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.
yacy_search_server/source/de/anomic/search/MetadataRepository.java

708 lines
29 KiB

// indexRepositoryReference.java
// (C) 2006 by Michael Peter Christen; mc@yacy.net, Frankfurt a. M., Germany
// first published 2006 as part of 'plasmaCrawlLURL.java' on http://yacy.net
//
// This is a part of YaCy, a peer-to-peer based web search engine
//
// $LastChangedDate: 2006-04-02 22:40:07 +0200 (So, 02 Apr 2006) $
// $LastChangedRevision: 1986 $
// $LastChangedBy: orbiter $
//
// LICENSE
//
// 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
package de.anomic.search;
import java.io.BufferedOutputStream;
import java.io.File;
import java.io.FileOutputStream;
import java.io.IOException;
import java.io.PrintWriter;
import java.util.ArrayList;
import java.util.HashMap;
import java.util.HashSet;
import java.util.Iterator;
import java.util.Map;
import java.util.TreeSet;
import de.anomic.crawler.CrawlStacker;
import net.yacy.cora.document.MultiProtocolURI;
import net.yacy.cora.protocol.http.HTTPClient;
import net.yacy.cora.storage.DynamicScore;
import net.yacy.cora.storage.ScoreCluster;
import net.yacy.document.parser.html.CharacterCoding;
import net.yacy.kelondro.data.meta.DigestURI;
import net.yacy.kelondro.data.meta.URIMetadataRow;
import net.yacy.kelondro.data.word.WordReferenceVars;
import net.yacy.kelondro.index.Cache;
import net.yacy.kelondro.index.HandleSet;
import net.yacy.kelondro.index.Index;
import net.yacy.kelondro.index.Row;
import net.yacy.kelondro.index.RowSpaceExceededException;
import net.yacy.kelondro.logging.Log;
import net.yacy.kelondro.order.CloneableIterator;
import net.yacy.kelondro.table.SplitTable;
import net.yacy.repository.Blacklist;
public final class MetadataRepository implements Iterable<byte[]> {
// class objects
protected Index urlIndexFile;
private Export exportthread; // will have a export thread assigned if exporter is running
private File location;
private ArrayList<hostStat> statsDump;
public MetadataRepository(
final File path,
final String tablename,
final boolean useTailCache,
final boolean exceed134217727) {
this.location = path;
Index backupIndex = null;
try {
backupIndex = new SplitTable(this.location, tablename, URIMetadataRow.rowdef, useTailCache, exceed134217727);
} catch (RowSpaceExceededException e) {
try {
backupIndex = new SplitTable(this.location, tablename, URIMetadataRow.rowdef, false, exceed134217727);
} catch (RowSpaceExceededException e1) {
Log.logException(e1);
}
}
this.urlIndexFile = backupIndex; //new Cache(backupIndex, 20000000, 20000000);
this.exportthread = null; // will have a export thread assigned if exporter is running
this.statsDump = null;
}
public void clearCache() {
if (urlIndexFile instanceof Cache) ((Cache) urlIndexFile).clearCache();
statsDump = null;
}
public void clear() throws IOException {
if (exportthread != null) exportthread.interrupt();
urlIndexFile.clear();
statsDump = null;
}
public int size() {
return urlIndexFile == null ? 0 : urlIndexFile.size();
}
public void close() {
statsDump = null;
if (urlIndexFile != null) {
urlIndexFile.close();
urlIndexFile = null;
}
}
public int writeCacheSize() {
if (urlIndexFile instanceof SplitTable) return ((SplitTable) urlIndexFile).writeBufferSize();
if (urlIndexFile instanceof Cache) return ((Cache) urlIndexFile).writeBufferSize();
return 0;
}
public URIMetadataRow load(final byte[] urlHash, final WordReferenceVars searchedWord, final long ranking) {
// generates an plasmaLURLEntry using the url hash
