True Object-Oriented and Immutable Java Web Framework
Java Other
Latest commit 07f1db2 Feb 19, 2017 @yegor256 many

README.md

Made By Teamed.io DevOps By Rultor.com We recommend IntelliJ IDEA

Build Status Build status Javadoc Coverage Status codebeat badge Codacy Badge License

Maven Central Dependencies PDD status

Takes is a true object-oriented and immutable Java7 web development framework. Its key benefits, comparing to all others, include these four fundamental principles:

  1. not a single null (why NULL is bad?)
  2. not a single public static method (why they are bad?)
  3. not a single mutable class (why they are bad?)
  4. not a single instanceof keyword, type casting, or reflection (why?)

Of course, there are no configuration files. Besides that, these are more traditional features, out of the box:

This is what is not supported and won't be supported:

These blog posts may help you too.

Contents

Quick Start

Create this App.java file:

import org.takes.http.Exit;
import org.takes.http.FtBasic;
import org.takes.facets.fork.FkRegex;
import org.takes.facets.fork.TkFork;
public final class App {
  public static void main(final String... args) throws Exception {
    new FtBasic(
      new TkFork(new FkRegex("/", "hello, world!")), 8080
    ).start(Exit.NEVER);
  }
}

Then, download takes.jar and compile your Java code:

$ javac -cp takes.jar App.java

Now, run it like this:

$ java -Dfile.encoding=UTF-8 -cp takes.jar:. App

Should work :)

This code starts a new HTTP server on port 8080 and renders a plain-text page on all requests at the root URI.

Important: Pay attention that UTF-8 encoding is set on the command line. The entire framework relies on your default Java encoding, which is not necessarily UTF-8 by default. To be sure, always set it on the command line with file.encoding Java argument. We decided not to hard-code "UTF-8" in our code mostly because this would be against the entire idea of Java localization, according to which a user always should have a choice of encoding and language selection. We're using Charset.defaultCharset() everywhere in the code.

Build and Run With Maven

If you're using Maven, this is how your pom.xml should look like:

<project>
  <dependencies>
    <dependency>
      <groupId>org.takes</groupId>
      <artifactId>takes</artifactId>
    </dependency>
  </dependencies>
  <profiles>
    <profile>
      <id>hit-refresh</id>
      <build>
        <plugins>
          <plugin>
            <groupId>org.codehaus.mojo</groupId>
            <artifactId>exec-maven-plugin</artifactId>
            <version>1.3</version>
            <executions>
              <execution>
                <id>start-server</id>
                <phase>pre-integration-test</phase>
                <goals>
                  <goal>java</goal>
                </goals>
                <configuration>
                  <mainClass>foo.App</mainClass> <!-- your main class -->
                  <cleanupDaemonThreads>false</cleanupDaemonThreads>
                  <arguments>
                    <argument>--port=${port}</argument>
                  </arguments>
                </configuration>
              </execution>
            </executions>
          </plugin>
        </plugins>
      </build>
    </profile>
  </profiles>
</project>

With this configutation you can run it from command line:

$ mvn clean integration-test -Phit-refresh -Dport=8080

Maven will start the server and you can see it at http://localhost:8080.

Unit Testing

This is how you can unit test the app, using JUnit 4.x and Hamcrest:

public final class AppTest {
  @Test
  public void returnsHttpResponse() throws Exception {
    MatcherAssert.assertThat(
      new RsPrint(
        new App().act(new RqFake("GET", "/"))
      ).printBody(),
      Matchers.equalsTo("hello, world!")
    );
  }
}

You can create a fake request with form parameters like this:

new RqForm.Fake(
  new RqFake(),
  "foo", "value-1",
  "bar", "value-2"
)

Integration Testing

Here is how you can test the entire server via HTTP, using JUnit and jcabi-http for making HTTP requests:

public final class AppITCase {
  @Test
  public void returnsTextPageOnHttpRequest() throws Exception {
    new FtRemote(new App()).exec(
      new FtRemote.Script() {
        @Override
        public void exec(final URI home) throws IOException {
          new JdkRequest(home)
            .fetch()
            .as(RestResponse.class)
            .assertStatus(HttpURLConnection.HTTP_OK)
            .assertBody(Matchers.equalTo("hello, world!"));
        }
      }
    );
  }
}

More complex integration testing examples you can find in one of the open source projects that are using Take, for example: rultor.com.

