

Loops

Given the examples above, it has to be said that a best practice with Karate is to avoid JavaScript for loops as far as possible. A common requirement is to build an array with n elements or do something n times where n is an integer (that could even be a variable reference). This is easily achieved with the karate.repeat() API:

* def fun = function(i){ return i * 2 }
* def foo = karate.repeat(5, fun)
* match foo == [0, 2, 4, 6, 8]

* def foo = []
* def fun = function(i){ karate.appendTo(foo, i) }
* karate.repeat(5, fun)
* match foo == [0, 1, 2, 3, 4]

# generate test data easily
* def fun = function(i){ return { name: 'User ' + (i + 1) } }
* def foo = karate.repeat(3, fun)
* match foo == [{ name: 'User 1' }, { name: 'User 2' }, { name: 'User 3' }]

# generate a range of numbers as a json array
* def foo = karate.range(4, 9)
* match foo == [4, 5, 6, 7, 8, 9]

And there's also karate.range() which can be useful to generate test-data.

Don't forget that Karate's data-driven testing capabilities can loop over arrays of JSON objects automatically.

XPath Functions

When handling XML, you sometimes need to call XPath functions, for example to get the count of a node-set. Any valid XPath expression is allowed on the left-hand-side of a match statement.

* def myXml =
  """
  <records>
    <record index="1">a</record>
    <record index="2">b</record>
    <record index="3" foo="bar">c</record>
  </records>
  """

* match foo count(/records//record) == 3
* match foo //record[@index=2] == 'b'
* match foo //record[@foo='bar'] == 'c'

Advanced XPath

Some XPath expressions return a list of nodes (instead of a single node). But since you can express a list of data-elements as a JSON array - even these XPath expressions can be used in match statements.

* def teachers = 
  """
  <teachers>
    <teacher department="science">
      <subject>math</subject>
      <subject>physics</subject>
    </teacher>
    <teacher department="arts">
      <subject>political education</subject>
      <subject>english</subject>
    </teacher>
  </teachers>
  """
* match teachers //teacher[@department='science']/subject == ['math', 'physics']

If your XPath is dynamic and has to be formed 'on the fly' perhaps by using some variable derived from previous steps, you can use the karate.xmlPath() helper:

* def xml = <query><name><foo>bar</foo></name></query>
* def elementName = 'name'
* def name = karate.xmlPath(xml, '/query/' + elementName + '/foo')
* match name == 'bar'
* def queryName = karate.xmlPath(xml, '/query/' + elementName)
* match queryName == <name><foo>bar</foo></name>

You can refer to this file (which is part of the Karate test-suite) for more XML examples: xml-and-xpath.feature

Special Variables

These are 'built-in' variables, there are only a few and all of them give you access to the HTTP response.

response

After every HTTP call this variable is set with the response body, and is available until the next HTTP request over-writes it. You can easily assign the whole response (or just parts of it using Json-Path or XPath) to a variable, and use it in later steps.

The response is automatically available as a JSON, XML or String object depending on what the response contents are.

As a short-cut, when running JsonPath expressions - $ represents the response. This has the advantage that you can use pure JsonPath and be more concise. For example:

# the three lines below are equivalent
Then match response $ == { name: 'Billie' }
Then match response == { name: 'Billie' }
Then match $ == { name: 'Billie' }

# the three lines below are equivalent
Then match response.name == 'Billie'
Then match response $.name == 'Billie'
Then match $.name == 'Billie'

And similarly for XML and XPath, '/' represents the response

# the four lines below are equivalent
Then match response / == <cat><name>Billie</name></cat>
Then match response/ == <cat><name>Billie</name></cat>
Then match response == <cat><name>Billie</name></cat>
Then match / == <cat><name>Billie</name></cat> 

# the three lines below are equivalent
Then match response /cat/name == 'Billie'
Then match response/cat/name == 'Billie'
Then match /cat/name == 'Billie'

JsonPath short-cuts

The $varName form is used on the right-hand-side of Karate expressions and is slightly different from pure JsonPath expressions which always begin with $. or $[. Here is a summary of what the different 'shapes' mean in Karate:

There is no need to prefix variable names with $ on the left-hand-side of match statements because it is implied. You can if you want to, but since only JsonPath (on variables) is allowed here, Karate ignores the $ and looks only at the variable name. None of the examples in the documentation use the $varName form on the LHS, and this is the recommended best-practice.

responseBytes

This will always hold the contents of the response as a byte-array. This is rarely used, unless you are expecting binary content returned by the server. The match keyword will work as you expect. Here is an example: binary.feature.

responseCookies

The responseCookies variable is set upon any HTTP response and is a map-like (or JSON-like) object. It can be easily inspected or used in expressions.