// if the url cannot be found, this returns null
if (urlHash == null) return null;
assert urlIndexFile != null : "urlHash = " + new String(urlHash);
if (urlIndexFile == null) return null;
try {
final Row.Entry entry = urlIndexFile.get(urlHash);
if (entry == null) return null;
return new URIMetadataRow(entry, searchedWord, ranking);
} catch (final IOException e) {
return null;
}
}
public void store(final URIMetadataRow entry) throws IOException {
// Check if there is a more recent Entry already in the DB
URIMetadataRow oldEntry;
if (urlIndexFile == null) return; // case may happen during shutdown or startup
try {
Row.Entry oe = urlIndexFile.get(entry.hash());
oldEntry = (oe == null) ? null : new URIMetadataRow(oe, null, 0);
} catch (final Exception e) {
Log.logException(e);
oldEntry = null;
}
if (oldEntry != null && entry.isOlder(oldEntry)) {
// the fetched oldEntry is better, so return its properties instead of the new ones
// this.urlHash = oldEntry.urlHash; // unnecessary, should be the same
// this.url = oldEntry.url; // unnecessary, should be the same
// doesn't make sense, since no return value:
//entry = oldEntry;
return; // this did not need to be stored, but is updated
}
try {
urlIndexFile.put(entry.toRowEntry());
} catch (RowSpaceExceededException e) {
throw new IOException("RowSpaceExceededException in " + this.urlIndexFile.filename() + ": " + e.getMessage());
}
statsDump = null;
}
public boolean remove(final byte[] urlHashBytes) {
if (urlHashBytes == null) return false;
try {
final Row.Entry r = urlIndexFile.remove(urlHashBytes);
if (r != null) statsDump = null;
return r != null;
} catch (final IOException e) {
return false;
}
}
public boolean exists(final byte[] urlHash) {
if (urlIndexFile == null) return false; // case may happen during shutdown
return urlIndexFile.has(urlHash);
}
public CloneableIterator<byte[]> keys(boolean up, byte[] firstKey) {
try {
return this.urlIndexFile.keys(up, firstKey);
} catch (IOException e) {
Log.logException(e);
return null;
}
}
public Iterator<byte[]> iterator() {
return keys(true, null);
}
public CloneableIterator<URIMetadataRow> entries() throws IOException {
// enumerates entry elements
return new kiter();
}
public CloneableIterator<URIMetadataRow> entries(final boolean up, final String firstHash) throws IOException {
// enumerates entry elements
return new kiter(up, firstHash);
}
public class kiter implements CloneableIterator<URIMetadataRow> {
// enumerates entry elements
private final Iterator<Row.Entry> iter;
private final boolean error;
boolean up;
public kiter() throws IOException {
this.up = true;
this.iter = urlIndexFile.rows();
this.error = false;
}
public kiter(final boolean up, final String firstHash) throws IOException {
this.up = up;
this.iter = urlIndexFile.rows(up, (firstHash == null) ? null : firstHash.getBytes());
this.error = false;
}
public kiter clone(final Object secondHash) {
try {
return new kiter(up, (String) secondHash);
} catch (final IOException e) {
return null;
}
}
public final boolean hasNext() {
if (this.error) return false;
if (this.iter == null) return false;
return this.iter.hasNext();
}
public final URIMetadataRow next() {
Row.Entry e = null;
if (this.iter == null) { return null; }
if (this.iter.hasNext()) { e = this.iter.next(); }
if (e == null) { return null; }
return new URIMetadataRow(e, null, 0);
}
public final void remove() {
this.iter.remove();
}
}
/**
* Uses an Iteration over urlHash.db to detect malformed URL-Entries.
* Damaged URL-Entries will be marked in a HashSet and removed at the end of the function.