A Bigger Example

Let's make it a bit more sophisticated:

public final class App {
  public static void main(final String... args) {
    new FtBasic(
      new TkFork(
        new FkRegex("/robots\\.txt", ""),
        new FkRegex("/", new TkIndex())
      ),
      8080
    ).start(Exit.NEVER);
  }
}

The FtBasic is accepting new incoming sockets on port 8080, parses them according to HTTP 1.1 specification and creates instances of class Request. Then, it gives requests to the instance of TkFork (tk stands for "take") and expects it to return an instance of Take back. As you probably understood already, the first regular expression that matches returns a take. TkIndex is our custom class, let's see how it looks:

public final class TkIndex implements Take {
  @Override
  public Response act(final Request req) {
    return new RsHtml("<html>Hello, world!</html>");
  }
}

It is immutable and must implement a single method act(), which is returning an instance of Response. So far so good, but this class doesn't have an access to an HTTP request. Here is how we solve this:

new TkFork(
  new FkRegex(
    "/file/(?<path>[^/]+)",
    new TkRegex() {
      @Override
      public Response act(final RqRegex request) throws IOException {
        final File file = new File(
          request.matcher().group("path")
        );
        return new RsHTML(
          FileUtils.readFileToString(file, Charsets.UTF_8)
        );
      }
    }
  )
)

We're using TkRegex instead of Take, in order to deal with RqRegex instead of a more generic Request. RqRegex gives an instance of Matcher used by FkRegex for pattern matching.

Here is a more complex and verbose example:

public final class App {
  public static void main(final String... args) {
    new FtBasic(
      new TkFork(
        new FkRegex("/robots.txt", ""),
        new FkRegex("/", new TkIndex()),
        new FkRegex(
          "/xsl/.*",
          new TkWithType(new TkClasspath(), "text/xsl")
        ),
        new FkRegex("/account", new TkAccount(users)),
        new FkRegex("/balance/(?<user>[a-z]+)", new TkBalance())
      )
    ).start(Exit.NEVER);
  }
}

Templates

Now let's see how we can render something more complex than an plain text. First, XML+XSLT is a recommended mechanism of HTML rendering. Even though it may be too complex, give it a try, you won't regret. Here is how we render a simple XML page that is transformed to HTML5 on-fly (more about RsXembly read below):

public final class TkAccount implements Take {
  private final Users users;
  public TkAccount(final Users users) {
    this.users = users;
  }
  @Override
  public Response act(final Request req) {
    final User user = this.users.find(new RqCookies(req).get("user"));
    return new RsLogin(
      new RsXSLT(
        new RsXembly(
          new XeStylesheet("/xsl/account.xsl"),
          new XeAppend("page", user)
        )
      ),
      user
    );
  }
}

This is how that User class may look like:

public final class User implements XeSource {
  private final String name;
  private final int balance;
  @Override
  public Iterable<Directive> toXembly() {
    return new Directives().add("user")
      .add("name").set(this.name).up()
      .add("balance").set(Integer.toString(this.balance));
  }
}

Here is how RsLogin may look like:

public final class RsLogin extends RsWrap {
  public RsLogin(final Response response, final User user) {
    super(
      new RsWithCookie(
        response, "user", user.toString()
      )
    );
  }
}

Velocity Templates

Let's say, you want to use Velocity:

public final class TkHelloWorld implements Take {
  @Override
  public Response act(final Request req) {
    return new RsVelocity(
      "hi, ${user.name}! You've got ${user.balance}",
      new RsVelocity.Pair("user", new User())
    );
  }
}

You will need this extra dependency in classpath:

<dependency>
  <groupId>org.apache.velocity</groupId>
  <artifactId>velocity-engine-core</artifactId>
  <scope>runtime</scope>
</dependency>

Static Resources

Very often you need to serve static resources to your web users, like CSS stylesheets, images, JavaScript files, etc. There are a few supplementary classes for that:

new TkFork(
  new FkRegex("/css/.+", new TkWithType(new TkClasspath(), "text/css")),
  new FkRegex("/data/.+", new TkFiles(new File("/usr/local/data"))
)

Class TkClasspath take static part of the request URI and finds a resource with this name in classpath.

