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.
383 lines
17 KiB
383 lines
17 KiB
//plasmaCrawlRobotsTxt.java
|
|
//-------------------------------------
|
|
//part of YACY
|
|
//(C) by Michael Peter Christen; mc@yacy.net
|
|
//first published on http://www.anomic.de
|
|
//Frankfurt, Germany, 2004
|
|
//
|
|
//This file is contributed by Martin Thelian
|
|
// [MC] moved some methods from robotsParser file that had been created by Alexander Schier to this class
|
|
//last major change: $LastChangedDate$ by $LastChangedBy$
|
|
//Revision: $LastChangedRevision$
|
|
//
|
|
//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 net.yacy.crawler.robots;
|
|
|
|
import java.io.IOException;
|
|
import java.net.MalformedURLException;
|
|
import java.util.ArrayList;
|
|
import java.util.Collection;
|
|
import java.util.Date;
|
|
import java.util.Map;
|
|
import java.util.concurrent.BlockingQueue;
|
|
import java.util.concurrent.ConcurrentHashMap;
|
|
import java.util.concurrent.ConcurrentMap;
|
|
import java.util.concurrent.LinkedBlockingQueue;
|
|
import java.util.regex.Pattern;
|
|
|
|
import net.yacy.cora.document.id.DigestURL;
|
|
import net.yacy.cora.document.id.MultiProtocolURL;
|
|
import net.yacy.cora.federate.yacy.CacheStrategy;
|
|
import net.yacy.cora.protocol.ClientIdentification;
|
|
import net.yacy.cora.protocol.HeaderFramework;
|
|
import net.yacy.cora.util.ConcurrentLog;
|
|
import net.yacy.cora.util.SpaceExceededException;
|
|
import net.yacy.crawler.retrieval.Request;
|
|
import net.yacy.crawler.retrieval.Response;
|
|
import net.yacy.data.WorkTables;
|
|
import net.yacy.kelondro.blob.BEncodedHeap;
|
|
import net.yacy.repository.LoaderDispatcher;
|
|
import net.yacy.repository.Blacklist.BlacklistType;
|
|
|
|
public class RobotsTxt {
|
|
|
|
private final static ConcurrentLog log = new ConcurrentLog(RobotsTxt.class.getName());
|
|
|
|
protected static final String ROBOTS_DB_PATH_SEPARATOR = ";";
|
|
protected static final Pattern ROBOTS_DB_PATH_SEPARATOR_MATCHER = Pattern.compile(ROBOTS_DB_PATH_SEPARATOR);
|
|
|
|
private final ConcurrentMap<String, DomSync> syncObjects;
|
|
//private static final HashSet<String> loadedRobots = new HashSet<String>(); // only for debugging
|
|
private final WorkTables tables;
|
|
private final LoaderDispatcher loader;
|
|
|
|
private static class DomSync {
|
|
private DomSync() {}
|
|
}
|
|
|
|
public RobotsTxt(final WorkTables worktables, LoaderDispatcher loader) {
|
|
this.syncObjects = new ConcurrentHashMap<String, DomSync>();
|
|
this.tables = worktables;
|
|
this.loader = loader;
|
|
try {
|
|
this.tables.getHeap(WorkTables.TABLE_ROBOTS_NAME);
|
|
//log.info("initiated robots table: " + this.tables.getHeap(WorkTables.TABLE_ROBOTS_NAME).getFile());
|
|
} catch (final IOException e) {
|
|
try {
|
|
this.tables.getHeap(WorkTables.TABLE_ROBOTS_NAME).clear();
|
|
} catch (final IOException e1) {
|
|
}
|
|
}
|
|
}
|
|
|
|
public void clear() throws IOException {
|
|
log.info("clearing robots table");
|
|
this.tables.getHeap(WorkTables.TABLE_ROBOTS_NAME).clear();
|
|
this.syncObjects.clear();
|
|
}
|
|
|
|
public int size() throws IOException {
|
|
return this.tables.getHeap(WorkTables.TABLE_ROBOTS_NAME).size();
|
|
}
|
|
|
|
public RobotsTxtEntry getEntry(final MultiProtocolURL theURL, final ClientIdentification.Agent agent) {
|
|
if (theURL == null) throw new IllegalArgumentException();
|
|
if (!theURL.getProtocol().startsWith("http")) return null;
|
|
return getEntry(getHostPort(theURL), agent, true);
|
|
}
|
|
|
|
