You can not select more than 25 topics
Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
1174 lines
59 KiB
1174 lines
59 KiB
// yacy.java
|
|
// -----------------------
|
|
// (C) by Michael Peter Christen; mc@anomic.de
|
|
// first published on http://www.yacy.net
|
|
// Frankfurt, Germany, 2004, 2005
|
|
//
|
|
// $LastChangedDate$
|
|
// $LastChangedRevision$
|
|
// $LastChangedBy$
|
|
//
|
|
// This program is free software; you can redistribute it and/or modify
|
|
// it under the terms of the GNU General Public License as published by
|
|
// the Free Software Foundation; either version 2 of the License, or
|
|
// (at your option) any later version.
|
|
//
|
|
// This program is distributed in the hope that it will be useful,
|
|
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
|
// GNU General Public License for more details.
|
|
//
|
|
// You should have received a copy of the GNU General Public License
|
|
// along with this program; if not, write to the Free Software
|
|
// Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
|
|
//
|
|
// Using this software in any meaning (reading, learning, copying, compiling,
|
|
// running) means that you agree that the Author(s) is (are) not responsible
|
|
// for cost, loss of data or any harm that may be caused directly or indirectly
|
|
// by usage of this softare or this documentation. The usage of this software
|
|
// is on your own risk. The installation and usage (starting/running) of this
|
|
// software may allow other people or application to access your computer and
|
|
// any attached devices and is highly dependent on the configuration of the
|
|
// software which must be done by the user of the software; the author(s) is
|
|
// (are) also not responsible for proper configuration and usage of the
|
|
// software, even if provoked by documentation provided together with
|
|
// the software.
|
|
//
|
|
// Any changes to this file according to the GPL as documented in the file
|
|
// gpl.txt aside this file in the shipment you received can be done to the
|
|
// lines that follows this copyright notice here, but changes must not be
|
|
// done inside the copyright notive above. A re-distribution must contain
|
|
// the intact and unchanged copyright notice.
|
|
// Contributions and changes to the program code must be marked as such.
|
|
import java.io.BufferedReader;
|
|
import java.io.BufferedWriter;
|
|
import java.io.ByteArrayOutputStream;
|
|
import java.io.File;
|
|
import java.io.FileInputStream;
|
|
import java.io.FileNotFoundException;
|
|
import java.io.FileWriter;
|
|
import java.io.IOException;
|
|
import java.io.InputStreamReader;
|
|
import java.io.PrintWriter;
|
|
import java.net.URL;
|
|
import java.util.HashMap;
|
|
import java.util.HashSet;
|
|
import java.util.Iterator;
|
|
import java.util.Properties;
|
|
import java.util.TreeSet;
|
|
import java.util.regex.Matcher;
|
|
import java.util.regex.Pattern;
|
|
|
|
import de.anomic.data.translator;
|
|
import de.anomic.http.httpHeader;
|
|
import de.anomic.http.httpc;
|
|
import de.anomic.http.httpd;
|
|
import de.anomic.http.httpdFileHandler;
|
|
import de.anomic.http.httpdProxyHandler;
|
|
import de.anomic.kelondro.kelondroMScoreCluster;
|
|
import de.anomic.plasma.plasmaCrawlLURL;
|
|
import de.anomic.plasma.plasmaSwitchboard;
|
|
import de.anomic.plasma.plasmaURL;
|
|
import de.anomic.plasma.plasmaWordIndex;
|
|
import de.anomic.plasma.plasmaWordIndexEntity;
|
|
import de.anomic.plasma.plasmaWordIndexEntry;
|
|
import de.anomic.plasma.plasmaWordIndexClassicDB;
|
|
import de.anomic.plasma.plasmaWordIndexCache;
|
|
import de.anomic.plasma.plasmaWordIndexEntryContainer;
|
|
import de.anomic.server.serverCore;
|
|
import de.anomic.server.serverDate;
|
|
import de.anomic.server.serverCodings;
|
|
import de.anomic.server.serverFileUtils;
|
|
import de.anomic.server.serverSystem;
|
|
import de.anomic.server.logging.serverLog;
|
|
import de.anomic.tools.enumerateFiles;
|
|
import de.anomic.yacy.yacyCore;
|
|
|
|
/**
|
|
* This is the main class of the proxy. Several threads are started from here:
|
|
* <ul>
|
|
* <li>one single instance of the plasmaSwitchboard is generated, which itself
|
|
* starts a thread with a plasmaHTMLCache object. This object simply counts
|
|
* files sizes in the cache and terminates them. It also generates a
|
|
* plasmaCrawlerLoader object, which may itself start some more httpc-calling
|
|
* threads to load web pages. They terminate automatically when a page has
|
|
* loaded.
|
|
* <li>one serverCore - thread is started, which implements a multi-threaded
|
|
* server. The process may start itself many more processes that handle
|
|
* connections.
|
|
* <li>finally, all idle-dependent processes are written in a queue in
|
|
* plasmaSwitchboard which are worked off inside an idle-sensitive loop of the
|
|
* main process. (here)
|
|
* </ul>
|
|
*
|
|
* On termination, the following must be done:
|
|
* <ul>
|
|
* <li>stop feeding of the crawling process because it othervise fills the
|
|
* indexing queue.
|
|
* <li>say goodbye to connected peers and disable new connections. Don't wait for
|
|
* success.
|
|
* <li>first terminate the serverCore thread. This prevents that new cache
|
|
* objects are queued.
|
|
* <li>wait that the plasmaHTMLCache terminates (it should be normal that this
|
|
* process already has terminated).
|
|
* <li>then wait for termination of all loader process of the
|
|
* plasmaCrawlerLoader.
|
|
* <li>work off the indexing and cache storage queue. These values are inside a
|
|
* RAM cache and would be lost otherwise.
|
|
* <li>write all settings.
|
|
* <li>terminate.
|
|
* </ul>
|
|
*/
|
|
|
|
public final class yacy {
|
|
// static objects
|
|
private static String vString = "@REPL_VERSION@";
|
|
private static float version = (float) 0.1;
|
|
|
|
private static final String vDATE = "@REPL_DATE@";
|
|
private static final String copyright = "[ YACY Proxy v" + vString + ", build " + vDATE + " by Michael Christen / www.yacy.net ]";
|
|
private static final String hline = "-------------------------------------------------------------------------------";
|
|
|
|
/**
|
|
* Convert the combined versionstring into a pretty string.
|
|
* FIXME: Why is this so complicated?
|
|
*
|
|
* @param s Combined version string
|
|
* @return Pretty string where version and svn-Version are separated by a
|
|
* slash
|
|
*/
|
|
public static String combinedVersionString2PrettyString(String s) {
|
|
long svn;
|
|
try {svn = (long) (100000000.0 * Double.parseDouble(s));} catch (NumberFormatException ee) {svn = 0;}
|
|
double version = (Math.floor((double) svn / (double) 100000) / (double) 1000);
|
|
String vStr = (version < 0.11) ? "dev" : Double.toString(version);
|
|
//while (vStr.length() < 5) vStr = vStr + "0";
|
|
svn = svn % 100000;
|
|
if (svn > 4000) svn=svn / 10; // fix a previous bug online
|
|
String svnStr = Long.toString(svn);
|
|
while (svnStr.length() < 5) svnStr = "0" + svnStr;
|
|
return vStr + "/" + svnStr;
|
|
}
|
|
|
|
/**
|
|
* Combines the version of the proxy with the versionnumber from SVN to a
|
|
* combined version
|
|
*
|
|
* @param version Current given version for this proxy.
|
|
* @param svn Current version given from svn.