TkFiles just looks by file name in the directory configured.

TkWithType sets content type of all responses coming out of the decorated take.

Hit Refresh Debugging

It is a very convenient feature. Once you start the app you want to be able to modify its static resources (CSS, JS, XSL, etc), refresh the page in a browser and immediately see the result. You don't want to re-compile the entire project and restart it. Here is what you need to do to your sources in order to enable that feature:

new TkFork(
  new FkRegex(
    "/css/.+",
    new TkWithType(
      new TkFork(
        new FkHitRefresh(
          "./src/main/resources/foo/scss/**", // what sources to watch
          "mvn sass:compile", // what to run when sources are modified
          new TkFiles("./target/css")
        )
        new FkFixed(new TkClasspath())
      ),
      "text/css"
    )
  )
)

This FkHitRefresh fork is a decorator of take. Once it sees X-Take-Refresh header in the request, it realizes that the server is running in "hit-refresh" mode and passes the request to the encapsulated take. Before it passes the request it tries to understand whether any of the resources are older than compiled files. If they are older, it tries to run compilation tool to build them again.

Request Methods (POST, PUT, HEAD, etc.)

Here is an example:

new TkFork(
  new FkRegex(
    "/user",
    new TkFork(
      new FkMethods("GET", new TkGetUser()),
      new FkMethods("POST,PUT", new TkPostUser()),
      new FkMethods("DELETE", new TkDeleteUser())
    )
  )
)

Request Parsing

Here is how you can parse an instance of Request:

Href href = new RqHref.Base(request).href();
URI uri = href.uri();
Iterable<String> values = href.param("key");

For a more complex parsing try to use Apache Http Client or something similar.

Form Processing

Here is an example:

public final class TkSavePhoto implements Take {
  @Override
  public Response act(final Request req) {
    final String name = new RqForm(req).param("name");
    return new RsWithStatus(HttpURLConnection.HTTP_NO_CONTENT);
  }
}

Exception Handling

By default, TkFork lets all exceptions bubble up. If one of your take crashes, a user will see a default error page. Here is how you can configure this behavior:

public final class App {
  public static void main(final String... args) {
    new FtBasic(
      new TkFallback(
        new TkFork(
          new FkRegex("/robots\\.txt", ""),
          new FkRegex("/", new TkIndex())
        ),
        new FbChain(
          new FbStatus(404, new RsText("sorry, page is absent")),
          new FbStatus(405, new RsText("this method is not allowed here")),
          new Fallback() {
            @Override
            public Iterator<Response> route(final RqFallback req) {
              return Collections.<Response>singleton(
                new RsHTML("oops, something went terribly wrong!")
              ).iterator();
            }
          }
        )
      ),
      8080
    ).start(Exit.NEVER);
  }
}

TkFallback decorates an instance of Take and catches all exceptions any of its take may throw. Once it's thrown, an instance of FbChain will find the most suitable fallback and will fetch a response from there.

Redirects

Sometimes it's very useful to return a redirect response (30x status code), either by a normal return or by throwing an exception. This example illustrates both methods:

public final class TkPostMessage implements Take {
  @Override
  public Response act(final Request req) {
    final String body = new RqPring(req).printBody();
    if (body.isEmpty()) {
      throw new RsForward(
        new RsFlash("message can't be empty")
      );
    }
    // save the message to the database
    return new RsForward(
      new RsFlash(
        "thanks, the message was posted"
      ),
      "/"
    );
  }
}

Then, you should decorate the entire TkFork with this TkForward and TkFlash:

public final class App {
  public static void main(final String... args) {
    new FtBasic(
      new TkFlash(
        new TkForward(
          new TkFork(new FkRegex("/", new TkPostMessage())
        )
      ),
      8080
    ).start(Exit.NEVER);
  }
}

