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.
367 lines
16 KiB
367 lines
16 KiB
// WorkTables.java
|
|
// (C) 2010 by Michael Peter Christen; mc@yacy.net, Frankfurt a. M., Germany
|
|
// first published 04.02.2010 on http://yacy.net
|
|
//
|
|
// This is a part of YaCy, a peer-to-peer based web search engine
|
|
//
|
|
// $LastChangedDate$
|
|
// $LastChangedRevision$
|
|
// $LastChangedBy$
|
|
//
|
|
// 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.data;
|
|
|
|
import java.io.File;
|
|
import java.io.IOException;
|
|
import java.util.ArrayList;
|
|
import java.util.Collection;
|
|
import java.util.Date;
|
|
import java.util.Iterator;
|
|
import java.util.LinkedHashMap;
|
|
import java.util.Map;
|
|
import java.util.TreeMap;
|
|
|
|
import net.yacy.cora.date.GenericFormatter;
|
|
import net.yacy.cora.document.UTF8;
|
|
import net.yacy.cora.protocol.http.HTTPClient;
|
|
import net.yacy.kelondro.blob.Tables;
|
|
import net.yacy.kelondro.data.meta.DigestURI;
|
|
import net.yacy.kelondro.data.word.WordReference;
|
|
import net.yacy.kelondro.index.HandleSet;
|
|
import net.yacy.kelondro.index.RowSpaceExceededException;
|
|
import net.yacy.kelondro.logging.Log;
|
|
import net.yacy.kelondro.order.Base64Order;
|
|
import net.yacy.kelondro.rwi.IndexCell;
|
|
import de.anomic.data.ymark.YMarkTables;
|
|
import de.anomic.search.Switchboard;
|
|
import de.anomic.server.serverObjects;
|
|
|
|
public class WorkTables extends Tables {
|
|
|
|
public final static String TABLE_API_NAME = "api";
|
|
public final static String TABLE_API_TYPE_STEERING = "steering";
|
|
public final static String TABLE_API_TYPE_CONFIGURATION = "configuration";
|
|
public final static String TABLE_API_TYPE_CRAWLER = "crawler";
|
|
|
|
public final static String TABLE_API_COL_TYPE = "type";
|
|
public final static String TABLE_API_COL_COMMENT = "comment";
|
|
public final static String TABLE_API_COL_DATE_RECORDING = "date_recording"; // if not present default to old date field
|
|
public final static String TABLE_API_COL_DATE_LAST_EXEC = "date_last_exec"; // if not present default to old date field
|
|
public final static String TABLE_API_COL_DATE_NEXT_EXEC = "date_next_exec"; // if not present default to zero
|
|
public final static String TABLE_API_COL_DATE = "date"; // old date; do not set in new records
|
|
public final static String TABLE_API_COL_URL = "url";
|
|
public final static String TABLE_API_COL_APICALL_PK = "apicall_pk"; // the primary key for the table entry of that api call (not really a database field, only a name in the apicall)
|
|
public final static String TABLE_API_COL_APICALL_COUNT = "apicall_count"; // counts how often the API was called (starts with 1)
|
|
public final static String TABLE_API_COL_APICALL_SCHEDULE_TIME = "apicall_schedule_time"; // factor for SCHEULE_UNIT time units
|
|
public final static String TABLE_API_COL_APICALL_SCHEDULE_UNIT= "apicall_schedule_unit"; // may be 'minutes', 'hours', 'days'
|
|
|
|
public final static String TABLE_ROBOTS_NAME = "robots";
|
|
|
|
public final static String TABLE_ACTIVECRAWLS_NAME = "crawljobsActive";
|
|
public final static String TABLE_PASSIVECRAWLS_NAME = "crawljobsPassive";
|
|
|
|
public final static String TABLE_SEARCH_FAILURE_NAME = "searchfl";
|
|
public final static String TABLE_SEARCH_FAILURE_COL_URL = "url";
|
|
public final static String TABLE_SEARCH_FAILURE_COL_DATE = "date";
|
|
public final static String TABLE_SEARCH_FAILURE_COL_WORDS = "words";
|
|
public final static String TABLE_SEARCH_FAILURE_COL_COMMENT = "comment";
|
|