|
|
* @return String with the combined version
|
|
*/
|
|
public static float versvn2combinedVersion(float version, int svn) {
|
|
return (float) (((double) version * 100000000.0 + ((double) svn)) / 100000000.0);
|
|
}
|
|
|
|
/**
|
|
* Starts up the whole application. Sets up all datastructures and starts
|
|
* the main threads.
|
|
*
|
|
* @param homePath Root-path where all information is to be found.
|
|
* @param startupFree free memory at startup time, to be used later for statistics
|
|
*/
|
|
private static void startup(String homePath, long startupMemFree, long startupMemTotal) {
|
|
long startup = System.currentTimeMillis();
|
|
String restart = "false";
|
|
|
|
try {
|
|
// start up
|
|
System.out.println(copyright);
|
|
System.out.println(hline);
|
|
|
|
// check java version
|
|
try {
|
|
String[] check = "a,b".split(","); // split needs java 1.4
|
|
} catch (NoSuchMethodError e) {
|
|
System.err.println("STARTUP: Java Version too low. You need at least Java 1.4.2 to run YACY");
|
|
Thread.currentThread().sleep(3000);
|
|
System.exit(-1);
|
|
}
|
|
|
|
// ensure that there is a DATA directory
|
|
File f = new File(homePath); if (!(f.exists())) f.mkdirs();
|
|
f = new File(homePath, "DATA/"); if (!(f.exists())) f.mkdirs();
|
|
|
|
// setting up logging
|
|
f = new File(homePath, "DATA/LOG/"); if (!(f.exists())) f.mkdirs();
|
|
if (!((new File(homePath, "DATA/LOG/yacy.logging")).exists())) try {
|
|
serverFileUtils.copy(new File(homePath, "yacy.logging"), new File(homePath, "DATA/LOG/yacy.logging"));
|
|
}catch (IOException e){
|
|
System.out.println("could not copy yacy.logging");
|
|
}
|
|
try{
|
|
serverLog.configureLogging(new File(homePath, "DATA/LOG/yacy.logging"));
|
|
} catch (IOException e) {
|
|
System.out.println("could not find logging properties in homePath=" + homePath);
|
|
e.printStackTrace();
|
|
}
|
|
serverLog.logConfig("STARTUP", copyright);
|
|
serverLog.logConfig("STARTUP", hline);
|
|
serverLog.logConfig("STARTUP", "java version " + System.getProperty("java.version", "no-java-version"));
|
|
serverLog.logConfig("STARTUP", "Application Root Path: " + homePath);
|
|
serverLog.logConfig("STARTUP", "Time Zone: UTC" + serverDate.UTCDiffString() + "; UTC+0000 is " + System.currentTimeMillis());
|
|
|
|
// create data folder
|
|
final File dataFolder = new File(homePath, "DATA");
|
|
if (!(dataFolder.exists())) dataFolder.mkdir();
|
|
|
|
final plasmaSwitchboard sb = new plasmaSwitchboard(homePath, "yacy.init", "DATA/SETTINGS/httpProxy.conf");
|
|
|
|
// save information about available memory at startup time
|
|
sb.setConfig("memoryFreeAfterStartup", startupMemFree);
|
|
sb.setConfig("memoryTotalAfterStartup", startupMemTotal);
|
|
|
|
// hardcoded, forced, temporary value-migration
|
|
sb.setConfig("htTemplatePath", "htroot/env/templates");
|
|
sb.setConfig("parseableExt", "html,htm,txt,php,shtml,asp");
|
|
|
|
// set default = no restart
|
|
sb.setConfig("restart", "false");
|
|
|
|
// if we are running an SVN version, we try to detect the used svn revision now ...
|
|
final Properties buildProp = new Properties();
|
|
File buildPropFile = null;
|
|
try {
|
|
buildPropFile = new File(homePath,"build.properties");
|
|
buildProp.load(new FileInputStream(buildPropFile));
|
|
} catch (Exception e) {
|
|
serverLog.logWarning("STARTUP", buildPropFile.toString() + " not found in settings path");
|
|
}
|
|
|
|
try {
|
|
if (buildProp.containsKey("releaseNr")) {
|
|
// this normally looks like this: $Revision$
|
|
final String svnReleaseNrStr = buildProp.getProperty("releaseNr");
|
|
final Pattern pattern = Pattern.compile("\\$Revision:\\s(.*)\\s\\$",Pattern.DOTALL+Pattern.CASE_INSENSITIVE);
|
|
final Matcher matcher = pattern.matcher(svnReleaseNrStr);
|
|
if (matcher.find()) {
|
|
final String svrReleaseNr = matcher.group(1);
|
|
try {
|
|
try {version = Float.parseFloat(vString);} catch (NumberFormatException e) {version = (float) 0.1;}
|
|
version = versvn2combinedVersion(version, Integer.parseInt(svrReleaseNr));
|
|
} catch (NumberFormatException e) {}
|
|
sb.setConfig("svnRevision", svrReleaseNr);
|
|
}
|
|
}
|
|
} catch (Exception e) {
|
|
System.err.println("Unable to determine the currently used SVN revision number.");
|
|
}
|
|
|
|
sb.setConfig("version", Float.toString(version));
|
|
sb.setConfig("vString", combinedVersionString2PrettyString(Float.toString(version)));
|
|
sb.setConfig("vdate", vDATE);
|
|
sb.setConfig("applicationRoot", homePath);
|
|
sb.setConfig("startupTime", Long.toString(startup));
|
|
serverLog.logConfig("STARTUP", "YACY Version: " + version + ", Built " + vDATE);
|
|
yacyCore.latestVersion = (float) version;
|
|
|
|
// read environment
|
|
//new
|
|
final int port = Integer.parseInt(sb.getConfig("port", "8080"));
|
|
int timeout = Integer.parseInt(sb.getConfig("httpdTimeout", "60000"));
|
|
if (timeout < 60000) timeout = 60000;
|
|
|
|
// create some directories
|
|
final File htRootPath = new File(homePath, sb.getConfig("htRootPath", "htroot"));
|
|
final File htDocsPath = new File(homePath, sb.getConfig("htDocsPath", "DATA/HTDOCS"));
|
|
File htTemplatePath = new File(homePath, sb.getConfig("htTemplatePath","htdocs"));
|
|
|
|
// create default notifier picture
|
|
if (!((new File(htRootPath, "env/grafics/notifier.gif")).exists())) try {
|
|
serverFileUtils.copy(new File(htRootPath, "env/grafics/empty.gif"),
|
|
new File(htRootPath, "env/grafics/notifier.gif"));
|
|
} catch (IOException e) {}
|
|
|
|
if (!(htDocsPath.exists())) htDocsPath.mkdir();
|
|
final File htdocsDefaultReadme = new File(htDocsPath, "readme.txt");
|
|
if (!(htdocsDefaultReadme.exists())) try {serverFileUtils.write((
|
|
"This is your root directory for individual Web Content\r\n" +
|
|
"\r\n" +
|
|
"Please place your html files into the www subdirectory.\r\n" +
|
|
"The URL of that path is either\r\n" +
|
|
"http://www.<your-peer-name>.yacy or\r\n" +
|
|
"http://<your-ip>:<your-port>/www\r\n" +
|
|
"\r\n" +
|
|
"Other subdirectories may be created; they map to corresponding sub-domains.\r\n" +
|
|
"This directory shares it's content with the applications htroot path, so you\r\n" +
|
|
"may access your yacy search page with\r\n" +
|
|
"http://<your-peer-name>.yacy/\r\n" +
|
|