RsJSON

Here is how we can deal with JSON:

public final class TkBalance extends TkFixed {
  @Override
  public Response act(final RqRegex request) {
    return new RsJSON(
      new User(request.matcher().group("user")))
    );
  }
}

This is the method to add to User:

public final class User implements XeSource, RsJSON.Source {
  @Override
  public JsonObject toJSON() {
    return Json.createObjectBuilder()
      .add("balance", this.balance)
      .build();
  }
}

RsXembly

Here is how you generate an XML page using Xembly:

Response response = new RsXembly(
  new XeAppend("page"),
  new XeDirectives("XPATH '/page'", this.user)
)

This is a complete example, with all possible options:

Response response = new RsXembly(
  new XeStylesheet("/xsl/account.xsl"), // add processing instruction
  new XeAppend(
    "page", // create a DOM document with "page" root element
    new XeMillis(false), // add "millis" attribute to the root, with current time
    user, // add user to the root element
    new XeSource() {
      @Override
      public Iterable<Directive> toXembly() {
        return new Directives().add("status").set("alive");
      }
    },
    new XeMillis(true), // replace "millis" attribute with take building time
  ),
)

This is the output that will be produced:

<?xml version='1.0'?>
<?xsl-stylesheet href='/xsl/account.xsl'?>
<page>
  <millis>5648</millis>
  <user>
    <name>Jeff Lebowski</name>
    <balance>123</balance>
  </user>
  <status>alive</status>
</page>

To avoid duplication of all this scaffolding in every page, you can create your own class, which will be used in every page, for example:

Response response = new RsXembly(
  new XeFoo(user)
)

This is how this XeFoo class would look like:

public final class XeFoo extends XeWrap {
  public XeFoo(final String stylesheet, final XeSource... sources) {
    super(
      new XeAppend(
        "page",
        new XeMillis(false),
        new XeStylesheet(stylesheet),
        new XeChain(sources),
        new XeSource() {
          @Override
          public Iterable<Directive> toXembly() {
            return new Directives().add("status").set("alive");
          }
        },
        new XeMillis(true)
      )
    );
  }
}

You will need this extra dependency in classpath:

<dependency>
  <groupId>com.jcabi.incubator</groupId>
  <artifactId>xembly</artifactId>
</dependency>

More about this mechanism in this blog post: XML Data and XSL Views in Takes Framework.

Cookies

Here is how we drop a cookie to the user:

public final class TkIndex implements Take {
  @Override
  public Response act(final Request req) {
    return new RsWithCookie("auth", "John Doe");
  }
}

An HTTP response will contain this header, which will place a auth cookie into the user's browser:

HTTP/1.1 200 OK
Set-Cookie: auth="John Doe"

This is how you read cookies from a request:

public final class TkIndex implements Take {
  @Override
  public Response act(final Request req) {
    // the list may be empty
    final Iterable<String> cookies = new RqCookies(req).cookie("my-cookie");
  }
}

GZIP Compression

If you want to compress all your responses with GZIP, wrap your take in TkGzip:

new TkGzip(take)

Now, each request that contains Accept-Encoding request header with gzip compression method inside will receive a GZIP-compressed response. Also, you can compress an individual response, using RsGzip decorator.

Content Negotiation

Say, you want to return different content based on Accept header of the request (a.k.a. content negotation):

public final class TkIndex implements Take {
  @Override
  public Response act(final Request req) {
    return new RsFork(
      req,
      new FkTypes("text/*", new RsText("it's a text"))
      new FkTypes("application/json", new RsJSON("{\"a\":1}"))
      new FkTypes("image/png", /* something else */)
    );
  }
}

Authentication

Here is an example of login via Facebook:

new TkAuth(
  new TkFork(
    new FkRegex("/", new TkHTML("hello, check <a href='/acc'>account</a>")),
    new FkRegex("/acc", new TkSecure(new TkAccount()))
  ),
  new PsChain(
    new PsCookie(
      new CcSafe(new CcHex(new CcXOR(new CcCompact(), "secret-code")))
    ),
    new PsByFlag(
      new PsByFlag.Pair(
        PsFacebook.class.getSimpleName(),
        new PsFacebook("facebook-app-id", "facebook-secret")
      ),
      new PsByFlag.Pair(
        PsLogout.class.getSimpleName(),
        new PsLogout()
      )
    )
  )
)