|
|
public YMarkTables bookmarks;
|
|
|
|
public WorkTables(final File workPath) {
|
|
super(workPath, 12);
|
|
this.bookmarks = new YMarkTables(this);
|
|
}
|
|
|
|
/**
|
|
* recording of a api call. stores the call parameters into the API database table
|
|
* @param post the post arguments of the api call
|
|
* @param servletName the name of the servlet
|
|
* @param type name of the servlet category
|
|
* @param comment visual description of the process
|
|
* @return the pk of the new entry in the api table
|
|
*/
|
|
public byte[] recordAPICall(final serverObjects post, final String servletName, final String type, final String comment) {
|
|
// remove the apicall attributes from the post object
|
|
String pks = post.remove(TABLE_API_COL_APICALL_PK);
|
|
byte[] pk = pks == null ? null : UTF8.getBytes(pks);
|
|
|
|
// generate the apicall url - without the apicall attributes
|
|
final String apiurl = /*"http://localhost:" + getConfig("port", "8090") +*/ "/" + servletName + "?" + post.toString();
|
|
|
|
// read old entry from the apicall table (if exists)
|
|
Row row = null;
|
|
try {
|
|
row = (pk == null) ? null : super.select(TABLE_API_NAME, pk);
|
|
} catch (IOException e) {
|
|
Log.logException(e);
|
|
} catch (RowSpaceExceededException e) {
|
|
Log.logException(e);
|
|
}
|
|
|
|
// insert or update entry
|
|
try {
|
|
if (row == null) {
|
|
// create and insert new entry
|
|
Data data = new Data();
|
|
data.put(TABLE_API_COL_TYPE, UTF8.getBytes(type));
|
|
data.put(TABLE_API_COL_COMMENT, UTF8.getBytes(comment));
|
|
byte[] date = UTF8.getBytes(GenericFormatter.SHORT_MILSEC_FORMATTER.format());
|
|
data.put(TABLE_API_COL_DATE_RECORDING, date);
|
|
data.put(TABLE_API_COL_DATE_LAST_EXEC, date);
|
|
data.put(TABLE_API_COL_URL, UTF8.getBytes(apiurl));
|
|
|
|
// insert APICALL attributes
|
|
data.put(TABLE_API_COL_APICALL_COUNT, "1");
|
|
pk = super.insert(TABLE_API_NAME, data);
|
|
} else {
|
|
// modify and update existing entry
|
|
|
|
// modify date attributes and patch old values
|
|
row.put(TABLE_API_COL_DATE_LAST_EXEC, UTF8.getBytes(GenericFormatter.SHORT_MILSEC_FORMATTER.format()));
|
|
if (!row.containsKey(TABLE_API_COL_DATE_RECORDING)) row.put(TABLE_API_COL_DATE_RECORDING, row.get(TABLE_API_COL_DATE));
|
|
row.remove(TABLE_API_COL_DATE);
|
|
|
|
// insert APICALL attributes
|
|
row.put(TABLE_API_COL_APICALL_COUNT, row.get(TABLE_API_COL_APICALL_COUNT, 1) + 1);
|
|
super.update(TABLE_API_NAME, row);
|
|
assert pk != null;
|
|
}
|
|
} catch (IOException e) {
|
|
Log.logException(e);
|
|
} catch (RowSpaceExceededException e) {
|
|
Log.logException(e);
|
|
}
|
|
Log.logInfo("APICALL", apiurl);
|
|
return pk;
|
|
}
|
|
|
|
/**
|
|
* store a API call and set attributes to schedule a re-call of that API call according to a given frequence
|
|
* This is the same as the previous method but it also computes a re-call time and stores that additionally
|
|
* @param post the post arguments of the api call
|
|
* @param servletName the name of the servlet
|
|
* @param type name of the servlet category
|
|
* @param comment visual description of the process
|
|
* @param time the time until next scheduled execution of this api call
|
|
* @param unit the time unit for the scheduled call
|
|
* @return the pk of the new entry in the api table
|
|
*/
|
|
public byte[] recordAPICall(final serverObjects post, final String servletName, final String type, final String comment, int time, String unit) {
|
|
if (post.containsKey(TABLE_API_COL_APICALL_PK)) {
|
|
// this api call has already been stored somewhere.