*
* @param proxyConfig
*/
public void deadlinkCleaner() {
final Log log = new Log("URLDBCLEANUP");
final HashSet<String> damagedURLS = new HashSet<String>();
try {
final Iterator<URIMetadataRow> eiter = entries(true, null);
int iteratorCount = 0;
while (eiter.hasNext()) try {
eiter.next();
iteratorCount++;
} catch (final RuntimeException e) {
if(e.getMessage() != null) {
final String m = e.getMessage();
damagedURLS.add(m.substring(m.length() - 12));
} else {
log.logSevere("RuntimeException:", e);
}
}
log.logInfo("URLs vorher: " + urlIndexFile.size() + " Entries loaded during Iteratorloop: " + iteratorCount + " kaputte URLs: " + damagedURLS.size());
final HTTPClient client = new HTTPClient();
final Iterator<String> eiter2 = damagedURLS.iterator();
byte[] urlHashBytes;
while (eiter2.hasNext()) {
urlHashBytes = eiter2.next().getBytes();
// trying to fix the invalid URL
String oldUrlStr = null;
try {
// getting the url data as byte array
final Row.Entry entry = urlIndexFile.get(urlHashBytes);
// getting the wrong url string
oldUrlStr = entry.getColString(1, null).trim();
int pos = -1;
if ((pos = oldUrlStr.indexOf("://")) != -1) {
// trying to correct the url
final String newUrlStr = "http://" + oldUrlStr.substring(pos + 3);
final DigestURI newUrl = new DigestURI(newUrlStr);
// doing a http head request to test if the url is correct
// final Client client = new Client(10000);
// ResponseContainer res = null;
// try {
// res = client.HEAD(newUrl.toString());
// } finally {
// if(res != null) {
// // release connection
// res.closeStream();
// }
// }
if (client.HEADResponse(newUrl.toString()) != null
&& client.getHttpResponse().getStatusLine().getStatusCode() == 200) {
entry.setCol(1, newUrl.toString().getBytes());
urlIndexFile.put(entry);
if (log.isInfo()) log.logInfo("UrlDB-Entry with urlHash '" + new String(urlHashBytes) + "' corrected\n\tURL: " + oldUrlStr + " -> " + newUrlStr);
} else {
remove(urlHashBytes);
if (log.isInfo()) log.logInfo("UrlDB-Entry with urlHash '" + new String(urlHashBytes) + "' removed\n\tURL: " + oldUrlStr + "\n\tConnection Status: " + (client.getHttpResponse() == null ? "null" : client.getHttpResponse().getStatusLine()));
}
}
} catch (final Exception e) {
remove(urlHashBytes);
if (log.isInfo()) log.logInfo("UrlDB-Entry with urlHash '" + new String(urlHashBytes) + "' removed\n\tURL: " + oldUrlStr + "\n\tExecption: " + e.getMessage());
}
}
log.logInfo("URLs nachher: " + size() + " kaputte URLs: " + damagedURLS.size());
} catch (final IOException e) {
log.logSevere("IOException", e);
}
}
public BlacklistCleaner getBlacklistCleaner(final Blacklist blacklist, final CrawlStacker crawlStacker) {
return new BlacklistCleaner(blacklist, crawlStacker);
}
public class BlacklistCleaner extends Thread {
private boolean run = true;
private boolean pause;
public int blacklistedUrls = 0;
public int totalSearchedUrls = 1;
public String lastBlacklistedUrl = "";
public String lastBlacklistedHash = "";
public String lastUrl = "";
public String lastHash = "";
private final Blacklist blacklist;
private final CrawlStacker crawlStacker;
public BlacklistCleaner(final Blacklist blacklist, final CrawlStacker crawlStacker) {
this.blacklist = blacklist;
this.crawlStacker = crawlStacker;
}
public void run() {
try {
Log.logInfo("URLDBCLEANER", "UrldbCleaner-Thread startet");
final Iterator<URIMetadataRow> eiter = entries(true, null);
while (eiter.hasNext() && run) {
synchronized (this) {
if (this.pause) {
try {
this.wait();
} catch (final InterruptedException e) {
Log.logWarning("URLDBCLEANER", "InterruptedException", e);
this.run = false;
return;
}
}
}
final URIMetadataRow entry = eiter.next();
if (entry == null) {
if (Log.isFine("URLDBCLEANER")) Log.logFine("URLDBCLEANER", "entry == null");