Then, you need to show a login link to the user, which he or she can click and get to the Facebook OAuth authentication page. Here is how you do this with XeResponse:

new RsXembly(
  new XeStylesheet("/xsl/index.xsl"),
  new XeAppend(
    "page",
    new XeFacebookLink(req, "facebook-app-id"),
    // ... other xembly sources
  )
)

The link will be add to the XML page like this:

<page>
  <links>
    <link rel="take:facebook" href="https://www.facebook.com/dialog/oauth..."/>
  </links>
</page>

Similar mechanism can be used for PsGithub, PsGoogle, PsLinkedin, PsTwitter, etc.

This is how you get currently logged in user:

public final class TkAccount implements Take {
  @Override
  public Response act(final Request req) {
    final Identity identity = new RqAuth(req).identity();
    if (this.identity.equals(Identity.ANONYMOUS)) {
      // returns "urn:facebook:1234567" for a user logged in via Facebook
      this.identity().urn();
    }
  }
}

More about it in this blog post: How Cookie-Based Authentication Works in the Takes Framework

Command Line Arguments

There is a convenient class FtCLI that parses command line arguments and starts the necessary Front accordingly.

There are a few command line arguments that should be passed to FtCLI constructor:

--port=1234         Tells the server to listen to TCP port 1234
--lifetime=5000     The server will die in five seconds (useful for integration testing)
--hit-refresh       Run the server in hit-refresh mode
--daemon            Runs the server in Java daemon thread (for integration testing)
--threads=30        Processes incoming HTTP requests in 30 parallel threads
--max-latency=5000  Maximum latency in milliseconds per each request
                    (longer requests will be interrupted)

For example:

public final class App {
  public static void main(final String... args) {
    new FtCLI(
      new TkFork(new FkRegex("/", "hello, world!")),
      args
    ).start(Exit.NEVER);
  }
}

Then, run it like this:

$ java -cp take.jar App.class --port=8080 --hit-refresh

You should see "hello, world!" at http://localhost:8080.

Parameter --port also accepts file name, instead of a number. If the file exists, FtCLI will try to read its content and use it as port number. If the file is absent, FtCLI will allocate a new random port number, use it to start a server, and save it to the file.

Logging

The framework sends all logs to SLF4J logging facility. If you want to see them, configure one of SLF4J bindings.

To make a Take log, wrap it in a TkSlf4j - for example:

 new TkSlf4j(
     new TkFork(...)
 )

Directory Layout

You are free to use any build tool, but we recommend Maven. This is how your project directory layout may/should look like:

/src
  /main
    /java
      /foo
        App.java
    /scss
    /coffeescript
    /resources
      /xsl
      /js
      /css
      robot.txt
      log4j.properties
  /test
    /java
      /foo
        AppTest.java
    /resources
      log4j.properties
pom.xml
LICENSE.txt

Optional dependencies

If you're using Maven and include Takes as a dependency in your own project, you can choose which of the optional dependencies to include in your project. The list of all of the optional dependencies can be seen in the Takes project pom.xml.

For example, to use the Facebook API shown above, simply add a dependency to the restfb API in your project:

<dependency>
  <groupId>com.restfb</groupId>
  <artifactId>restfb</artifactId>
  <version>1.15.0</version>
  <scope>runtime</scope>
</dependency>

How to contribute

Fork repository, make changes, send us a pull request. We will review your changes and apply them to the master branch shortly, provided they don't violate our quality standards. To avoid frustration, before sending us your pull request please run full Maven build:

$ mvn clean install -Pqulice

To avoid build errors use maven 3.2+.

Pay attention that our pom.xml inherits a lot of configuration from jcabi-parent. This article explains why it's done this way.

Got questions?

If you have questions or general suggestions, don't hesitate to submit a new Github issue.