How to use java.net.URLConnection to fire and handle HTTP requests

Use of java.net.URLConnection is asked about pretty often here, and the Oracle tutorial is too concise about it.

That tutorial basically only shows how to fire a GET request and read the response. It doesn't explain anywhere how to use it to among others perform a POST request, set request headers, read response headers, deal with cookies, submit a HTML form, upload a file, etc.

So, how can I use java.net.URLConnection to fire and handle "advanced" HTTP requests?

First a disclaimer beforehand: the posted code snippets are all basic examples. You'll need to handle trivial IOException s and RuntimeException s like NullPointerException , ArrayIndexOutOfBoundsException and consorts yourself.

Preparing

We first need to know at least the URL and the charset. The parameters are optional and depend on the functional requirements.

String url = "http://example.com";
String charset = "UTF-8";  // Or in Java 7 and later, use the constant: java.nio.charset.StandardCharsets.UTF_8.name()
String param1 = "value1";
String param2 = "value2";
// ...

String query = String.format("param1=%s&param2=%s", 
     URLEncoder.encode(param1, charset), 
     URLEncoder.encode(param2, charset));

The query parameters must be in name=value format and be concatenated by & . You would normally also URL-encode the query parameters with the specified charset using URLEncoder#encode() .

The String#format() is just for convenience. I prefer it when I would need the String concatenation operator + more than twice.

Firing a HTTP GET request with (optionally) query parameters

It's a trivial task. It's the default request method.

URLConnection connection = new URL(url + "?" + query).openConnection();
connection.setRequestProperty("Accept-Charset", charset);
InputStream response = connection.getInputStream();
// ...

Any query string should be concatenated to the URL using ? . The Accept-Charset header may hint the server what encoding the parameters are in. If you don't send any query string, then you can leave the Accept-Charset header away. If you don't need to set any headers, then you can even use the URL#openStream() shortcut method.

InputStream response = new URL(url).openStream();
// ...

Either way, if the other side is a HttpServlet , then its doGet() method will be called and the parameters will be available by HttpServletRequest#getParameter() .

For testing purposes, you can print the response body to stdout as below:

try (Scanner scanner = new Scanner(response)) {
    String responseBody = scanner.useDelimiter("A").next();
    System.out.println(responseBody);
}

Firing a HTTP POST request with query parameters

Setting the URLConnection#setDoOutput() to true implicitly sets the request method to POST. The standard HTTP POST as web forms do is of type application/x-www-form-urlencoded wherein the query string is written to the request body.

URLConnection connection = new URL(url).openConnection();
connection.setDoOutput(true); // Triggers POST.
connection.setRequestProperty("Accept-Charset", charset);
connection.setRequestProperty("Content-Type", "application/x-www-form-urlencoded;charset=" + charset);

try (OutputStream output = connection.getOutputStream()) {
    output.write(query.getBytes(charset));
}

InputStream response = connection.getInputStream();
// ...

Note: whenever you'd like to submit a HTML form programmatically, don't forget to take the name=value pairs of any <input type="hidden"> elements into the query string and of course also the name=value pair of the <input type="submit"> element which you'd like to "press" programmatically (because that's usually been used in the server side to distinguish if a button was pressed and if so, which one).

You can also cast the obtained URLConnection to HttpURLConnection and use its HttpURLConnection#setRequestMethod() instead. But if you're trying to use the connection for output you still need to set URLConnection#setDoOutput() to true .

HttpURLConnection httpConnection = (HttpURLConnection) new URL(url).openConnection();
httpConnection.setRequestMethod("POST");
// ...

Either way, if the other side is a HttpServlet , then its doPost() method will be called and the parameters will be available by HttpServletRequest#getParameter() .

Actually firing the HTTP request

You can fire the HTTP request explicitly with URLConnection#connect() , but the request will automatically be fired on demand when you want to get any information about the HTTP response, such as the response body using URLConnection#getInputStream() and so on. The above examples does exactly that, so the connect() call is in fact superfluous.

Gathering HTTP response information

int status = httpConnection.getResponseCode();

for (Entry<String, List<String>> header : connection.getHeaderFields().entrySet()) {
    System.out.println(header.getKey() + "=" + header.getValue());
}

String contentType = connection.getHeaderField("Content-Type");
String charset = null;

for (String param : contentType.replace(" ", "").split(";")) {
    if (param.startsWith("charset=")) {
        charset = param.split("=", 2)[1];
        break;
    }
}

if (charset != null) {
    try (BufferedReader reader = new BufferedReader(new InputStreamReader(response, charset))) {
        for (String line; (line = reader.readLine()) != null;) {
            // ... System.out.println(line) ?
        }
    }
} else {
    // It's likely binary content, use InputStream/OutputStream.
}

Maintaining the session

The server side session is usually backed by a cookie. Some web forms require that you're logged in and/or are tracked by a session. You can use the CookieHandler API to maintain cookies. You need to prepare a CookieManager with a CookiePolicy of ACCEPT_ALL before sending all HTTP requests.

