private void buildUrlSets(String url) {
try {
outputMessage("\nFetching " + url, TEST_SUMMARY_MESSAGE);
URL srcUrl = new URL(url);
// URLConnection conn = srcUrl.openConnection();
// String type = conn.getContentType();
// type = conn.getHeaderField("content-type");
// InputStream istr = conn.getInputStream();
LockssUrlConnection conn = UrlUtil.openConnection(url, connectionPool);
if (proxyHost != null) {
conn.setProxy(proxyHost, proxyPort);
}
if (userAgent != null) {
conn.setRequestProperty("user-agent", userAgent);
}
try {
conn.execute();
int resp = conn.getResponseCode();
if (resp != 200) {
outputMessage("Resp: " + resp + ": " + conn.getResponseMessage(), TEST_SUMMARY_MESSAGE);
return;
}
depth_fetched[m_curDepth - 1]++;
String cookies = conn.getResponseHeaderValue("Set-Cookie");
if (cookies != null) {
outputMessage("Cookies: " + cookies, PLAIN_MESSAGE);
}
String type = conn.getResponseContentType();
if (type == null || !type.toLowerCase().startsWith("text/html")) {
outputMessage("Type: " + type + ", not parsing", URL_SUMMARY_MESSAGE);
return;
}
outputMessage("Type: " + type + ", extracting Urls", URL_SUMMARY_MESSAGE);
InputStream istr = conn.getResponseInputStream();
InputStreamReader reader = new InputStreamReader(istr);
// MyMockCachedUrl mcu = new MyMockCachedUrl(srcUrl.toString(), reader);
GoslingHtmlLinkExtractor extractor = new GoslingHtmlLinkExtractor();
extractor.extractUrls(null, istr, null, srcUrl.toString(), new MyLinkExtractorCallback());
istr.close();
depth_parsed[m_curDepth - 1]++;
} finally {
conn.release();
}
} catch (MalformedURLException murle) {
murle.printStackTrace();
outputErrResults(url, "Malformed URL:" + murle.getMessage());
} catch (IOException ex) {
ex.printStackTrace();
outputErrResults(url, "IOException: " + ex.getMessage());
}
}
private ExternalizableMap loadMap(String extMapName, ClassLoader loader)
throws FileNotFoundException {
String first = null;
String next = extMapName;
List<String> urls = new ArrayList<String>();
ExternalizableMap res = null;
while (next != null) {
// convert the plugin class name to an xml file name
String mapFile = next.replace('.', '/') + MAP_SUFFIX;
URL url = loader.getResource(mapFile);
if (url != null && urls.contains(url.toString())) {
throw new PluginException.InvalidDefinition("Plugin inheritance loop: " + next);
}
// load into map
ExternalizableMap oneMap = new ExternalizableMap();
oneMap.loadMapFromResource(mapFile, loader);
urls.add(url.toString());
// apply overrides one plugin at a time in inheritance chain
processOverrides(oneMap);
if (res == null) {
res = oneMap;
} else {
for (Map.Entry ent : oneMap.entrySet()) {
String key = (String) ent.getKey();
Object val = ent.getValue();
if (!res.containsKey(key)) {
res.setMapElement(key, val);
}
}
}
if (oneMap.containsKey(KEY_PLUGIN_PARENT)) {
next = oneMap.getString(KEY_PLUGIN_PARENT);
} else {
next = null;
}
}
loadedFromUrls = urls;
return res;
}
/**
* mapUrlToFileLocation() is the method used to resolve urls into file names. This maps a given
* url to a file location, using the au top directory as the base. It creates directories which
* mirror the html string, so 'http://www.journal.org/issue1/index.html' would be cached in the
* file: <rootLocation>/www.journal.org/http/issue1/index.html
*
* @param rootLocation the top directory for ArchivalUnit this URL is in
* @param urlStr the url to translate
* @return the url file location
* @throws java.net.MalformedURLException
*/
public static String mapUrlToFileLocation(String rootLocation, String urlStr)
throws MalformedURLException {
int totalLength = rootLocation.length() + urlStr.length();
URL url = new URL(urlStr);
StringBuilder buffer = new StringBuilder(totalLength);
buffer.append(rootLocation);
if (!rootLocation.endsWith(File.separator)) {
buffer.append(File.separator);
}
buffer.append(url.getHost().toLowerCase());
int port = url.getPort();
if (port != -1) {
buffer.append(PORT_SEPARATOR);
buffer.append(port);
}
buffer.append(File.separator);
buffer.append(url.getProtocol());
if (RepositoryManager.isEnableLongComponents()) {
String escapedPath =
escapePath(
StringUtil.replaceString(url.getPath(), UrlUtil.URL_PATH_SEPARATOR, File.separator));
String query = url.getQuery();
if (query != null) {
escapedPath = escapedPath + "?" + escapeQuery(query);
}
String encodedPath = RepositoryNodeImpl.encodeUrl(escapedPath);
// encodeUrl strips leading / from path
buffer.append(File.separator);
buffer.append(encodedPath);
} else {
buffer.append(
escapePath(
StringUtil.replaceString(url.getPath(), UrlUtil.URL_PATH_SEPARATOR, File.separator)));
String query = url.getQuery();
if (query != null) {
buffer.append("?");
buffer.append(escapeQuery(query));
}
}
return buffer.toString();
}