* assert responseCookies['my.key'].value == 'someValue'

# karate's unified data handling means that even 'match' works
* match responseCookies contains { time: '#notnull' }

# ... which means that checking if a cookie does NOT exist is a piece of cake
* match responseCookies !contains { blah: '#notnull' }

# save a response cookie for later use
* def time = responseCookies.time.value

As a convenience, cookies from the previous response are collected and passed as-is as part of the next HTTP request. This is what is normally expected and simulates a web-browser - which makes it easy to script things like HTML-form based authentication into test-flows. Refer to the documentation for cookie for details and how you can disable this if need be.

Each item within responseCookies is itself a 'map-like' object. Typically you would examine the value property as in the example above, but domain and path are also available.

responseHeaders

See also match header which is what you would normally need.

But if you need to use values in the response headers - they will be in a variable named responseHeaders. Note that it is a 'map of lists' so you will need to do things like this:

* def contentType = responseHeaders['Content-Type'][0]

And just as in the responseCookies example above, you can use match to run complex validations on the responseHeaders.

Finally, using karate.response.header(name) can be simpler to just get a header value string by name, and it will ignore-case for the name passed as the argument:

* match karate.response.header('content-type') == 'application/json'

responseStatus

You would normally only need to use the status keyword. But if you really need to use the HTTP response code in an expression or save it for later, you can get it as an integer:

* def uploadStatusCode = responseStatus

# check if the response status is either of two values
Then assert responseStatus == 200 || responseStatus == 204

Note that match can give you some extra readable options:

* match [200, 201, 204] contains responseStatus

# this may be sufficient to check a range of values
* assert responseStatus >= 200
* assert responseStatus < 300

# but using karate.range() you can even do this !
* match karate.range(200, 299) contains responseStatus

responseTime

The response time (in milliseconds) for the current response would be available in a variable called responseTime. You can use this to assert that it was returned within the expected time like so:

When method post
Then status 201
And assert responseTime < 1000

responseType

Karate will attempt to parse the raw HTTP response body as JSON or XML and make it available as the response value. If parsing fails, Karate will log a warning and the value of response will then be a plain string. You can still perform string comparisons such as a match contains and look for error messages etc. In rare cases, you may want to check what the "type" of the response is and it can be one of 3 different values: json, xml and string.

So if you really wanted to assert that the HTTP response body is well-formed JSON or XML you can do this:

When method post
Then status 201
And match responseType == 'json'

requestTimeStamp

Very rarely used - but you can get the Java system-time (for the current response) at the point when the HTTP request was initiated (the value of System.currentTimeMillis()) which can be used for detailed logging or custom framework / stats calculations.

HTTP Header Manipulation

configure headers

Custom header manipulation for every HTTP request is something that Karate makes very easy and pluggable. For every HTTP request made from Karate, the internal flow is as follows:

• did we configure the value of headers ?
• if so, is the configured value a JavaScript function ?
  • if so, a call is made to that function.
  • did the function invocation return a map-like (or JSON) object ?
    • all the key-value pairs are added to the HTTP headers.
• or is the configured value a JSON object ?
  • all the key-value pairs are added to the HTTP headers.

This makes setting up of complex authentication schemes for your test-flows really easy. It typically ends up being a one-liner that appears in the Background section at the start of your test-scripts. You can re-use the function you create across your whole project.

Here is an example JavaScript function that uses some variables in the context (which have been possibly set as the result of a sign-in) to build the Authorization header. Note how even calls to Java code can be made if needed.