"\r\n").getBytes(), htdocsDefaultReadme);} catch (IOException e) {
|
|
System.out.println("Error creating htdocs readme: " + e.getMessage());
|
|
}
|
|
|
|
final File wwwDefaultPath = new File(htDocsPath, "www");
|
|
if (!(wwwDefaultPath.exists())) wwwDefaultPath.mkdir();
|
|
|
|
final File wwwDefaultClass = new File(wwwDefaultPath, "welcome.class");
|
|
//if ((!(wwwDefaultClass.exists())) || (wwwDefaultClass.length() != (new File(htRootPath, "htdocsdefault/welcome.class")).length())) try {
|
|
if((new File(htRootPath, "htdocsdefault/welcome.java")).exists())
|
|
serverFileUtils.copy(new File(htRootPath, "htdocsdefault/welcome.java"), new File(wwwDefaultPath, "welcome.java"));
|
|
serverFileUtils.copy(new File(htRootPath, "htdocsdefault/welcome.class"), wwwDefaultClass);
|
|
serverFileUtils.copy(new File(htRootPath, "htdocsdefault/welcome.html"), new File(wwwDefaultPath, "welcome.html"));
|
|
//} catch (IOException e) {}
|
|
|
|
final File shareDefaultPath = new File(htDocsPath, "share");
|
|
if (!(shareDefaultPath.exists())) shareDefaultPath.mkdir();
|
|
|
|
final File shareDefaultClass = new File(shareDefaultPath, "dir.class");
|
|
//if ((!(shareDefaultClass.exists())) || (shareDefaultClass.length() != (new File(htRootPath, "htdocsdefault/dir.class")).length())) try {
|
|
if((new File(htRootPath, "htdocsdefault/dir.java")).exists())
|
|
serverFileUtils.copy(new File(htRootPath, "htdocsdefault/dir.java"), new File(shareDefaultPath, "dir.java"));
|
|
serverFileUtils.copy(new File(htRootPath, "htdocsdefault/dir.class"), shareDefaultClass);
|
|
serverFileUtils.copy(new File(htRootPath, "htdocsdefault/dir.html"), new File(shareDefaultPath, "dir.html"));
|
|
//} catch (IOException e) {}
|
|
|
|
|
|
// set preset accounts/passwords
|
|
String acc;
|
|
if ((acc = sb.getConfig("proxyAccount", "")).length() > 0) {
|
|
sb.setConfig("proxyAccountBase64MD5", de.anomic.server.serverCodings.encodeMD5Hex(serverCodings.standardCoder.encodeBase64String(acc)));
|
|
sb.setConfig("proxyAccount", "");
|
|
}
|
|
if ((acc = sb.getConfig("serverAccount", "")).length() > 0) {
|
|
sb.setConfig("serverAccountBase64MD5", de.anomic.server.serverCodings.encodeMD5Hex(serverCodings.standardCoder.encodeBase64String(acc)));
|
|
sb.setConfig("serverAccount", "");
|
|
}
|
|
if ((acc = sb.getConfig("adminAccount", "")).length() > 0) {
|
|
sb.setConfig("adminAccountBase64MD5", de.anomic.server.serverCodings.encodeMD5Hex(serverCodings.standardCoder.encodeBase64String(acc)));
|
|
sb.setConfig("adminAccount", "");
|
|
}
|
|
|
|
// fix unsafe old passwords
|
|
if ((acc = sb.getConfig("proxyAccountBase64", "")).length() > 0) {
|
|
sb.setConfig("proxyAccountBase64MD5", de.anomic.server.serverCodings.encodeMD5Hex(acc));
|
|
sb.setConfig("proxyAccountBase64", "");
|
|
}
|
|
if ((acc = sb.getConfig("serverAccountBase64", "")).length() > 0) {
|
|
sb.setConfig("serverAccountBase64MD5", de.anomic.server.serverCodings.encodeMD5Hex(acc));
|
|
sb.setConfig("serverAccountBase64", "");
|
|
}
|
|
if ((acc = sb.getConfig("adminAccountBase64", "")).length() > 0) {
|
|
sb.setConfig("adminAccountBase64MD5", de.anomic.server.serverCodings.encodeMD5Hex(acc));
|
|
sb.setConfig("adminAccountBase64", "");
|
|
}
|
|
if ((acc = sb.getConfig("uploadAccountBase64", "")).length() > 0) {
|
|
sb.setConfig("uploadAccountBase64MD5", de.anomic.server.serverCodings.encodeMD5Hex(acc));
|
|
sb.setConfig("uploadAccountBase64", "");
|
|
}
|
|
if ((acc = sb.getConfig("downloadAccountBase64", "")).length() > 0) {
|
|
sb.setConfig("downloadAccountBase64MD5", de.anomic.server.serverCodings.encodeMD5Hex(acc));
|
|
sb.setConfig("downloadAccountBase64", "");
|
|
}
|
|
|
|
// start main threads
|
|
try {
|
|
final httpd protocolHandler = new httpd(sb, new httpdFileHandler(sb), new httpdProxyHandler(sb));
|
|
final serverCore server = new serverCore(port,
|
|
timeout /*control socket timeout in milliseconds*/,
|
|
true /* block attacks (wrong protocol) */,
|
|
protocolHandler /*command class*/,
|
|
sb,
|
|
30000 /*command max length incl. GET args*/);
|
|
server.setName("httpd:"+port);
|
|
server.setPriority(Thread.MAX_PRIORITY);
|
|
server.setObeyIntermission(false);
|
|
if (server == null) {
|
|
serverLog.logSevere("STARTUP", "Failed to start server. Probably port " + port + " already in use.");
|
|
} else {
|
|
// first start the server
|
|
sb.deployThread("10_httpd", "HTTPD Server/Proxy", "the HTTPD, used as web server and proxy", null, server, 0, 0, 0, 0);
|
|
//server.start();
|
|
|
|
// open the browser window
|
|
final boolean browserPopUpTrigger = sb.getConfig("browserPopUpTrigger", "true").equals("true");
|
|
if (browserPopUpTrigger) {
|
|
final String browserPopUpPage = sb.getConfig("browserPopUpPage", "Status.html");
|
|
final String browserPopUpApplication = sb.getConfig("browserPopUpApplication", "netscape");
|
|
serverSystem.openBrowser("http://localhost:" + port + "/" + browserPopUpPage, browserPopUpApplication);
|
|
}
|
|
|
|
//Copy the shipped locales into DATA
|
|
final File localesPath = new File(homePath, sb.getConfig("localesPath", "DATA/LOCALE"));
|
|
final File defaultLocalesPath = new File(homePath, "locales");
|
|
|
|
try{
|
|
final File[] defaultLocales = defaultLocalesPath.listFiles();
|
|
localesPath.mkdirs();
|
|
for(int i=0;i < defaultLocales.length; i++){
|
|
if(defaultLocales[i].getName().endsWith(".lng"))
|
|
serverFileUtils.copy(defaultLocales[i], new File(localesPath, defaultLocales[i].getName()));
|
|
}
|
|
serverLog.logInfo("STARTUP", "Copied the default lokales to DATA/LOCALE");
|
|
}catch(NullPointerException e){
|
|
serverLog.logSevere("STARTUP", "Nullpointer Exception while copying the default Locales");
|
|
}
|
|
|
|
//regenerate Locales from Translationlist, if needed
|
|
final String lang = sb.getConfig("htLocaleSelection", "");
|
|
if(! lang.equals("") && ! lang.equals("default") ){ //locale is used
|
|
String currentRev = "";
|
|
try{
|
|
final BufferedReader br = new BufferedReader(new InputStreamReader(new FileInputStream(new File( sb.getConfig("htRootPath", "htroot"), "locale/"+lang+"/version" ))));
|
|
currentRev = br.readLine();
|
|
br.close();
|
|
}catch(IOException e){
|
|
//Error
|
|
}
|
|
|
|
try{ //seperate try, because we want this, even when the File "version2 does not exist.