} else if (entry.hash() == null) {
if (Log.isFine("URLDBCLEANER")) Log.logFine("URLDBCLEANER", ++blacklistedUrls + " blacklisted (" + ((double) blacklistedUrls / totalSearchedUrls) * 100 + "%): " + "hash == null");
} else {
final URIMetadataRow.Components metadata = entry.metadata();
totalSearchedUrls++;
if (metadata == null) {
if (Log.isFine("URLDBCLEANER")) Log.logFine("URLDBCLEANER", "corrupted entry for hash = " + new String(entry.hash()));
remove(entry.hash());
continue;
}
if (metadata.url() == null) {
if (Log.isFine("URLDBCLEANER")) Log.logFine("URLDBCLEANER", ++blacklistedUrls + " blacklisted (" + ((double) blacklistedUrls / totalSearchedUrls) * 100 + "%): " + new String(entry.hash()) + "URL == null");
remove(entry.hash());
continue;
}
if (blacklist.isListed(Blacklist.BLACKLIST_CRAWLER, metadata.url()) ||
blacklist.isListed(Blacklist.BLACKLIST_DHT, metadata.url()) ||
(crawlStacker.urlInAcceptedDomain(metadata.url()) != null)) {
lastBlacklistedUrl = metadata.url().toNormalform(true, true);
lastBlacklistedHash = new String(entry.hash());
if (Log.isFine("URLDBCLEANER")) Log.logFine("URLDBCLEANER", ++blacklistedUrls + " blacklisted (" + ((double) blacklistedUrls / totalSearchedUrls) * 100 + "%): " + new String(entry.hash()) + " " + metadata.url().toNormalform(false, true));
remove(entry.hash());
if (blacklistedUrls % 100 == 0) {
Log.logInfo("URLDBCLEANER", "Deleted " + blacklistedUrls + " URLs until now. Last deleted URL-Hash: " + lastBlacklistedUrl);
}
}
lastUrl = metadata.url().toNormalform(true, true);
lastHash = new String(entry.hash());
}
}
} catch (final RuntimeException e) {
if (e.getMessage() != null && e.getMessage().indexOf("not found in LURL") != -1) {
Log.logWarning("URLDBCLEANER", "urlHash not found in LURL", e);
}
else {
Log.logWarning("URLDBCLEANER", "RuntimeException", e);
run = false;
}
} catch (final IOException e) {
Log.logException(e);
run = false;
} catch (final Exception e) {
Log.logException(e);
run = false;
}
Log.logInfo("URLDBCLEANER", "UrldbCleaner-Thread stopped");
}
public void abort() {
synchronized(this) {
run = false;
this.notifyAll();
}
}
public void pause() {
synchronized(this) {
if (!pause) {
pause = true;
Log.logInfo("URLDBCLEANER", "UrldbCleaner-Thread paused");
}
}
}
public void endPause() {
synchronized(this) {
if (pause) {
pause = false;
this.notifyAll();
Log.logInfo("URLDBCLEANER", "UrldbCleaner-Thread resumed");
}
}
}
}
// export methods
public Export export(final File f, final String filter, HandleSet set, final int format, final boolean dom) {
if ((exportthread != null) && (exportthread.isAlive())) {
Log.logWarning("LURL-EXPORT", "cannot start another export thread, already one running");
return exportthread;
}
this.exportthread = new Export(f, filter, set, format, dom);
this.exportthread.start();
return exportthread;
}
public Export export() {
return this.exportthread;
}
public class Export extends Thread {
private final File f;
private final String filter;
private int count;
private String failure;
private final int format;
private final boolean dom;
private final HandleSet set;
public Export(final File f, final String filter, HandleSet set, final int format, boolean dom) {
// format: 0=text, 1=html, 2=rss/xml
this.f = f;
this.filter = filter;
this.count = 0;
this.failure = null;
this.format = format;
this.dom = dom;
this.set = set;
if ((dom) && (format == 2)) dom = false;
}
public void run() {
try {
File parentf = f.getParentFile();
if (parentf != null) parentf.mkdirs();
final PrintWriter pw = new PrintWriter(new BufferedOutputStream(new FileOutputStream(f)));
if (format == 1) {
pw.println("<html><head></head><body>");
}
if (format == 2) {
pw.println("<?xml version=\"1.0\" encoding=\"UTF-8\"?>");