|
|
return recordAPICall(post, servletName, type, comment);
|
|
}
|
|
if (time < 0 || unit == null || unit.length() == 0 || "minutes,hours,days".indexOf(unit) < 0) {
|
|
time = 0; unit = "";
|
|
} else {
|
|
if (unit.equals("minutes") && time < 10) time = 10;
|
|
}
|
|
|
|
// generate the apicall url - without the apicall attributes
|
|
final String apiurl = /*"http://localhost:" + getConfig("port", "8090") +*/ "/" + servletName + "?" + post.toString();
|
|
byte[] pk = null;
|
|
// insert entry
|
|
try {
|
|
// create and insert new entry
|
|
Data data = new Data();
|
|
data.put(TABLE_API_COL_TYPE, UTF8.getBytes(type));
|
|
data.put(TABLE_API_COL_COMMENT, UTF8.getBytes(comment));
|
|
byte[] date = UTF8.getBytes(GenericFormatter.SHORT_MILSEC_FORMATTER.format());
|
|
data.put(TABLE_API_COL_DATE_RECORDING, date);
|
|
data.put(TABLE_API_COL_DATE_LAST_EXEC, date);
|
|
data.put(TABLE_API_COL_URL, UTF8.getBytes(apiurl));
|
|
|
|
// insert APICALL attributes
|
|
data.put(TABLE_API_COL_APICALL_COUNT, UTF8.getBytes("1"));
|
|
data.put(TABLE_API_COL_APICALL_SCHEDULE_TIME, UTF8.getBytes(Integer.toString(time)));
|
|
data.put(TABLE_API_COL_APICALL_SCHEDULE_UNIT, UTF8.getBytes(unit));
|
|
calculateAPIScheduler(data, false); // set next execution time
|
|
pk = super.insert(TABLE_API_NAME, data);
|
|
} catch (IOException e) {
|
|
Log.logException(e);
|
|
} catch (RowSpaceExceededException e) {
|
|
Log.logException(e);
|
|
}
|
|
Log.logInfo("APICALL", apiurl);
|
|
return pk;
|
|
}
|
|
|
|
/**
|
|
* execute an API call using a api table row which contains all essentials
|
|
* to access the server also the host, port and the authentication realm must be given
|
|
* @param pks a collection of primary keys denoting the rows in the api table
|
|
* @param host the host where the api shall be called
|
|
* @param port the port on the host
|
|
* @param realm authentification realm
|
|
* @return a map of the called urls and the http status code of the api call or -1 if any other IOException occurred
|
|
*/
|
|
public Map<String, Integer> execAPICalls(String host, int port, String realm, Collection<String> pks) {
|
|
// now call the api URLs and store the result status
|
|
final HTTPClient client = new HTTPClient();
|
|
client.setRealm(realm);
|
|
client.setTimout(120000);
|
|
LinkedHashMap<String, Integer> l = new LinkedHashMap<String, Integer>();
|
|
for (String pk: pks) {
|
|
Tables.Row row = null;
|
|
try {
|
|
row = select(WorkTables.TABLE_API_NAME, UTF8.getBytes(pk));
|
|
} catch (IOException e) {
|
|
Log.logException(e);
|
|
} catch (RowSpaceExceededException e) {
|
|
Log.logException(e);
|
|
}
|
|
if (row == null) continue;
|
|
String url = "http://" + host + ":" + port + UTF8.String(row.get(WorkTables.TABLE_API_COL_URL));
|
|
url += "&" + WorkTables.TABLE_API_COL_APICALL_PK + "=" + UTF8.String(row.getPK());
|
|
Log.logInfo("WorkTables", "executing url: " + url);
|
|
try {
|
|
client.GETbytes(url);
|
|
l.put(url, client.getStatusCode());
|
|
} catch (IOException e) {
|
|
Log.logException(e);
|
|
l.put(url, -1);
|
|
}
|
|
}
|
|
return l;
|
|
}
|
|
|
|
public static int execAPICall(String host, int port, String realm, String path, byte[] pk) {
|
|
// now call the api URLs and store the result status
|
|
final HTTPClient client = new HTTPClient();
|
|
client.setRealm(realm);
|
|
client.setTimout(120000);
|
|
String url = "http://" + host + ":" + port + path;
|
|
if (pk != null) url += "&" + WorkTables.TABLE_API_COL_APICALL_PK + "=" + UTF8.String(pk);
|
|
try {
|
|
client.GETbytes(url);
|
|
return client.getStatusCode();
|
|
} catch (IOException e) {