|
|
if(! currentRev.equals(sb.getConfig("svnRevision", "")) ){ //is this another version?!
|
|
final File sourceDir = new File(sb.getConfig("htRootPath", "htroot"));
|
|
final File destDir = new File(sourceDir, "locale/"+lang);
|
|
if(translator.translateFiles(sourceDir, destDir, new File("DATA/LOCALE/"+lang+".lng"), "html")){ //translate it
|
|
//if(translator.translateFilesRecursive(sourceDir, destDir, new File("DATA/LOCALE/"+lang+".lng"), "html")){ //translate it
|
|
//write the new Versionnumber
|
|
final BufferedWriter bw = new BufferedWriter(new PrintWriter(new FileWriter(new File(destDir, "version"))));
|
|
bw.write(sb.getConfig("svnRevision", "Error getting Version"));
|
|
bw.close();
|
|
}
|
|
}
|
|
}catch(IOException e){
|
|
//Error
|
|
}
|
|
}
|
|
|
|
// registering shutdown hook
|
|
serverLog.logConfig("STARTUP", "Registering Shutdown Hook");
|
|
final Runtime run = Runtime.getRuntime();
|
|
run.addShutdownHook(new shutdownHookThread(Thread.currentThread(), sb));
|
|
|
|
// save information about available memory after all initializations
|
|
sb.setConfig("memoryFreeAfterInitBGC", Runtime.getRuntime().freeMemory());
|
|
sb.setConfig("memoryTotalAfterInitBGC", Runtime.getRuntime().totalMemory());
|
|
System.gc();
|
|
sb.setConfig("memoryFreeAfterInitAGC", Runtime.getRuntime().freeMemory());
|
|
sb.setConfig("memoryTotalAfterInitAGC", Runtime.getRuntime().totalMemory());
|
|
|
|
// wait for server shutdown
|
|
try {
|
|
sb.waitForShutdown();
|
|
} catch (Exception e) {
|
|
serverLog.logSevere("MAIN CONTROL LOOP", "PANIK: " + e.getMessage(),e);
|
|
}
|
|
restart = sb.getConfig("restart", "false");
|
|
|
|
// shut down
|
|
serverLog.logConfig("SHUTDOWN", "caught termination signal");
|
|
server.terminate(false);
|
|
server.interrupt();
|
|
if (server.isAlive()) try {
|
|
httpc.wget(new URL("http://localhost:" + port), 1000, null, null, null); // kick server
|
|
serverLog.logConfig("SHUTDOWN", "sent termination signal to server socket");
|
|
} catch (IOException ee) {
|
|
serverLog.logConfig("SHUTDOWN", "termination signal to server socket missed (server shutdown, ok)");
|
|
}
|
|
|
|
// idle until the processes are down
|
|
while (server.isAlive()) {
|
|
Thread.currentThread().sleep(2000); // wait a while
|
|
}
|
|
serverLog.logConfig("SHUTDOWN", "server has terminated");
|
|
sb.close();
|
|
}
|
|
} catch (Exception e) {
|
|
serverLog.logSevere("STARTUP", "Unexpected Error: " + e.getClass().getName(),e);
|
|
//System.exit(1);
|
|
}
|
|
} catch (Exception ee) {
|
|
serverLog.logSevere("STARTUP", "FATAL ERROR: " + ee.getMessage(),ee);
|
|
}
|
|
|
|
// restart YaCy
|
|
if (restart.equals("true")) {
|
|
serverLog.logConfig("SHUTDOWN", "RESTART...");
|
|
long count = 0;
|
|
if (Thread.activeCount() > 1) {
|
|
serverLog.logConfig("SHUTDOWN", "Wait maximally 5 minutes for " + (Thread.activeCount() - 1) + " running threads to restart YaCy");
|
|
while (Thread.activeCount() > 1 && count <= 60) { // wait 5 minutes
|
|
count++;
|
|
try { Thread.currentThread().sleep(5000); } catch (InterruptedException e) {}
|
|
}
|
|
}
|
|
if (count < 60) {
|
|
System.gc();
|
|
try { Thread.currentThread().sleep(5000); } catch (InterruptedException e) {}
|
|
startupMemFree = Runtime.getRuntime().freeMemory(); // the amount of free memory in the Java Virtual Machine
|
|
startupMemTotal = Runtime.getRuntime().totalMemory(); // the total amount of memory in the Java virtual machine; may vary over time
|
|
startup(homePath, startupMemFree, startupMemTotal);
|
|
} else {
|
|
serverLog.logConfig("SHUTDOWN", "RESTART BREAK, more than 5 minutes waited to try a restart, goodbye. (this is the last line)");
|
|
// serverLog.logConfig("SHUTDOWN", "RESTART BREAK, getAllStackTraces()\n" + Thread.getAllStackTraces()); // needs java 1.5
|
|
}
|
|
} else {
|
|
serverLog.logConfig("SHUTDOWN", "goodbye. (this is the last line)");
|
|
}
|
|
try {
|
|
System.exit(0);
|
|
} catch (Exception e) {} // was once stopped by de.anomic.net.ftpc$sm.checkExit(ftpc.java:1790)
|
|
}
|
|
|
|
/**
|
|
* Loads the configuration from the data-folder.
|
|
* FIXME: Why is this called over and over again from every method, instead
|
|
* of setting the configurationdata once for this class in main?
|
|
*
|
|
* @param mes Where are we called from, so that the errormessages can be
|
|
* more descriptive.
|
|
* @param homePath Root-path where all the information is to be found.
|
|
* @return Properties read from the configurationfile.
|
|
*/
|
|
private static Properties configuration(String mes, String homePath) {
|
|
serverLog.logConfig(mes, "Application Root Path: " + homePath.toString());
|
|
|
|
// read data folder
|
|
File dataFolder = new File(homePath, "DATA");
|
|
if (!(dataFolder.exists())) {
|
|
serverLog.logSevere(mes, "Application was never started or root path wrong.");
|
|
System.exit(-1);
|
|
}
|
|
|
|
Properties config = new Properties();
|
|
try {
|
|
config.load(new FileInputStream(new File(homePath, "DATA/SETTINGS/httpProxy.conf")));
|
|
} catch (FileNotFoundException e) {
|
|
serverLog.logSevere(mes, "could not find configuration file.");
|
|
System.exit(-1);
|
|
} catch (IOException e) {
|
|
serverLog.logSevere(mes, "could not read configuration file.");
|
|
System.exit(-1);
|
|
}
|
|
|
|
return config;
|
|
}
|
|
|
|
/**
|
|
* Call the shutdown-page from yacy to tell it to shut down. This method is
|
|
* called if you start yacy with the argument -shutdown.
|
|
*
|
|
* @param homePath Root-path where all the information is to be found.
