An intranet site has a search form which uses AJAX to call a servlet on a different domain for search suggestions.
This works in Internet Explorer with the intranet domain being a “trusted site” and with cross-domain requests enabled for trusted sites, but doesn’t work in Firefox.
I have tried to work around the problem by creating a servlet on the intranet server, so there’s a JS call to my servlet on the same domain, then my servlet calls the suggestions servlet on the other domain. The cross-domain call is server-side, so it should work regardless of browser settings.
The AJAX call and my servlet’s call to the other servlet both use a HTTP POST request with arguments in the URL and empty request-content.
The reason I’m sticking with POST requests is that the JS code is all in files on the search server, which I can’t modify, and that code uses POST requests.
I’ve tried calling the customer’s existing suggestions servlet with a GET request, and it produces a 404 error.
The problem is that the result is inconsistent.
I’ve used System.out.println calls to show the full URL and size of the result on the server log.
The output first seemed to change depending on the calling browser and/or website, but now seems to change even between sessions of the same browser.
E.g. entering “g” in the search box, I got this output from the first few tries on the Development environment using Firefox:
Search suggestion URL: http://searchdev.companyname.com.au/suggest?q=g&max=10&site=All&client=ie&access=p&format=rich
Search suggestion result length: 64
Initial tries with Firefox on the Test environment (different intranet server but same search server) produced a result length of 0 for the same search URL.
Initial tries with Internet Explorer produced a result length of 0 in both environments.
Then I tried searching for different letters, and found that “t” produced a result in IE when “g” hadn’t.
After closing the browsers and leaving it for a while, I tried again and got different results.
E.g. Using Firefox and trying “g” in the Development environment now produces no result when it was previously producing one.
The inconsistency makes me think something is wrong with my servlet code, which is shown below. What could be causing the problem?
I think the search suggestions are being provided by a Google Search Appliance, and the JS files on the search server all seem to have come from Google.
The actual AJAX call is this line in one file:
XH_XmlHttpPOST(xmlhttp, url, '', handler);
The XH_XmlHttpPOST function is as follows in another file:
function XH_XmlHttpPOST(xmlHttp, url, data, handler) {
xmlHttp.open("POST", url, true);
xmlHttp.onreadystatechange = handler;
xmlHttp.setRequestHeader("Content-Type", "application/x-www-form-urlencoded");
xmlHttp.setRequestHeader("Content-Length",
/** @type {string} */ (data.length));
XH_XmlHttpSend(xmlHttp, data);
}
Here is my servlet code:
package com.companyname.theme;
import java.io.IOException;
import java.io.InputStream;
import java.io.OutputStream;
import java.io.PrintWriter;
import java.net.HttpURLConnection;
import java.net.MalformedURLException;
import java.net.URL;
import java.util.Properties;
import javax.servlet.ServletException;
import javax.servlet.http.HttpServlet;
import javax.servlet.http.HttpServletRequest;
import javax.servlet.http.HttpServletResponse;
public class suggest extends HttpServlet {
Properties props=null;
@Override
protected void doPost(HttpServletRequest req, HttpServletResponse resp)
throws ServletException, IOException {
String result = "";
String args = req.getQueryString();
String baseURL = props.getProperty("searchFormBaseURL");
String urlStr = baseURL + "/suggest?" + args;
System.out.println("Search suggestion URL: " + urlStr);
try {
int avail, rCount;
int totalCount = 0;
byte[] ba = null;
byte[] bCopy;
URL url = new URL(urlStr);
HttpURLConnection conn = (HttpURLConnection) url.openConnection();
conn.setRequestMethod("POST");
conn.setDoOutput(true);
OutputStream os = conn.getOutputStream();
os.write("".getBytes());
os.close();
InputStream is = conn.getInputStream();
while ((avail = is.available()) > 0) {
if (ba == null) ba = new byte[avail];
else if (totalCount + avail > ba.length) {
// Resize ba if there's more data available.
bCopy = new byte[totalCount + avail];
System.arraycopy(ba, 0, bCopy, 0, totalCount);
ba = bCopy;
bCopy = null;
}
rCount = is.read(ba, totalCount, avail);
if (rCount < 0) break;
totalCount += rCount;
}
is.close();
conn.disconnect();
result = (ba == null ? "" : new String(ba));
System.out.println("Search suggestion result length: " + Integer.toString(result.length()));
} catch(MalformedURLException e) {
e.printStackTrace();
} catch(IOException e) {
e.printStackTrace();
}
PrintWriter pw = resp.getWriter();
pw.print(result);
}
@Override
public void init() throws ServletException {
super.init();
InputStream stream = this.getClass().getResourceAsStream("/WEB-INF/lib/endeavour.properties");
props = new Properties();
try {
props.load(stream);
stream.close();
} catch (Exception e) {
// TODO: handle exception
}
}
}
Solution: don’t rely on
InputStream.available().The JavaDoc for that method says it always returns 0.
HttpURLConnection.getInputStream()actually returns a HttpInputStream, in whichavailable()seems to work but apparently sometimes returns 0 when there is more data.I changed my read loop to not use
available()at all, and now it consistently returns the expected results.The working servlet is below.