In the example below, note the use of the karate.get() helper for getting the value of a dynamic variable (which was not set at the time this JS function was declared). This is preferred because it takes care of situations such as if the value is undefined in JavaScript. In rare cases you may need to set a variable from this routine, and a good example is to make the generated UUID "visible" to the currently executing script or feature. You can easily do this via karate.set('someVarName', value).

function fn() {
  var uuid = '' + java.util.UUID.randomUUID(); // convert to string
  var out = { // so now the txid_header would be a unique uuid for each request
    txid_header: uuid,
    ip_header: '123.45.67.89', // hard coded here, but also can be as dynamic as you want   
  };
  var authString = '';
  var authToken = karate.get('authToken'); // use the 'karate' helper to do a 'safe' get of a 'dynamic' variable
  if (authToken) { // and if 'authToken' is not null ... 
    authString = ',auth_type=MyAuthScheme'
        + ',auth_key=' + authToken.key
        + ',auth_user=' + authToken.userId
        + ',auth_project=' + authToken.projectId;
  }
  // the 'appId' variable here is expected to have been set via karate-config.js (bootstrap init) and will never change
  out['Authorization'] = 'My_Auth app_id=' + appId + authString;
  return out;
}

Assuming the above code is in a file called my-headers.js, the next section on calling other feature files shows how it looks like in action at the beginning of a test script.

Notice how once the authToken variable is initialized, it is used by the above function to generate headers for every HTTP call made as part of the test flow.

If a few steps in your flow need to temporarily change (or completely bypass) the currently-set header-manipulation scheme, just update configure headers to a new value (or set it to null) in the middle of a script. Then use the header keyword to do a custom 'over-ride' if needed.

The karate-demo has an example showing various ways to configure or set headers: headers.feature

The karate object

A JavaScript function or Karate expression at runtime has access to a utility object in a variable named: karate. This provides the following methods:

Code Reuse / Common Routines

call

In any complex testing endeavor, you would find yourself needing 'common' code that needs to be re-used across multiple test scripts. A typical need would be to perform a 'sign in', or create a fresh user as a pre-requisite for the scenarios being tested.

There are two types of code that can be call-ed. *.feature files and JavaScript functions.

Calling other *.feature files

When you have a sequence of HTTP calls that need to be repeated for multiple test scripts, Karate allows you to treat a *.feature file as a re-usable unit. You can also pass parameters into the *.feature file being called, and extract variables out of the invocation result.

Here is an example of using the call keyword to invoke another feature file, loaded using the read function:

If you find this hard to understand at first, try looking at this set of examples.

Feature: which makes a 'call' to another re-usable feature

Background:
  * configure headers = read('classpath:my-headers.js')
  * def signIn = call read('classpath:my-signin.feature') { username: 'john', password: 'secret' }
  * def authToken = signIn.authToken

Scenario: some scenario
  # main test steps

Note that def can be used to assign a feature to a variable. For example look at how "creator" has been defined in the Background in this example, and used later in a call statement. This is very close to how "custom keywords" work in other frameworks. See this other example for more ideas: dsl.feature.

The contents of my-signin.feature are shown below. A few points to note:

• Karate creates a new 'context' for the feature file being invoked but passes along all variables and configuration. This means that all your config variables and configure settings would be available to use, for example loginUrlBase in the example below.
• When you use def in the 'called' feature, it will not over-write variables in the 'calling' feature (unless you explicitly choose to use shared scope). But note that JSON, XML, Map-like or List-like variables are 'passed by reference' which means that 'called' feature steps can update or 'mutate' them using the set keyword. Use the copy keyword to 'clone' a JSON or XML payload if needed, and refer to this example for more details: copy.feature.
• You can add (or over-ride) variables by passing a call 'argument' as shown above. Only one JSON argument is allowed, but this does not limit you in any way as you can use any complex JSON structure. You can even initialize the JSON in a separate step and pass it by name, especially if it is complex. Observe how using JSON for parameter-passing makes things super-readable. In the 'called' feature, the argument can also be accessed using the built-in variable: __arg.
• All variables that were defined (using def) in the 'called' script would be returned as 'keys' within a JSON-like object. Note that this includes 'built-in' variables, which means that things like the last value of response would also be present. In the example above you can see that the JSON 'envelope' returned - is assigned to the variable named signIn. And then getting hold of any data that was generated by the 'called' script is as simple as accessing it by name, for example signIn.authToken as shown above. This design has the following advantages:
  • 'called' Karate scripts don't need to use any special keywords to 'return' data and can behave like 'normal' Karate tests in 'stand-alone' mode if needed
  • the data 'return' mechanism is 'safe', there is no danger of the 'called' script over-writing any variables in the 'calling' (or parent) script (unless you use shared scope)
  • the need to explicitly 'unpack' variables by name from the returned 'envelope' keeps things readable and maintainable in the 'caller' script

Note that only variables and configuration settings will be passed. You can't do things such as * url 'http://foo.bar' and expect the URL to be set in the "called" feature. Use a variable in the "called" feature instead, for e.g. * url myUrl or take a look at configure url.