|
|
*/
|
|
static void shutdown(String homePath) {
|
|
// start up
|
|
System.out.println(copyright);
|
|
System.out.println(hline);
|
|
|
|
Properties config = configuration("REMOTE-SHUTDOWN", homePath);
|
|
|
|
// read port
|
|
int port = Integer.parseInt((String) config.get("port"));
|
|
|
|
// read password
|
|
String encodedPassword = (String) config.get("adminAccountBase64MD5");
|
|
if (encodedPassword == null) encodedPassword = ""; // not defined
|
|
|
|
// send 'wget' to web interface
|
|
httpHeader requestHeader = new httpHeader();
|
|
requestHeader.put("Authorization", "realm=" + encodedPassword); // for http-authentify
|
|
try {
|
|
httpc con = httpc.getInstance("localhost", port, 10000, false);
|
|
httpc.response res = con.GET("Steering.html?shutdown=", requestHeader);
|
|
|
|
// read response
|
|
if (res.status.startsWith("2")) {
|
|
serverLog.logConfig("REMOTE-SHUTDOWN", "YACY accepted shutdown command.");
|
|
serverLog.logConfig("REMOTE-SHUTDOWN", "Stand by for termination, which may last some seconds.");
|
|
ByteArrayOutputStream bos = new ByteArrayOutputStream();
|
|
res.writeContent(bos, null);
|
|
con.close();
|
|
} else {
|
|
serverLog.logSevere("REMOTE-SHUTDOWN", "error response from YACY socket: " + res.status);
|
|
System.exit(-1);
|
|
}
|
|
} catch (IOException e) {
|
|
serverLog.logSevere("REMOTE-SHUTDOWN", "could not establish connection to YACY socket: " + e.getMessage());
|
|
System.exit(-1);
|
|
}
|
|
|
|
// finished
|
|
serverLog.logConfig("REMOTE-SHUTDOWN", "SUCCESSFULLY FINISHED remote-shutdown:");
|
|
serverLog.logConfig("REMOTE-SHUTDOWN", "YACY will terminate after working off all enqueued tasks.");
|
|
}
|
|
|
|
/**
|
|
* This method gets all found words and outputs a statistic about the score
|
|
* of the words. The output of this method can be used to create stop-word
|
|
* lists. This method will be called if you start yacy with the argument
|
|
* -genwordstat.
|
|
* FIXME: How can stop-word list be created from this output? What type of
|
|
* score is output?
|
|
*
|
|
* @param homePath Root-Path where all the information is to be found.
|
|
*/
|
|
private static void genWordstat(String homePath) {
|
|
// start up
|
|
System.out.println(copyright);
|
|
System.out.println(hline);
|
|
|
|
Properties config = configuration("GEN-WORDSTAT", homePath);
|
|
|
|
// load words
|
|
serverLog.logInfo("GEN-WORDSTAT", "loading words...");
|
|
HashMap words = loadWordMap(new File(homePath, "yacy.words"));
|
|
|
|
// find all hashes
|
|
serverLog.logInfo("GEN-WORDSTAT", "searching all word-hash databases...");
|
|
File dbRoot = new File(homePath, config.getProperty("dbPath"));
|
|
enumerateFiles ef = new enumerateFiles(new File(dbRoot, "WORDS"), true, false, true, true);
|
|
File f;
|
|
String h;
|
|
kelondroMScoreCluster hs = new kelondroMScoreCluster();
|
|
while (ef.hasMoreElements()) {
|
|
f = (File) ef.nextElement();
|
|
h = f.getName().substring(0, plasmaURL.urlHashLength);
|
|
hs.addScore(h, (int) f.length());
|
|
}
|
|
|
|
// list the hashes in reverse order
|
|
serverLog.logInfo("GEN-WORDSTAT", "listing words in reverse size order...");
|
|
String w;
|
|
Iterator i = hs.scores(false);
|
|
while (i.hasNext()) {
|
|
h = (String) i.next();
|
|
w = (String) words.get(h);
|
|
if (w == null) System.out.print("# " + h); else System.out.print(w);
|
|
System.out.println(" - " + hs.getScore(h));
|
|
}
|
|
|
|
// finished
|
|
serverLog.logConfig("GEN-WORDSTAT", "FINISHED");
|
|
}
|
|
|
|
/**
|
|
* Migrates the PLASMA WORDS structure to the assortment cache if possible.
|
|
* This method will be called if you start yacy with the argument
|
|
* -migratewords.
|
|
* Caution: This might take a long time to finish. Don't interrupt it!
|
|
* FIXME: Shouldn't this method be private?
|
|
*
|
|
* @param homePath Root-path where all the information is to be found.
|
|
*/
|
|
public static void migrateWords(String homePath) {
|
|
// run with "java -classpath classes yacy -migratewords"
|
|
try {serverLog.configureLogging(new File(homePath, "yacy.logging"));} catch (Exception e) {}
|
|
File dbroot = new File(new File(homePath), "DATA/PLASMADB");
|
|
try {
|
|
serverLog log = new serverLog("WORDMIGRATION");
|
|
log.logInfo("STARTING MIGRATION");
|
|
plasmaWordIndexCache wordIndexCache = new plasmaWordIndexCache(dbroot, new plasmaWordIndexClassicDB(dbroot, log), 20000, log);
|
|
enumerateFiles words = new enumerateFiles(new File(dbroot, "WORDS"), true, false, true, true);
|
|
String wordhash;
|
|
File wordfile;
|
|
int migration;
|
|
while (words.hasMoreElements()) try {
|
|
wordfile = (File) words.nextElement();
|
|
wordhash = wordfile.getName().substring(0, 12);
|
|
migration = wordIndexCache.migrateWords2Assortment(wordhash);
|
|
if (migration == 0)
|
|
log.logInfo("SKIPPED " + wordhash + ": " + ((wordfile.exists()) ? "too big" : "database corrupted; deleted"));
|
|
else if (migration > 0)
|
|
log.logInfo("MIGRATED " + wordhash + ": " + migration + " entries");
|
|
else
|
|
log.logInfo("REVERSED " + wordhash + ": " + (-migration) + " entries");
|
|
} catch (Exception e) {
|
|
e.printStackTrace();
|
|
}
|
|
log.logInfo("FINISHED MIGRATION JOB, WAIT FOR DUMP");
|
|
wordIndexCache.close(60);
|
|
log.logInfo("TERMINATED MIGRATION");
|
|
} catch (IOException e) {
|
|
e.printStackTrace();
|
|
}
|
|
}
|
|
|
|
public static void importDB(String homePath, String importPath) {
|
|
if (homePath == null) throw new NullPointerException();
|
|
if (importPath == null) throw new NullPointerException();
|
|
if (homePath.equals(importPath)) throw new IllegalArgumentException("Import and home DB directory must not be equal");
|
|
|
|
// configure logging
|
|
try {serverLog.configureLogging(new File(homePath, "yacy.logging"));} catch (Exception e) {}
|
|
serverLog log = new serverLog("DB-IMPORT");
|
|
log.logInfo("STARTING DB-IMPORT");
|
|
log.logInfo("Import can be aborted using <ctrl>+<c>");
|
|
|
|
plasmaWordIndex homeWordIndex = null, importWordIndex = null;
|
|
plasmaCrawlLURL homeUrlDB = null, importUrlDB = null;
|
|
try {
|
|
//
|
|
Runtime rt = Runtime.getRuntime();
|
|
String errorMsg = null;
|
|
|
|
// configure destination DB
|
|
File homeDBroot = new File(new File(homePath), "DATA/PLASMADB");
|
|
if (!homeDBroot.exists()) errorMsg = "Home DB directory does not exist.";
|
|
if (!homeDBroot.canRead()) errorMsg = "Home DB directory is not readable.";
|
|
if (!homeDBroot.canWrite()) errorMsg = "Home DB directory is not writeable";