pw.println("<?xml-stylesheet type='text/xsl' href='/yacysearch.xsl' version='1.0'?>");
pw.println("<rss version=\"2.0\" xmlns:yacy=\"http://www.yacy.net/\" xmlns:opensearch=\"http://a9.com/-/spec/opensearch/1.1/\" xmlns:atom=\"http://www.w3.org/2005/Atom\">");
pw.println("<channel>");
pw.println("<title>YaCy Peer-to-Peer - Web-Search LURL Export</title>");
pw.println("<description></description>");
pw.println("<link>http://yacy.net</link>");
}
if (dom) {
TreeSet<String> set = domainNameCollector(-1);
for (String host: set) {
if (!host.matches(filter)) continue;
if (format == 0) pw.println(host);
if (format == 1) pw.println("<a href=\"http://" + host + "\">" + host + "</a><br>");
count++;
}
} else {
final Iterator<URIMetadataRow> i = entries(); // iterates indexURLEntry objects
URIMetadataRow entry;
URIMetadataRow.Components metadata;
String url;
while (i.hasNext()) {
entry = i.next();
if (this.set != null && !set.has(entry.hash())) continue;
metadata = entry.metadata();
url = metadata.url().toNormalform(true, false);
if (!url.matches(filter)) continue;
if (format == 0) {
pw.println(url);
}
if (format == 1) {
pw.println("<a href=\"" + url + "\">" + CharacterCoding.unicode2xml(metadata.dc_title(), true) + "</a><br>");
}
if (format == 2) {
pw.println("<item>");
pw.println("<title>" + CharacterCoding.unicode2xml(metadata.dc_title(), true) + "</title>");
pw.println("<link>" + MultiProtocolURI.escape(url) + "</link>");
if (metadata.dc_creator().length() > 0) pw.println("<author>" + CharacterCoding.unicode2xml(metadata.dc_creator(), true) + "</author>");
if (metadata.dc_subject().length() > 0) pw.println("<description>" + CharacterCoding.unicode2xml(metadata.dc_subject(), true) + "</description>");
pw.println("<pubDate>" + entry.moddate().toString() + "</pubDate>");
pw.println("<yacy:size>" + entry.size() + "</yacy:size>");
pw.println("<guid isPermaLink=\"false\">" + new String(entry.hash()) + "</guid>");
pw.println("</item>");
}
count++;
}
}
if (format == 1) {
pw.println("</body></html>");
}
if (format == 2) {
pw.println("</channel>");
pw.println("</rss>");
}
pw.close();
} catch (final IOException e) {
Log.logException(e);
this.failure = e.getMessage();
} catch (final Exception e) {
Log.logException(e);
this.failure = e.getMessage();
}
// terminate process
}
public File file() {
return this.f;
}
public String failed() {
return this.failure;
}
public int count() {
return this.count;
}
}
private Map<String, hashStat> domainSampleCollector() throws IOException {
Map<String, hashStat> map = new HashMap<String, hashStat>();
// first collect all domains and calculate statistics about it
CloneableIterator<byte[]> i = this.urlIndexFile.keys(true, null);
String urlhash, hosthash;
hashStat ds;
if (i != null) while (i.hasNext()) {
urlhash = new String(i.next());
hosthash = urlhash.substring(6);
ds = map.get(hosthash);
if (ds == null) {
ds = new hashStat(urlhash);
map.put(hosthash, ds);
} else {
ds.count++;
}
}
return map;
}
public TreeSet<String> domainNameCollector(int count) throws IOException {
// collect hashes from all domains
Map<String, hashStat> map = domainSampleCollector();
// fetch urls from the database to determine the host in clear text
URIMetadataRow urlref;
if (count < 0 || count > map.size()) count = map.size();
statsDump = new ArrayList<hostStat>();
TreeSet<String> set = new TreeSet<String>();
for (hashStat hs: map.values()) {
if (hs == null) continue;
urlref = this.load(hs.urlhash.getBytes(), null, 0);
if (urlref == null || urlref.metadata() == null || urlref.metadata().url() == null || urlref.metadata().url().getHost() == null) continue;
set.add(urlref.metadata().url().getHost());
count--;
if (count == 0) break;
}
return set;
}
public Iterator<hostStat> statistics(int count) throws IOException {
// prevent too heavy IO.