public RobotsTxtEntry getEntry(final String urlHostPort, final ClientIdentification.Agent agent, final boolean fetchOnlineIfNotAvailableOrNotFresh) {
|
|
// this method will always return a non-null value
|
|
RobotsTxtEntry robotsTxt4Host = null;
|
|
Map<String, byte[]> record;
|
|
BEncodedHeap robotsTable = null;
|
|
try {
|
|
robotsTable = this.tables.getHeap(WorkTables.TABLE_ROBOTS_NAME);
|
|
} catch (final IOException e1) {
|
|
log.severe("tables not available", e1);
|
|
}
|
|
try {
|
|
record = robotsTable.get(robotsTable.encodedKey(urlHostPort));
|
|
} catch (final SpaceExceededException e) {
|
|
log.warn("memory exhausted", e);
|
|
record = null;
|
|
} catch (final IOException e) {
|
|
log.warn("cannot get robotstxt from table", e);
|
|
record = null;
|
|
}
|
|
if (record != null) robotsTxt4Host = new RobotsTxtEntry(urlHostPort, record);
|
|
|
|
if (fetchOnlineIfNotAvailableOrNotFresh && (
|
|
robotsTxt4Host == null ||
|
|
robotsTxt4Host.getLoadedDate() == null ||
|
|
System.currentTimeMillis() - robotsTxt4Host.getLoadedDate().getTime() > 7*24*60*60*1000
|
|
)) {
|
|
|
|
// make or get a synchronization object
|
|
DomSync syncObj = this.syncObjects.get(urlHostPort);
|
|
if (syncObj == null) {
|
|
syncObj = new DomSync();
|
|
this.syncObjects.put(urlHostPort, syncObj);
|
|
}
|
|
|
|
// we can now synchronize for each host separately
|
|
synchronized (syncObj) {
|
|
// if we have not found any data or the data is older than 7 days, we need to load it from the remote server
|
|
// check the robots table again for all threads that come here because they waited for another one
|
|
// to complete a download
|
|
try {
|
|
record = robotsTable.get(robotsTable.encodedKey(urlHostPort));
|
|
} catch (final SpaceExceededException e) {
|
|
log.warn("memory exhausted", e);
|
|
record = null;
|
|
} catch (final IOException e) {
|
|
log.warn("cannot get robotstxt from table", e);
|
|
record = null;
|
|
}
|
|
if (record != null) robotsTxt4Host = new RobotsTxtEntry(urlHostPort, record);
|
|
if (robotsTxt4Host != null &&
|
|
robotsTxt4Host.getLoadedDate() != null &&
|
|
System.currentTimeMillis() - robotsTxt4Host.getLoadedDate().getTime() <= 1*24*60*60*1000) {
|
|
return robotsTxt4Host;
|
|
}
|
|
|
|
// generating the proper url to download the robots txt
|
|
DigestURL robotsURL = null;
|
|
try {
|
|
robotsURL = new DigestURL((urlHostPort.endsWith(":443") ? "https://" : "http://") + urlHostPort + "/robots.txt");
|
|
} catch (final MalformedURLException e) {
|
|
log.severe("Unable to generate robots.txt URL for host:port '" + urlHostPort + "'.", e);
|
|
robotsURL = null;
|
|
}
|
|
|
|
Response response = null;
|
|
if (robotsURL != null) {
|
|
if (log.isFine()) log.fine("Trying to download the robots.txt file from URL '" + robotsURL + "'.");
|
|
Request request = new Request(robotsURL, null);
|
|
try {
|
|
response = RobotsTxt.this.loader.load(request, CacheStrategy.NOCACHE, null, agent);
|
|
} catch (final Throwable e) {
|
|
log.info("Trying to download the robots.txt file from URL '" + robotsURL + "' failed - " + e.getMessage());
|
|
response = null;
|
|
}
|
|
}
|
|
|
|
if (response == null) {
|
|
processOldEntry(robotsTxt4Host, robotsURL, robotsTable);
|
|
} else {
|
|
processNewEntry(robotsURL, response, agent.robotIDs);
|
|
}
|
|
}
|
|
}
|
|
|
|
return robotsTxt4Host;
|
|
}
|
|
|
|
public void ensureExist(final MultiProtocolURL theURL, final ClientIdentification.Agent agent, boolean concurrent) {
|
|
if (theURL.isLocal()) return;
|
|
final String urlHostPort = getHostPort(theURL);
|
|