|
|
if (!homeDBroot.isDirectory()) errorMsg = "Home DB Directory is not a directory.";
|
|
if (errorMsg != null) {
|
|
log.logSevere(errorMsg + "\nName: " + homeDBroot.getAbsolutePath());
|
|
return;
|
|
}
|
|
|
|
if ((!homeDBroot.exists())&&(!homeDBroot.canRead())&&(!homeDBroot.isDirectory())) {
|
|
log.logSevere("DB home directory can not be opened.");
|
|
return;
|
|
}
|
|
log.logFine("Initializing destination word index db.");
|
|
homeWordIndex = new plasmaWordIndex(homeDBroot, 8*1024*1024, log);
|
|
log.logFine("Initializing destination URL db.");
|
|
homeUrlDB = new plasmaCrawlLURL(new File(homeDBroot, "urlHash.db"), 4*1024*1024);
|
|
|
|
// configure import DB
|
|
errorMsg = null;
|
|
File importDBroot = new File(importPath);
|
|
if (!importDBroot.exists()) errorMsg = "Import directory does not exist.";
|
|
if (!importDBroot.canRead()) errorMsg = "Import directory is not readable.";
|
|
if (!importDBroot.canWrite()) errorMsg = "Import directory is not writeable";
|
|
if (!importDBroot.isDirectory()) errorMsg = "ImportDirectory is not a directory.";
|
|
if (errorMsg != null) {
|
|
log.logSevere(errorMsg + "\nName: " + homeDBroot.getAbsolutePath());
|
|
return;
|
|
}
|
|
|
|
log.logFine("Initializing source word index db.");
|
|
importWordIndex = new plasmaWordIndex(importDBroot, 8*1024*1024, log);
|
|
log.logFine("Initializing source URL db.");
|
|
importUrlDB = new plasmaCrawlLURL(new File(importDBroot, "urlHash.db"), 4*1024*1024);
|
|
int startSize = importWordIndex.size();
|
|
|
|
log.logInfo("Importing DB from '" + importDBroot.getAbsolutePath() + "' to '" + homeDBroot.getAbsolutePath() + "'.");
|
|
log.logInfo("Home word index contains " + homeWordIndex.size() + " words and " + homeUrlDB.size() + " URLs.");
|
|
log.logInfo("Import word index contains " + importWordIndex.size() + " words and " + importUrlDB.size() + " URLs.");
|
|
|
|
// iterate over all words from import db
|
|
String wordHash = "";
|
|
long urlCounter = 0, wordCounter = 0, entryCounter = 0;
|
|
long globalStart = System.currentTimeMillis(), wordChunkStart = System.currentTimeMillis(), wordChunkEnd = 0;
|
|
String wordChunkStartHash = "------------", wordChunkEndHash;
|
|
|
|
Iterator importWordHashIterator = importWordIndex.wordHashes(wordChunkStartHash, true, true);
|
|
while (importWordHashIterator.hasNext()) {
|
|
|
|
// testing if import process was aborted
|
|
if (Thread.interrupted()) break;
|
|
|
|
plasmaWordIndexEntity importWordIdxEntity = null;
|
|
try {
|
|
wordCounter++;
|
|
wordHash = (String) importWordHashIterator.next();
|
|
importWordIdxEntity = importWordIndex.getEntity(wordHash, true, -1);
|
|
|
|
if (importWordIdxEntity.size() == 0) {
|
|
importWordIdxEntity.deleteComplete();
|
|
continue;
|
|
}
|
|
|
|
// creating a container used to hold the imported entries
|
|
plasmaWordIndexEntryContainer newContainer = new plasmaWordIndexEntryContainer(wordHash,importWordIdxEntity.size());
|
|
|
|
// the combined container will fit, read the container
|
|
Iterator importWordIdxEntries = importWordIdxEntity.elements(true);
|
|
plasmaWordIndexEntry importWordIdxEntry;
|
|
while (importWordIdxEntries.hasNext()) {
|
|
|
|
// testing if import process was aborted
|
|
if (Thread.interrupted()) break;
|
|
|
|
// getting next word index entry
|
|
entryCounter++;
|
|
importWordIdxEntry = (plasmaWordIndexEntry) importWordIdxEntries.next();
|
|
String urlHash = importWordIdxEntry.getUrlHash();
|
|
if ((importUrlDB.exists(urlHash)) && (!homeUrlDB.exists(urlHash))) {
|
|
urlCounter++;
|
|
|
|
// importing the new url
|
|
plasmaCrawlLURL.Entry urlEntry = importUrlDB.getEntry(urlHash);
|
|
homeUrlDB.newEntry(urlEntry);
|
|
|
|
if (urlCounter % 500 == 0) {
|
|
log.logFine(urlCounter + " URLs processed so far.");
|
|
}
|
|
}
|
|
|
|
// adding word index entity to container
|
|
newContainer.add(importWordIdxEntry,System.currentTimeMillis());
|
|
|
|
if (entryCounter % 500 == 0) {
|
|
log.logFine(entryCounter + " word entries and " + wordCounter + " word entries processed so far.");
|
|
}
|
|
}
|
|
|
|
// testing if import process was aborted
|
|
if (Thread.interrupted()) break;
|
|
|
|
// importing entity container to home db
|
|
homeWordIndex.addEntries(newContainer, true);
|
|
|
|
// delete complete index entity file
|
|
importWordIdxEntity.close();
|
|
importWordIndex.deleteIndex(wordHash);
|
|
|
|
// print out some statistical information
|
|
if (wordCounter%500 == 0) {
|
|
wordChunkEndHash = wordHash;
|
|
wordChunkEnd = System.currentTimeMillis();
|
|
long duration = wordChunkEnd - wordChunkStart;
|
|
log.logInfo(wordCounter + " word entities imported " +
|
|
"[" + wordChunkStartHash + " .. " + wordChunkEndHash + "] " +
|
|
((startSize-importWordIndex.size())/(importWordIndex.size()/100)) +
|
|
"%\n" +
|
|
"Speed: "+ 500*1000/duration + " word entities/s" +
|
|
" | Elapsed time: " + serverDate.intervalToString(wordChunkEnd-globalStart) +
|
|
" | Estimated time: " + serverDate.intervalToString(importWordIndex.size()*((wordChunkEnd-globalStart)/wordCounter)) + "\n" +
|
|
"Free memory: " + rt.freeMemory() +
|
|
" | Total memory: " + rt.totalMemory() + "\n" +
|
|
"Home Words = " + homeWordIndex.size() +
|
|
" | Import Words = " + importWordIndex.size());
|
|
wordChunkStart = wordChunkEnd;
|
|
wordChunkStartHash = wordChunkEndHash;
|
|
}
|
|
|
|
} catch (Exception e) {
|
|
log.logSevere("Import of word entity '" + wordHash + "' failed.",e);
|
|
} finally {
|
|
if (importWordIdxEntity != null) try { importWordIdxEntity.close(); } catch (Exception e) {}
|
|
}
|
|
}
|
|
|
|
log.logInfo("Home word index contains " + homeWordIndex.size() + " words and " + homeUrlDB.size() + " URLs.");
|
|
log.logInfo("Import word index contains " + importWordIndex.size() + " words and " + importUrlDB.size() + " URLs.");
|
|
|
|
log.logInfo("DB-IMPORT FINISHED");
|
|
} catch (Exception e) {
|
|
log.logSevere("Database import failed.",e);
|
|
e.printStackTrace();
|
|
} finally {
|
|
if (homeUrlDB != null) try { homeUrlDB.close(); } catch (Exception e){}
|
|
if (importUrlDB != null) try { importUrlDB.close(); } catch (Exception e){}
|
|
if (homeWordIndex != null) try { homeWordIndex.close(5000); } catch (Exception e){}
|
|
if (importWordIndex != null) try { importWordIndex.close(5000); } catch (Exception e){}
|
|
}
|
|
}
|
|
|
|
public static void minimizeUrlDB(String homePath) {
|
|
// run with "java -classpath classes yacy -migratewords"
|
|
