Apache Sling Testing Clients
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
src
.gitignore
CODE_OF_CONDUCT.md
CONTRIBUTING.md
LICENSE
README.md Updating badges for org-apache-sling-testing-clients Jun 8, 2018
pom.xml SLING-7724: Change scope to QueryServlet dependencies Jun 12, 2018

README.md

Build Status Test Status Maven Central JavaDocs License testing

Apache Sling Testing Clients

This module is part of the Apache Sling project.

SlingClient is a specialized HttpClient that provides additional functionalities specific to Sling. It is designed to be easy to use out of the box, but also fully customizable. This library comes with a bunch of other specialized clients (built on top of SlingClient) that are ready to use.

Architecture

SlingClient implements the HttpClient interface, but deletegates this functionality to a private final CloseableHttpClient http field. The config is stored in a private final SlingClientConfig config field which is immutable and may be shared across multiple clients (more about it in the How to configure a SlingClient section). These two objects define the state of the client and are built to make the client thread safe.

SlingClient is designed in two layers:

  • The base class AbstractSlingClient implements HttpClient provides an overlay of basic http methods such as doGet(), doPost() & similar. These are meant to be full replacements of the ones in HttpClient for Sling specific needs, and they add specific customizations. One particularity is that they all return SlingHttpResponse, an augmented HttpResponse.

    Still, all the methods from HttpClient are exposed (through inheritance and delegation) in case one needs the raw functionality. Some useful methods to manipulate Sling paths and URLs have also been added (getUrl(), getPath()).

    This class encapsulates the mechanisms for extensibility (immutable config field, delegate client field, package private constructor, adaptTo()), but it is defined as abstract and should never be used directly.

  • The main class SlingClient extends AbstractSlingClient is the one that adds Sling specific methods (createNode(), deletePath() etc.). It has no fields, but makes use of everything that was defined in the super class. Another main functionality defined in SlingClient are the mechanisms to instantiate a SlingClient (and any other sub-class):

    • constructor: public SlingClient(URI url, String user, String password) throws ClientException

    • builder: public final static class Builder extends InternalBuilder<SlingClient> (more in How to write a Builder)

Any client you write should extend SlingClient (more in How to extend SlingClient)

How to instantiate SlingClient

There are several ways to obtain a SlingClient (and sub-client) instance, depending on the resources available:

  • constructor SlingClient(URI url, String user, String password) - handy for obtaining a simple client from the url:

    SlingClient c = new SlingClient(URI.create("localhost:8080"), "admin", "admin");
  • builder class Builder<T extends Builder> extends HttpClientBuilder - this allows for more complex clients to be created, e.g. with different authentication mechanism, or additional interceptors:

    SlingClient c = SlingClient.Builder.create("localhost:8080", "admin", "admin").build();

    This gives the possibility to customize the HttpClient (e.g. add interceptors, change auth method) before constructing it.

  • public <T extends AbstractSlingClient> T adaptTo(Class<T> clientClass) is the convenient method to obtain another specialized client form an existing one. The advantage is that the two will share the same configuration and http handler, so they will behave like two different "facets" of the same client (think about the analogy of a Web browser with multiple tabs).

Although the constructor and the builder are very handy, the preferred way of working with clients is to obtain it using one of the Junit Rules provided (e.g. ExistingQuickstart) and then call adaptTo().

How to configure SlingClient

All the configs specific to SlingClient are stored in private final SlingClientConfig config which contains fields such as url, cookieStore and authentication parameters. These fields can be set only indirectly, through constructor or Builder, and only before constructing the client. They cannot be changed, so if you need to change something, you must instantiate another client.

SlingClient was designed to be immutable, so thread safe. You don't have to worry about synchronizing it when running tests in parallel. Also, the immutable config is the base for the adaptTo() mechanism, since the two clients can share the same config.

How to extend SlingClient

The SlingClient was designed with extensibility in mind. That's why it provides only basic functionality, leaving other specialized clients to implement the rest. To create a new client class (let's call it MyClient), you need to:

  • extend SlingClient: class MyClient extends SlingClient
  • implement the two constructors:
    • the one for simple uses:
      public MyClient(URI serverUrl, String userName, String password) throws ClientException {
        super(serverUrl, userName, password);
      }
    • the one used by adaptTo() (so don't forget it!):
      public MyClient(CloseableHttpClient http, SlingClientConfig config) throws ClientException {
        super(http, config);
      }

A good example of how SlingClient can be extended is OsgiConsoleClient. Note you can further extend the sub-clients in the same way.

How to write a Builder

If you need to make your client customizable you will have to write your own Builder (you were no thinking to break the immutability by adding a setter, right?). Below is an example of how to create the Builder mechanism that you can take and adapt for your needs. In this case, we try to expose only one field foo, but it can be extended to any number of fields. Although it seems complicated, if you follow exactly the example, you cannot fail. Trying to simplify it will burn you (sooner or later), you have been warned!