if (urlHostPort == null) return;
|
|
final BEncodedHeap robotsTable;
|
|
try {
|
|
robotsTable = this.tables.getHeap(WorkTables.TABLE_ROBOTS_NAME);
|
|
} catch (final IOException e1) {
|
|
log.severe("tables not available", e1);
|
|
return;
|
|
}
|
|
if (robotsTable == null || robotsTable.containsKey(robotsTable.encodedKey(urlHostPort))) return;
|
|
Thread t = new Thread() {
|
|
public void run(){
|
|
// make or get a synchronization object
|
|
DomSync syncObj = RobotsTxt.this.syncObjects.get(urlHostPort);
|
|
if (syncObj == null) {
|
|
syncObj = new DomSync();
|
|
RobotsTxt.this.syncObjects.put(urlHostPort, syncObj);
|
|
}
|
|
// we can now synchronize for each host separately
|
|
synchronized (syncObj) {
|
|
if (robotsTable.containsKey(robotsTable.encodedKey(urlHostPort))) return;
|
|
|
|
// generating the proper url to download the robots txt
|
|
DigestURL robotsURL = null;
|
|
try {
|
|
robotsURL = new DigestURL((urlHostPort.endsWith(":443") ? "https://" : "http://") + urlHostPort + "/robots.txt");
|
|
} catch (final MalformedURLException e) {
|
|
log.severe("Unable to generate robots.txt URL for host:port '" + urlHostPort + "'.", e);
|
|
robotsURL = null;
|
|
}
|
|
|
|
Response response = null;
|
|
if (robotsURL != null) {
|
|
if (log.isFine()) log.fine("Trying to download the robots.txt file from URL '" + robotsURL + "'.");
|
|
Request request = new Request(robotsURL, null);
|
|
try {
|
|
response = RobotsTxt.this.loader.load(request, CacheStrategy.NOCACHE, null, agent);
|
|
} catch (final IOException e) {
|
|
response = null;
|
|
}
|
|
}
|
|
|
|
if (response == null) {
|
|
processOldEntry(null, robotsURL, robotsTable);
|
|
} else {
|
|
processNewEntry(robotsURL, response, agent.robotIDs);
|
|
}
|
|
}
|
|
}
|
|
};
|
|
if (concurrent) t.start(); else t.run();
|
|
}
|
|
|
|
private void processOldEntry(RobotsTxtEntry robotsTxt4Host, DigestURL robotsURL, BEncodedHeap robotsTable) {
|
|
// no robots.txt available, make an entry to prevent that the robots loading is done twice
|
|
if (robotsTxt4Host == null) {
|
|
// generate artificial entry
|
|
robotsTxt4Host = new RobotsTxtEntry(
|
|
robotsURL,
|
|
new ArrayList<String>(),
|
|
new ArrayList<String>(),
|
|
new Date(),
|
|
new Date(),
|
|
null,
|
|
null,
|
|
Integer.valueOf(0),
|
|
null);
|
|
} else {
|
|
robotsTxt4Host.setLoadedDate(new Date());
|
|
}
|
|
|
|
// store the data into the robots DB
|
|
final int sz = robotsTable.size();
|
|
addEntry(robotsTxt4Host);
|
|
if (robotsTable.size() <= sz) {
|
|
log.severe("new entry in robots.txt table failed, resetting database");
|
|
try {clear();} catch (final IOException e) {}
|
|
addEntry(robotsTxt4Host);
|
|
}
|
|
}
|
|
|
|
private void processNewEntry(DigestURL robotsURL, Response response, final String[] thisAgents) {
|
|
final byte[] robotsTxt = response.getContent();
|
|
//Log.logInfo("RobotsTxt", "robots of " + robotsURL.toNormalform(true, true) + ":\n" + ((robotsTxt == null) ? "null" : UTF8.String(robotsTxt))); // debug TODO remove
|
|
RobotsTxtParser parserResult;
|
|
ArrayList<String> denyPath;
|
|
if (response.getResponseHeader().getStatusCode() == 401 || response.getResponseHeader().getStatusCode() == 403) {
|
|
parserResult = new RobotsTxtParser(thisAgents);
|
|
// create virtual deny path
|
|
denyPath = new ArrayList<String>();
|
|
denyPath.add("/");
|
|
} else {
|
|
parserResult = new RobotsTxtParser(thisAgents, robotsTxt);
|
|
denyPath = parserResult.denyList();
|
|
}
|
|
|
|
// store the data into the robots DB
|
|