Feature: here are the contents of 'my-signin.feature'

Scenario:
  Given url loginUrlBase
  And request { userId: '#(username)', userPass: '#(password)' }
  When method post
  Then status 200
  And def authToken = response

  # second HTTP call, to get a list of 'projects'
  Given path 'users', authToken.userId, 'projects'
  When method get
  Then status 200
  # logic to 'choose' first project
  And set authToken.projectId = response.projects[0].projectId;

The above example actually makes two HTTP requests - the first is a standard 'sign-in' POST and then (for illustrative purposes) another HTTP call (a GET) is made for retrieving a list of projects for the signed-in user, and the first one is 'selected' and added to the returned 'auth token' JSON object.

So you get the picture, any kind of complicated 'sign-in' flow can be scripted and re-used.

If the second HTTP call above expects headers to be set by my-headers.js - which in turn depends on the authToken variable being updated, you will need to duplicate the line * configure headers = read('classpath:my-headers.js') from the 'caller' feature here as well. The above example does not use shared scope, which means that the variables in the 'calling' (parent) feature are not shared by the 'called' my-signin.feature. The above example can be made more simpler with the use of call (or callonce) without a def-assignment to a variable, and is the recommended pattern for implementing re-usable authentication setup flows.

Do look at the documentation and example for configure headers also as it goes hand-in-hand with call. In the above example, the end-result of the call to my-signin.feature resulted in the authToken variable being initialized. Take a look at how the configure headers example uses the authToken variable.

Call Tag Selector

You can "select" a single Scenario (or Scenario-s or Scenario Outline-s or even specific Examples rows) by appending a "tag selector" at the end of the feature-file you are calling. For example:

call read('classpath:my-signin.feature@name=someScenarioName')

While the tag does not need to be in the @key=value form, it is recommended for readability when you start getting into the business of giving meaningful names to your Scenario-s.

This "tag selection" capability is designed for you to be able to "compose" flows out of existing test-suites when using the Karate Gatling integration. Normally we recommend that you keep your "re-usable" features lightweight - by limiting them to just one Scenario.

Call Same Feature

As a convenience, you can call a tag directly, which is a short-cut to call another Scenario within the same feature file. Note that you would typically want to use the @ignore tag for such cases.

Scenario: one
* call read('@two')

@ignore @two
Scenario: two
* print 'called'

Data-Driven Features

If the argument passed to the call of a *.feature file is a JSON array, something interesting happens. The feature is invoked for each item in the array. Each array element is expected to be a JSON object, and for each object - the behavior will be as described above.

But this time, the return value from the call step will be a JSON array of the same size as the input array. And each element of the returned array will be the 'envelope' of variables that resulted from each iteration where the *.feature got invoked.

Here is an example that combines the table keyword with calling a *.feature. Observe how the get shortcut is used to 'distill' the result array of variable 'envelopes' into an array consisting only of response payloads.

* table kittens 
  | name   | age |
  | 'Bob'  |   2 |
  | 'Wild' |   1 |
  | 'Nyan' |   3 |

* def result = call read('cat-create.feature') kittens
* def created = $result[*].response
* match each created == { id: '#number', name: '#string', age: '#number' }
* match created[*].name contains only ['Bob', 'Wild', 'Nyan']

And here is how cat-create.feature could look like:

@ignore
Feature:

Scenario:
  Given url someUrlFromConfig
  And path 'cats'
  And request { name: '#(name)', age: '#(age)' }
  When method post
  Then status 200

If you replace the table with perhaps a JavaScript function call that gets some JSON data from some data-source, you can imagine how you could go about dynamic data-driven testing.

Although it is just a few lines of code, take time to study the above example carefully. It is a great example of how to effectively use the unique combination of syntax and JsonPath that Karate provides.

Also look at the demo examples, especially dynamic-params.feature - to compare the above approach with how the Gherkin Scenario Outline: can be alternatively used for data-driven tests.

Built-in variables for call

Although all properties in the passed JSON-like argument are 'unpacked' into the current scope as separate 'named' variables, it sometimes makes sense to access the whole argument and this can be done via __arg. And if being called in a loop, a built-in variable called __loop will also be available that will hold the value of the current loop index. So you can do things like this: * def name = name + __loop - or you can use the loop index value for looking up other values that may be in scope - in a data-driven style.