try {serverLog.configureLogging(new File(homePath, "yacy.logging"));} catch (Exception e) {}
|
|
File dbroot = new File(new File(homePath), "DATA/PLASMADB");
|
|
try {
|
|
serverLog log = new serverLog("URL-CLEANUP");
|
|
log.logInfo("STARTING URL CLEANUP");
|
|
|
|
// db containing all currently loades urls
|
|
plasmaCrawlLURL currentUrlDB = new plasmaCrawlLURL(new File(dbroot, "urlHash.db"), 4194304);
|
|
|
|
// db used to hold all neede urls
|
|
plasmaCrawlLURL minimizedUrlDB = new plasmaCrawlLURL(new File(dbroot, "urlHash.temp.db"), 4194304);
|
|
|
|
Runtime rt = Runtime.getRuntime();
|
|
int cacheMem = (int)(rt.maxMemory()-rt.totalMemory())-5*1024*1024;
|
|
plasmaWordIndex wordIndex = new plasmaWordIndex(dbroot, cacheMem, log);
|
|
Iterator wordHashIterator = wordIndex.wordHashes("------------", true, true);
|
|
|
|
String wordhash;
|
|
long urlCounter = 0, wordCounter = 0;
|
|
long wordChunkStart = System.currentTimeMillis(), wordChunkEnd = 0;
|
|
String wordChunkStartHash = "------------", wordChunkEndHash;
|
|
|
|
while (wordHashIterator.hasNext()) {
|
|
plasmaWordIndexEntity wordIdxEntity = null;
|
|
try {
|
|
wordCounter++;
|
|
wordhash = (String) wordHashIterator.next();
|
|
wordIdxEntity = wordIndex.getEntity(wordhash, true, -1);
|
|
|
|
// the combined container will fit, read the container
|
|
Iterator wordIdxEntries = wordIdxEntity.elements(true);
|
|
plasmaWordIndexEntry wordIdxEntry;
|
|
while (wordIdxEntries.hasNext()) {
|
|
wordIdxEntry = (plasmaWordIndexEntry) wordIdxEntries.next();
|
|
String urlHash = wordIdxEntry.getUrlHash();
|
|
if ((currentUrlDB.exists(urlHash)) && (!minimizedUrlDB.exists(urlHash))) {
|
|
urlCounter++;
|
|
plasmaCrawlLURL.Entry urlEntry = currentUrlDB.getEntry(urlHash);
|
|
minimizedUrlDB.newEntry(urlEntry);
|
|
if (urlCounter % 500 == 0) {
|
|
log.logInfo(urlCounter + " URLs found so far.");
|
|
}
|
|
}
|
|
}
|
|
// we have read all elements, now delete the entity
|
|
wordIdxEntity.close(); wordIdxEntity = null;
|
|
|
|
if (wordCounter%500 == 0) {
|
|
wordChunkEndHash = wordhash;
|
|
wordChunkEnd = System.currentTimeMillis();
|
|
long duration = wordChunkEnd - wordChunkStart;
|
|
log.logInfo(wordCounter + " words scanned " +
|
|
"[" + wordChunkStartHash + " .. " + wordChunkEndHash + "]\n" +
|
|
"Duration: "+ 500*1000/duration + " words/s" +
|
|
" | Free memory: " + rt.freeMemory() +
|
|
" | Total memory: " + rt.totalMemory());
|
|
wordChunkStart = wordChunkEnd;
|
|
wordChunkStartHash = wordChunkEndHash;
|
|
}
|
|
|
|
|
|
} catch (Exception e) {
|
|
e.printStackTrace();
|
|
} finally {
|
|
if (wordIdxEntity != null) try { wordIdxEntity.close(); } catch (Exception e) {}
|
|
}
|
|
}
|
|
currentUrlDB.close();
|
|
minimizedUrlDB.close();
|
|
wordIndex.close(600);
|
|
|
|
log.logInfo("current LURL DB contains " + currentUrlDB.size() + " entries.");
|
|
log.logInfo("mimimized LURL DB contains " + minimizedUrlDB.size() + " entries.");
|
|
|
|
// TODO: rename the mimimized UrlDB to the name of the previous UrlDB
|
|
|
|
log.logInfo("FINISHED URL CLEANUP, WAIT FOR DUMP");
|
|
log.logInfo("TERMINATED URL CLEANUP");
|
|
} catch (IOException e) {
|
|
e.printStackTrace();
|
|
}
|
|
}
|
|
|
|
/**
|
|
* Reads all words from the given file and creates a hashmap, where key is
|
|
* the plasma word hash and value is the word itself.
|
|
*
|
|
* @param wordlist File where the words are stored.
|
|
* @return HashMap with the hash-word - relation.
|
|
*/
|
|
private static HashMap loadWordMap(File wordlist) {
|
|
// returns a hash-word - Relation
|
|
HashMap wordmap = new HashMap();
|
|
try {
|
|
String word;
|
|
BufferedReader br = new BufferedReader(new InputStreamReader(new FileInputStream(wordlist)));
|
|
while ((word = br.readLine()) != null) wordmap.put(plasmaWordIndexEntry.word2hash(word),word);
|
|
br.close();
|
|
} catch (IOException e) {}
|
|
return wordmap;
|
|
}
|
|
|
|
/**
|
|
* Reads all words from the given file and creats as HashSet, which contains
|
|
* all found words.
|
|
*
|
|
* @param wordlist File where the words are stored.
|
|
* @return HashSet with the words
|
|
*/
|
|
private static HashSet loadWordSet(File wordlist) {
|
|
// returns a set of words
|
|
HashSet wordset = new HashSet();
|
|
try {
|
|
String word;
|
|
BufferedReader br = new BufferedReader(new InputStreamReader(new FileInputStream(wordlist)));
|
|
while ((word = br.readLine()) != null) wordset.add(word);
|
|
br.close();
|
|
} catch (IOException e) {}
|
|
return wordset;
|
|
}
|
|
|
|
/**
|
|
* Cleans a wordlist in a file according to the length of the words. The
|
|
* file with the given filename is read and then only the words in the given
|
|
* length-range are written back to the file.
|
|
*
|
|
* @param wordlist Name of the file the words are stored in.
|
|
* @param minlength Minimal needed length for each word to be stored.
|
|
* @param maxlength Maximal allowed length for each word to be stored.
|
|
*/
|
|
private static void cleanwordlist(String wordlist, int minlength, int maxlength) {
|
|
// start up
|
|
System.out.println(copyright);
|
|
System.out.println(hline);
|
|
serverLog.logConfig("CLEAN-WORDLIST", "START");
|
|
|
|
String word;
|
|
TreeSet wordset = new TreeSet();
|
|
int count = 0;
|
|
try {
|
|
BufferedReader br = new BufferedReader(new InputStreamReader(new FileInputStream(wordlist)));
|
|
String seps = "' .,:/-&";
|
|
while ((word = br.readLine()) != null) {
|
|
word = word.toLowerCase().trim();
|
|
for (int i = 0; i < seps.length(); i++) {
|
|
if (word.indexOf(seps.charAt(i)) >= 0) word = word.substring(0, word.indexOf(seps.charAt(i)));
|
|
}
|
|
if ((word.length() >= minlength) && (word.length() <= maxlength)) wordset.add(word);
|
|
count++;
|
|
}
|
|
br.close();
|
|
|
|
if (wordset.size() != count) {
|
|
count = count - wordset.size();
|
|
BufferedWriter bw = new BufferedWriter(new PrintWriter(new FileWriter(wordlist)));
|
|
while (wordset.size() > 0) {
|
|
word = (String) wordset.first();
|
|
bw.write(word + "\n");
|
|
wordset.remove(word);
|
|
}
|
|
bw.close();
|
|
serverLog.logInfo("CLEAN-WORDLIST", "shrinked wordlist by " + count + " words.");
|
|
} else {
|
|
serverLog.logInfo("CLEAN-WORDLIST", "not necessary to change wordlist");
|
|
}
|
|
} catch (IOException e) {
|
|
serverLog.logSevere("CLEAN-WORDLIST", "ERROR: " + e.getMessage());
|
|
System.exit(-1);
|
|
}
|
|
|
|
// finished
|
|
serverLog.logConfig("CLEAN-WORDLIST", "FINISHED");
|
|
}
|
|
|
|
/**
|
|
* Gets all words from the stopword-list and removes them in the databases.