// First set the default cookie manager.
CookieHandler.setDefault(new CookieManager(null, CookiePolicy.ACCEPT_ALL));

// All the following subsequent URLConnections will use the same cookie manager.
URLConnection connection = new URL(url).openConnection();
// ...

connection = new URL(url).openConnection();
// ...

connection = new URL(url).openConnection();
// ...

Note that this is known to not always work properly in all circumstances. If it fails for you, then best is to manually gather and set the cookie headers. You basically need to grab all Set-Cookie headers from the response of the login or the first GET request and then pass this through the subsequent requests.

// Gather all cookies on the first request.
URLConnection connection = new URL(url).openConnection();
List<String> cookies = connection.getHeaderFields().get("Set-Cookie");
// ...

// Then use the same cookies on all subsequent requests.
connection = new URL(url).openConnection();
for (String cookie : cookies) {
    connection.addRequestProperty("Cookie", cookie.split(";", 2)[0]);
}
// ...

The split(";", 2)[0] is there to get rid of cookie attributes which are irrelevant for the server side like expires , path , etc. Alternatively, you could also use cookie.substring(0, cookie.indexOf(';')) instead of split() .

Streaming mode

The HttpURLConnection will by default buffer the entire request body before actually sending it, regardless of whether you've set a fixed content length yourself using connection.setRequestProperty("Content-Length", contentLength); . This may cause OutOfMemoryException s whenever you concurrently send large POST requests (eg uploading files). To avoid this, you would like to set the HttpURLConnection#setFixedLengthStreamingMode() .

httpConnection.setFixedLengthStreamingMode(contentLength);

But if the content length is really not known beforehand, then you can make use of chunked streaming mode by setting the HttpURLConnection#setChunkedStreamingMode() accordingly. This will set the HTTP Transfer-Encoding header to chunked which will force the request body being sent in chunks. The below example will send the body in chunks of 1KB.

httpConnection.setChunkedStreamingMode(1024);

User-Agent

It can happen that a request returns an unexpected response, while it works fine with a real web browser. The server side is probably blocking requests based on the User-Agent request header. The URLConnection will by default set it to Java/1.6.0_19 where the last part is obviously the JRE version. You can override this as follows:

connection.setRequestProperty("User-Agent", "Mozilla/5.0 (Windows NT 6.1) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/41.0.2228.0 Safari/537.36"); // Do as if you're using Chrome 41 on Windows 7.

Use the User-Agent string from a recent browser.

Error handling

If the HTTP response code is 4nn (Client Error) or 5nn (Server Error), then you may want to read the HttpURLConnection#getErrorStream() to see if the server has sent any useful error information.

InputStream error = ((HttpURLConnection) connection).getErrorStream();

If the HTTP response code is -1, then something went wrong with connection and response handling. The HttpURLConnection implementation is in older JREs somewhat buggy with keeping connections alive. You may want to turn it off by setting the http.keepAlive system property to false . You can do this programmatically in the beginning of your application by:

System.setProperty("http.keepAlive", "false");

Uploading files

You'd normally use multipart/form-data encoding for mixed POST content (binary and character data). The encoding is in more detail described in RFC2388.

String param = "value";
File textFile = new File("/path/to/file.txt");
File binaryFile = new File("/path/to/file.bin");
String boundary = Long.toHexString(System.currentTimeMillis()); // Just generate some unique random value.
String CRLF = "rn"; // Line separator required by multipart/form-data.
URLConnection connection = new URL(url).openConnection();
connection.setDoOutput(true);
connection.setRequestProperty("Content-Type", "multipart/form-data; boundary=" + boundary);

try (
    OutputStream output = connection.getOutputStream();
    PrintWriter writer = new PrintWriter(new OutputStreamWriter(output, charset), true);
) {
    // Send normal param.
    writer.append("--" + boundary).append(CRLF);
    writer.append("Content-Disposition: form-data; name="param"").append(CRLF);
    writer.append("Content-Type: text/plain; charset=" + charset).append(CRLF);
    writer.append(CRLF).append(param).append(CRLF).flush();

    // Send text file.
    writer.append("--" + boundary).append(CRLF);
    writer.append("Content-Disposition: form-data; name="textFile"; filename="" + textFile.getName() + """).append(CRLF);
    writer.append("Content-Type: text/plain; charset=" + charset).append(CRLF); // Text file itself must be saved in this charset!
    writer.append(CRLF).flush();
    Files.copy(textFile.toPath(), output);
    output.flush(); // Important before continuing with writer!
    writer.append(CRLF).flush(); // CRLF is important! It indicates end of boundary.