Refer to this demo feature for an example: kitten-create.feature

Default Values

Some users need "callable" features that are re-usable even when variables have not been defined by the calling feature. Normally an undefined variable results in nasty JavaScript errors. But there is an elegant way you can specify a default value using the karate.get() API:

# if foo is not defined, it will default to 42
* def foo = karate.get('foo', 42)

A word of caution: we recommend that you should not over-use Karate's capability of being able to re-use features. Re-use can sometimes result in negative benefits - especially when applied to test-automation. Prefer readability over re-use. See this for an example.

copy

For a call (or callonce) - payload / data structures (JSON, XML, Map-like or List-like) variables are 'passed by reference' which means that steps within the 'called' feature can update or 'mutate' them, for e.g. using the set keyword. This is actually the intent most of the time and is convenient. If you want to pass a 'clone' to a 'called' feature, you can do so using the rarely used copy keyword that works very similar to type conversion. This is best explained in this example: copy.feature.

Calling JavaScript Functions

Examples of defining and using JavaScript functions appear in earlier sections of this document. Being able to define and re-use JavaScript functions is a powerful capability of Karate. For example, you can:

• call re-usable functions that take complex data as an argument and return complex data that can be stored in a variable
• call and interoperate with Java code if needed
• share and re-use test utilities or 'helper' functionality across your organization

For an advanced example of how you can build and re-use a common set of JS functions, refer to this answer on Stack Overflow.

In real-life scripts, you would typically also use this capability of Karate to configure headers where the specified JavaScript function uses the variables that result from a sign in to manipulate headers for all subsequent HTTP requests. And it is worth mentioning that the Karate configuration 'bootstrap' routine is itself a JavaScript function.

Also refer to the eval keyword for a simpler way to execute arbitrary JavaScript that can be useful in some situations.

JS function argument rules for call

When using call (or callonce), only one argument is allowed. But this does not limit you in any way, because similar to how you can call *.feature files, you can pass a whole JSON object as the argument. In the case of the call of a JavaScript function, you can also pass a JSON array or a primitive (string, number, boolean) as the solitary argument, and the function implementation is expected to handle whatever is passed.

Instead of using call (or callonce) you are always free to call JavaScript functions 'normally' and then you can use more than one argument.

* def adder = function(a, b){ return a + b }
* assert adder(1, 2) == 3

Return types

Naturally, only one value can be returned. But again, you can return a JSON object. There are two things that can happen to the returned value.

Either - it can be assigned to a variable like so.

* def returnValue = call myFunction

Or - if a call is made without an assignment, and if the function returns a map-like object, it will add each key-value pair returned as a new variable into the execution context.

# while this looks innocent ...
# ... behind the scenes, it could be creating (or over-writing) a bunch of variables !
* call someFunction

While this sounds dangerous and should be used with care (and limits readability), the reason this feature exists is to quickly set (or over-write) a bunch of config variables when needed. In fact, this is the mechanism used when karate-config.js is processed on start-up.

Shared Scope

This behavior where all key-value pairs in the returned map-like object get automatically added as variables - applies to the calling of *.feature files as well. In other words, when call or callonce is used without a def, the 'called' script not only shares all variables (and configure settings) but can update the shared execution context. This is very useful to boil-down those 'common' steps that you may have to perform at the start of multiple test-scripts - into one-liners. But use wisely, because called scripts will now over-write variables that may have been already defined.

* def config = { user: 'john', password: 'secret' }
# this next line may perform many steps and result in multiple variables set for the rest of the script
* call read('classpath:common-setup.feature') config

You can use callonce instead of call within the Background in case you have multiple Scenario sections or Examples. Note the 'inline' use of the read function as a short-cut above. This applies to JS functions as well:

* call read('my-function.js')

These heavily commented demo examples can help you understand 'shared scope' better, and are designed to get you started with creating re-usable 'sign-in' or authentication flows:

Once you get comfortable with Karate, you can consider moving your authentication flow into a 'global' one-time flow using karate.callSingle(), think of it as 'callonce on steroids'.

call vs read()

Since this is a frequently asked question, the different ways of being able to re-use code (or data) are summarized below.

Calling Java

There are examples of calling JVM classes in the section on Java Interop and in the file-upload demo. Also look at the section on commonly needed utilities for more ideas.