|
|
* FIXME: Really? Don't know if I read this correctly.
|
|
*
|
|
* @param homePath Root-Path where all information is to be found.
|
|
*/
|
|
private static void deleteStopwords(String homePath) {
|
|
// start up
|
|
System.out.println(copyright);
|
|
System.out.println(hline);
|
|
serverLog.logConfig("DELETE-STOPWORDS", "START");
|
|
|
|
Properties config = configuration("DELETE-STOPWORDS", homePath);
|
|
File dbRoot = new File(homePath, config.getProperty("dbPath"));
|
|
|
|
// load stopwords
|
|
HashSet stopwords = loadWordSet(new File(homePath, "yacy.stopwords"));
|
|
serverLog.logInfo("DELETE-STOPWORDS", "loaded stopwords, " + stopwords.size() + " entries in list, starting scanning");
|
|
|
|
// find all hashes
|
|
File f;
|
|
String w;
|
|
int count = 0;
|
|
long thisamount, totalamount = 0;
|
|
Iterator i = stopwords.iterator();
|
|
while (i.hasNext()) {
|
|
w = (String) i.next();
|
|
f = plasmaWordIndexEntity.wordHash2path(dbRoot, plasmaWordIndexEntry.word2hash(w));
|
|
if (f.exists()) {
|
|
thisamount = f.length();
|
|
if (f.delete()) {
|
|
count++;
|
|
totalamount += thisamount;
|
|
serverLog.logInfo("DELETE-STOPWORDS", "deleted index for word '" + w + "', " + thisamount + " bytes");
|
|
}
|
|
}
|
|
}
|
|
|
|
serverLog.logInfo("DELETE-STOPWORDS", "TOTALS: deleted " + count + " indexes; " + (totalamount / 1024) + " kbytes");
|
|
|
|
// finished
|
|
serverLog.logConfig("DELETE-STOPWORDS", "FINISHED");
|
|
}
|
|
|
|
/**
|
|
* Main-method which is started by java. Checks for special arguments or
|
|
* starts up the application.
|
|
*
|
|
* @param args Given arguments from the command line.
|
|
*/
|
|
public static void main(String args[]) {
|
|
|
|
// check memory amount
|
|
System.gc();
|
|
long startupMemFree = Runtime.getRuntime().freeMemory(); // the amount of free memory in the Java Virtual Machine
|
|
long startupMemTotal = Runtime.getRuntime().totalMemory(); // the total amount of memory in the Java virtual machine; may vary over time
|
|
|
|
// go into headless awt mode
|
|
System.setProperty("java.awt.headless", "true");
|
|
|
|
String applicationRoot = System.getProperty("user.dir").replace('\\', '/');
|
|
//System.out.println("args.length=" + args.length);
|
|
//System.out.print("args=["); for (int i = 0; i < args.length; i++) System.out.print(args[i] + ", "); System.out.println("]");
|
|
if ((args.length >= 1) && ((args[0].equals("-startup")) || (args[0].equals("-start")))) {
|
|
// normal start-up of yacy
|
|
if (args.length == 2) applicationRoot= args[1];
|
|
startup(applicationRoot, startupMemFree, startupMemTotal);
|
|
} else if ((args.length >= 1) && ((args[0].equals("-shutdown")) || (args[0].equals("-stop")))) {
|
|
// normal shutdown of yacy
|
|
if (args.length == 2) applicationRoot= args[1];
|
|
shutdown(applicationRoot);
|
|
} else if ((args.length >= 1) && (args[0].equals("-migratewords"))) {
|
|
// migrate words from DATA/PLASMADB/WORDS path to assortment cache, if possible
|
|
// attention: this may run long and should not be interrupted!
|
|
if (args.length == 2) applicationRoot= args[1];
|
|
migrateWords(applicationRoot);
|
|
} else if ((args.length >= 1) && (args[0].equals("-minimizeUrlDB"))) {
|
|
// migrate words from DATA/PLASMADB/WORDS path to assortment cache, if possible
|
|
// attention: this may run long and should not be interrupted!
|
|
if (args.length == 2) applicationRoot= args[1];
|
|
minimizeUrlDB(applicationRoot);
|
|
} else if ((args.length >= 1) && (args[0].equals("-importDB"))) {
|
|
// attention: this may run long and should not be interrupted!
|
|
String importRoot = null;
|
|
if (args.length == 3) {
|
|
applicationRoot= args[1];
|
|
importRoot = args[2];
|
|
} else if (args.length == 2) {
|
|
importRoot = args[1];
|
|
} else {
|
|
System.err.println("Usage: -importDB [homeDbRoot] importDbRoot");
|
|
}
|
|
importDB(applicationRoot, importRoot);
|
|
} else if ((args.length >= 1) && (args[0].equals("-deletestopwords"))) {
|
|
// delete those words in the index that are listed in the stopwords file
|
|
if (args.length == 2) applicationRoot= args[1];
|
|
deleteStopwords(applicationRoot);
|
|
} else if ((args.length >= 1) && (args[0].equals("-genwordstat"))) {
|
|
// this can help to create a stop-word list
|
|
// to use this, you need a 'yacy.words' file in the root path
|
|
// start this with "java -classpath classes yacy -genwordstat [<rootdir>]"
|
|
if (args.length == 2) applicationRoot= args[1];
|
|
genWordstat(applicationRoot);
|
|
} else if ((args.length == 4) && (args[0].equals("-cleanwordlist"))) {
|
|
// this can be used to organize and clean a word-list
|
|
// start this with "java -classpath classes yacy -cleanwordlist <word-file> <minlength> <maxlength>"
|
|
int minlength = Integer.parseInt(args[2]);
|
|
int maxlength = Integer.parseInt(args[3]);
|
|
cleanwordlist(args[1], minlength, maxlength);
|
|
} else {
|
|
if (args.length == 1) applicationRoot= args[0];
|
|
startup(applicationRoot, startupMemFree, startupMemTotal);
|
|
}
|
|
}
|
|
}
|
|
|
|
/**
|
|
* This class is a helper class whose instance is started, when the java virtual
|
|
* machine shuts down. Signals the plasmaSwitchboard to shut down.
|
|
*/
|
|
class shutdownHookThread extends Thread {
|
|
private plasmaSwitchboard sb = null;
|
|
private Thread mainThread = null;
|
|
|
|
public shutdownHookThread(Thread mainThread, plasmaSwitchboard sb) {
|
|
super();
|
|
this.sb = sb;
|
|
this.mainThread = mainThread;
|
|
}
|
|
|
|
public void run() {
|
|
try {
|
|
if (!this.sb.isTerminated()) {
|
|
serverLog.logConfig("SHUTDOWN","Shutdown via shutdown hook.");
|
|
|
|
// sending the yacy main thread a shutdown signal
|
|
this.sb.terminate();
|
|
|
|
// waiting for the yacy thread to finish execution
|
|
this.mainThread.join();
|
|
}
|
|
} catch (Exception e) {
|
|
serverLog.logSevere("SHUTDOWN","Unexpected error. " + e.getClass().getName(),e);
|
|
}
|
|
}
|
|
}
|