    // Send binary file.
    writer.append("--" + boundary).append(CRLF);
    writer.append("Content-Disposition: form-data; name="binaryFile"; filename="" + binaryFile.getName() + """).append(CRLF);
    writer.append("Content-Type: " + URLConnection.guessContentTypeFromName(binaryFile.getName())).append(CRLF);
    writer.append("Content-Transfer-Encoding: binary").append(CRLF);
    writer.append(CRLF).flush();
    Files.copy(binaryFile.toPath(), output);
    output.flush(); // Important before continuing with writer!
    writer.append(CRLF).flush(); // CRLF is important! It indicates end of boundary.

    // End of multipart/form-data.
    writer.append("--" + boundary + "--").append(CRLF).flush();
}

If the other side is a HttpServlet , then its doPost() method will be called and the parts will be available by HttpServletRequest#getPart() (note, thus not getParameter() and so on!). The getPart() method is however relatively new, it's introduced in Servlet 3.0 (Glassfish 3, Tomcat 7, etc). Prior to Servlet 3.0, your best choice is using Apache Commons FileUpload to parse a multipart/form-data request. Also see this answer for examples of both the FileUpload and the Servelt 3.0 approaches.

Dealing with untrusted or misconfigured HTTPS sites

Sometimes you need to connect a HTTPS URL, perhaps because you're writing a web scraper. In that case, you may likely face a javax.net.ssl.SSLException: Not trusted server certificate on some HTTPS sites who doesn't keep their SSL certificates up to date, or a java.security.cert.CertificateException: No subject alternative DNS name matching [hostname] found or javax.net.ssl.SSLProtocolException: handshake alert: unrecognized_name on some misconfigured HTTPS sites.

The following one-time-run static initializer in your web scraper class should make HttpsURLConnection more lenient as to those HTTPS sites and thus not throw those exceptions anymore.

static {
    TrustManager[] trustAllCertificates = new TrustManager[] {
        new X509TrustManager() {
            @Override
            public X509Certificate[] getAcceptedIssuers() {
                return null; // Not relevant.
            }
            @Override
            public void checkClientTrusted(X509Certificate[] certs, String authType) {
                // Do nothing. Just allow them all.
            }
            @Override
            public void checkServerTrusted(X509Certificate[] certs, String authType) {
                // Do nothing. Just allow them all.
            }
        }
    };

    HostnameVerifier trustAllHostnames = new HostnameVerifier() {
        @Override
        public boolean verify(String hostname, SSLSession session) {
            return true; // Just allow them all.
        }
    };

    try {
        System.setProperty("jsse.enableSNIExtension", "false");
        SSLContext sc = SSLContext.getInstance("SSL");
        sc.init(null, trustAllCertificates, new SecureRandom());
        HttpsURLConnection.setDefaultSSLSocketFactory(sc.getSocketFactory());
        HttpsURLConnection.setDefaultHostnameVerifier(trustAllHostnames);
    }
    catch (GeneralSecurityException e) {
        throw new ExceptionInInitializerError(e);
    }
}

Last words

The Apache HttpComponents HttpClient is much more convenient in this all :)

Parsing and extracting HTML

If all you want is parsing and extracting data from HTML, then better use a HTML parser like Jsoup

When working with HTTP it's almost always more useful to refer to HttpURLConnection rather than the base class URLConnection (since URLConnection is an abstract class when you ask for URLConnection.openConnection() on a HTTP URL that's what you'll get back anyway).

Then you can instead of relying on URLConnection#setDoOutput(true) to implicitly set the request method to POST instead do httpURLConnection.setRequestMethod("POST") which some might find more natural (and which also allows you to specify other request methods such as PUT, DELETE, ...).

It also provides useful HTTP constants so you can do:

int responseCode = httpURLConnection.getResponseCode();

if (responseCode == HttpURLConnection.HTTP_OK) {

Inspired by this and other questions on SO, I've created a minimal open source basic-http-client that embodies most of the techniques found here.

google-http-java-client is also a great open source resource.

上一篇: 为移动应用程序创建一个API

下一篇: 如何使用java.net.URLConnection来触发和处理HTTP请求