if (statsDump != null && count <= statsDump.size()) return statsDump.iterator();
// collect hashes from all domains
Map<String, hashStat> map = domainSampleCollector();
// order elements by size
DynamicScore<String> s = new ScoreCluster<String>();
for (Map.Entry<String, hashStat> e: map.entrySet()) {
s.inc(e.getValue().urlhash, e.getValue().count);
}
// fetch urls from the database to determine the host in clear text
Iterator<String> j = s.keys(false); // iterate urlhash-examples in reverse order (biggest first)
URIMetadataRow urlref;
String urlhash;
count += 10; // make some more to prevent that we have to do this again after deletions too soon.
if (count < 0 || count > s.size()) count = s.size();
statsDump = new ArrayList<hostStat>();
URIMetadataRow.Components comps;
DigestURI url;
while (j.hasNext()) {
urlhash = j.next();
if (urlhash == null) continue;
urlref = this.load(urlhash.getBytes(), null, 0);
if (urlref == null || urlref.metadata() == null || urlref.metadata().url() == null || urlref.metadata().url().getHost() == null) continue;
if (statsDump == null) return new ArrayList<hostStat>().iterator(); // some other operation has destroyed the object
comps = urlref.metadata();
url = comps.url();
statsDump.add(new hostStat(url.getHost(), url.getPort(), urlhash.substring(6), s.get(urlhash)));
count--;
if (count == 0) break;
}
// finally return an iterator for the result array
return (statsDump == null) ? new ArrayList<hostStat>().iterator() : statsDump.iterator();
}
private static class hashStat {
public String urlhash;
public int count;
public hashStat(String urlhash) {
this.urlhash = urlhash;
this.count = 1;
}
}
public static class hostStat {
public String hostname, hosthash;
public int port;
public int count;
public hostStat(String host, int port, String urlhashfragment, int count) {
assert urlhashfragment.length() == 6;
this.hostname = host;
this.port = port;
this.hosthash = urlhashfragment;
this.count = count;
}
}
/**
* using a fragment of the url hash (5 bytes: bytes 6 to 10) it is possible to address all urls from a specific domain
* here such a fragment can be used to delete all these domains at once
* @param hosthash
* @return number of deleted domains
* @throws IOException
*/
public int deleteDomain(String hosthash) throws IOException {
// first collect all url hashes that belong to the domain
assert hosthash.length() == 6;
ArrayList<String> l = new ArrayList<String>();
CloneableIterator<byte[]> i = this.urlIndexFile.keys(true, null);
String hash;
while (i != null && i.hasNext()) {
hash = new String(i.next());
if (hosthash.equals(hash.substring(6))) l.add(hash);
}
// then delete the urls using this list
int cnt = 0;
for (String h: l) {
if (urlIndexFile.delete(h.getBytes())) cnt++;
}
// finally remove the line with statistics
if (statsDump != null) {
Iterator<hostStat> hsi = statsDump.iterator();
hostStat hs;
while (hsi.hasNext()) {
hs = hsi.next();
if (hs.hosthash.equals(hosthash)) {
hsi.remove();
break;
}
}
}
return cnt;
}
}