A short description of the Builder architecture would be: the InternalBuilder contains all the logic while staying extensible, while Builder takes all the credit by exposing the build() method. Yet, the Builder cannot be extended because all the sub-classes would return a SlingClient when calling build() (and not a subclass instance).

@Immutable
public class MyClient extends SlingClient {

    private final String foo;

    public MyClient(URI serverUrl, String user, String password) throws ClientException {
        super(serverUrl, user, password);
    }

    /**
     * Constructor used by Builders and adaptTo(). <b>Should never be called directly from the code.</b>
     *
     * @see AbstractSlingClient#AbstractSlingClient(CloseableHttpClient, SlingClientConfig)
     */
    public MyClient(CloseableHttpClient http, SlingClientConfig config) throws ClientException {
        super(http, config);
    }

    public static abstract class InternalBuilder<T extends MyClient> extends SlingClient.InternalBuilder<T> {
        protected String foo;

        protected InternalBuilder(URI url, String user, String password) {
            super(url, user, password);
        }

        public InternalBuilder<T> withFoo(String foo) {
          this.foo = foo;
        }
    }

    public final static class Builder extends InternalBuilder<MyClient> {

        private Builder(URI url, String user, String password) {
            super(url, user, password);
        }

        @Override
        public MyClient build() throws ClientException {
            MyClient client = new MyClient(buildHttpClient(), buildSlingClientConfig());
            client.foo = this.foo;
            return client;
        }

        public static Builder create(URI url, String user, String password) {
            return new Builder(url, user, password);
        }
    }
}

FAQ

How can I change the server url of an existing client?

You don't. As described in How to configure a SlingClient, you have to instantiate another client to change the config.

How can I create a client for a server url with context path?

The server url (passed in the constructor or builder) must contain all the elements, including protocol, hostname, port and eventually the context path, e.g.: http://localhost:8080/mycontextpath/. The url may (or may not) contain the trailing slash. Yet, the client will always store it with a trailing slash:

SlingClient client = new SlingClient("http://localhost:4502/mycontextpath", "user", "pass");
System.out.println(client.getUrl());
// prints http://localhost:4502/mycontextpath/
How can I customize the underlying HttpClient?

The SlingClient.Builder directly exposes the most useful methods from HttpClientBuilder, but not all of them. First, check if you can find it there. If you haven't found your specific method, then the Builder exposes an HttpClientBuilder through public HttpClientBuilder httpClientBuilder() which you can use to config it. Note that in this case you cannot chain the methods to build the client, so you will need to keep a reference to the SlingClient.Builder:

SlingClient.Builder builder = SlingClient.Builder.create("http://localhost:8080", "user", "pass");
HttpClientBuilder httpBuilder = builder.httpClientBuilder();
httpBuilder.setProxy(myProxy);
builder.setUser("another");
SlingClient client = builder.build();
Why is the Builder pattern so complicated? Do I really need two classes?

Don't try to get creative here. Respect the examples provided and don't take shortcuts, otherwise you will hurt yourself.

We have tried different ways of designing the Builder. This is the best compromise between extensibility and simplicity. The HttpClientBuilder does not offer any extensibility support, so SlingClient.Builder does not extend it, it just uses it internally. Always remember that you don't need to create your Builder, unless you want to add custom fields to the client.

Why I cannot use the entity's content InputStream?

SlingClient#doRequest(), SlingClient#doGet(), SlingClient#doPost() & co. are all consuming the entity and caching it as String. This is by design, since there's a big risk to forget closing the connections and to run out of sockets quickly. If you need the response content as InputStream (e.g. for downloading a binary), you can use doStreamGet() or similar. These methods were written specially to not consume the entity so it's the caller's responsibility to close it when finished. Remember to use them with caution and only when needed.

Can my client use another authentication method?

The username and password required by the constructor and builder are there for convenience (since more than 90% of cases will use basic auth). But you can easily overwrite the CredentialsProvider in Builder so those will be ignored. Or do anything you want with that HttpClientBuilder...

How can I obtain the context path?

client.getUrl().getPath()

How can I obtain the "relative" url (excluding hostname and port, but including context path)?

client.getUrl(path).getPath()

How can I remove the context path from a path?

client.getPath(path)

What if I pass an url or a path with or without context path to getUrl() or getPath()?

We have tried to make these methods as robust as possible. Their job is very clear:

  • getUrl(String path) to transform a Sling path into a full url
  • getPath(String url) to transform a full url into a Sling path

Any input that does not respect the contract might not work. Check AbstractSlingClientGetPathTest and AbstractSlingClientGetUrlTest for an extensive list of cases that we have considered when writing these methods.