String etag = response.getResponseHeader().containsKey(HeaderFramework.ETAG) ? (response.getResponseHeader().get(HeaderFramework.ETAG)).trim() : null;
|
|
boolean isBrowserAgent = thisAgents.length == 1 && thisAgents[0].equals("Mozilla");
|
|
if (isBrowserAgent) denyPath.clear();
|
|
final RobotsTxtEntry robotsTxt4Host = new RobotsTxtEntry(
|
|
robotsURL,
|
|
parserResult.allowList(),
|
|
denyPath,
|
|
new Date(),
|
|
response.getResponseHeader().lastModified(),
|
|
etag,
|
|
parserResult.sitemap(),
|
|
parserResult.crawlDelayMillis(),
|
|
parserResult.agentName());
|
|
addEntry(robotsTxt4Host);
|
|
}
|
|
|
|
private String addEntry(final RobotsTxtEntry entry) {
|
|
// writes a new page and returns key
|
|
try {
|
|
final BEncodedHeap robotsTable = this.tables.getHeap(WorkTables.TABLE_ROBOTS_NAME);
|
|
robotsTable.insert(robotsTable.encodedKey(entry.getHostName()), entry.getMem());
|
|
return entry.getHostName();
|
|
} catch (final Exception e) {
|
|
log.warn("cannot write robots.txt entry", e);
|
|
return null;
|
|
}
|
|
}
|
|
|
|
static final String getHostPort(final MultiProtocolURL theURL) {
|
|
int port = theURL.getPort();
|
|
if (port == -1) {
|
|
if (theURL.getProtocol().equalsIgnoreCase("http")) {
|
|
port = 80;
|
|
} else if (theURL.getProtocol().equalsIgnoreCase("https")) {
|
|
port = 443;
|
|
} else {
|
|
port = 80;
|
|
}
|
|
}
|
|
String host = theURL.getHost();
|
|
if (host == null) return null;
|
|
StringBuilder sb = new StringBuilder(host.length() + 6);
|
|
sb.append(host).append(':').append(Integer.toString(port));
|
|
return sb.toString();
|
|
}
|
|
|
|
public static class CheckEntry {
|
|
public final DigestURL digestURL;
|
|
public final RobotsTxtEntry robotsTxtEntry;
|
|
public final Response response;
|
|
public final String error;
|
|
public CheckEntry(DigestURL digestURL, RobotsTxtEntry robotsTxtEntry, Response response, String error) {
|
|
this.digestURL = digestURL;
|
|
this.robotsTxtEntry = robotsTxtEntry;
|
|
this.response = response;
|
|
this.error = error;
|
|
}
|
|
}
|
|
|
|
public Collection<CheckEntry> massCrawlCheck(final Collection<DigestURL> rootURLs, final ClientIdentification.Agent userAgent, final int concurrency) {
|
|
// put the rootURLs into a blocking queue as input for concurrent computation
|
|
final BlockingQueue<DigestURL> in = new LinkedBlockingQueue<DigestURL>();
|
|
try {
|
|
for (DigestURL u: rootURLs) in.put(u);
|
|
for (int i = 0; i < concurrency; i++) in.put(DigestURL.POISON);
|
|
} catch (InterruptedException e) {}
|
|
final BlockingQueue<CheckEntry> out = new LinkedBlockingQueue<CheckEntry>();
|
|
final Thread[] threads = new Thread[concurrency];
|
|
for (int i = 0; i < concurrency; i++) {
|
|
threads[i] = new Thread() {
|
|
public void run() {
|
|
DigestURL u;
|
|
try {
|
|
while ((u = in.take()) != DigestURL.POISON) {
|
|
// try to load the robots
|
|
RobotsTxtEntry robotsEntry = getEntry(u, userAgent);
|
|
boolean robotsAllowed = robotsEntry == null ? true : !robotsEntry.isDisallowed(u);
|
|
if (robotsAllowed) try {
|
|
Request request = loader.request(u, true, false);
|
|
Response response = loader.load(request, CacheStrategy.NOCACHE, BlacklistType.CRAWLER, userAgent);
|
|
out.put(new CheckEntry(u, robotsEntry, response, null));
|
|
} catch (final IOException e) {
|
|
out.put(new CheckEntry(u, robotsEntry, null, "error response: " + e.getMessage()));
|
|
}
|
|
}
|
|
} catch (InterruptedException e) {}
|
|
}
|
|
};
|
|
threads[i].start();
|
|
}
|
|
// wait for termiation
|
|
try {for (Thread t: threads) t.join();} catch (InterruptedException e1) {}
|
|
return out;
|
|
}
|
|
}
|