|
|
Log.logException(e);
|
|
return -1;
|
|
}
|
|
}
|
|
|
|
/**
|
|
* simplified call to execute a single entry in the api database table
|
|
* @param pk the primary key of the entry
|
|
* @param host the host where the api shall be called
|
|
* @param port the port on the host
|
|
* @param realm authentification realm
|
|
* @return the http status code of the api call or -1 if any other IOException occurred
|
|
*/
|
|
public int execAPICall(String pk, String host, int port, String realm) {
|
|
ArrayList<String> pks = new ArrayList<String>();
|
|
pks.add(pk);
|
|
Map<String, Integer> m = execAPICalls(host, port, realm, pks);
|
|
if (m.isEmpty()) return -1;
|
|
return m.values().iterator().next().intValue();
|
|
}
|
|
|
|
/**
|
|
* calculate the execution time in a api call table based on given scheduling time and last execution time
|
|
* @param row the database row in the api table
|
|
* @param update if true then the next execution time is based on the latest computed execution time; othervise it is based on the last execution time
|
|
*/
|
|
public static void calculateAPIScheduler(Tables.Data row, boolean update) {
|
|
Date date = row.containsKey(WorkTables.TABLE_API_COL_DATE) ? row.get(WorkTables.TABLE_API_COL_DATE, (Date) null) : null;
|
|
date = update ? row.get(WorkTables.TABLE_API_COL_DATE_NEXT_EXEC, date) : row.get(WorkTables.TABLE_API_COL_DATE_LAST_EXEC, date);
|
|
int time = row.get(WorkTables.TABLE_API_COL_APICALL_SCHEDULE_TIME, 1);
|
|
if (time <= 0) {
|
|
row.put(WorkTables.TABLE_API_COL_DATE_NEXT_EXEC, "");
|
|
return;
|
|
}
|
|
String unit = row.get(WorkTables.TABLE_API_COL_APICALL_SCHEDULE_UNIT, "days");
|
|
long d = date.getTime();
|
|
if (unit.equals("minutes")) d += 60000L * Math.max(10, time);
|
|
if (unit.equals("hours")) d += 60000L * 60L * time;
|
|
if (unit.equals("days")) d += 60000L * 60L * 24L * time;
|
|
if (d < System.currentTimeMillis()) d = System.currentTimeMillis() + 600000L;
|
|
d -= d % 60000; // remove seconds
|
|
row.put(WorkTables.TABLE_API_COL_DATE_NEXT_EXEC, new Date(d));
|
|
}
|
|
|
|
public void failURLsRegisterMissingWord(IndexCell<WordReference> indexCell, final DigestURI url, HandleSet queryHashes, final String reason) {
|
|
|
|
// remove words from index
|
|
for (byte[] word: queryHashes) {
|
|
indexCell.removeDelayed(word, url.hash());
|
|
}
|
|
|
|
// insert information about changed url into database
|
|
try {
|
|
// create and insert new entry
|
|
Data data = new Data();
|
|
byte[] date = UTF8.getBytes(GenericFormatter.SHORT_MILSEC_FORMATTER.format());
|
|
data.put(TABLE_SEARCH_FAILURE_COL_URL, url.toNormalform(true, false));
|
|
data.put(TABLE_SEARCH_FAILURE_COL_DATE, date);
|
|
data.put(TABLE_SEARCH_FAILURE_COL_WORDS, queryHashes.export());
|
|
data.put(TABLE_SEARCH_FAILURE_COL_COMMENT, UTF8.getBytes(reason));
|
|
super.insert(TABLE_SEARCH_FAILURE_NAME, url.hash(), data);
|
|
} catch (IOException e) {
|
|
Log.logException(e);
|
|
}
|
|
}
|
|
|
|
public boolean failURLsContains(byte[] urlhash) {
|
|
try {
|
|
return super.has(TABLE_SEARCH_FAILURE_NAME, urlhash);
|
|
} catch (IOException e) {
|
|
Log.logException(e);
|
|
return false;
|
|
}
|
|
}
|
|
|
|
/**
|
|
* cleanup cached failed searchs older then timeout
|
|
*/
|
|
public void cleanFailURLS(long timeout) {
|
|
if (timeout >= 0) {
|
|
try {
|
|
Iterator<Row> iter = this.iterator(WorkTables.TABLE_SEARCH_FAILURE_NAME);
|
|
while (iter.hasNext()) {
|
|
Row row = iter.next();
|
|
Date date = new Date();
|
|
date = row.get(TABLE_SEARCH_FAILURE_COL_DATE, date);
|
|
if(date.before(new Date(System.currentTimeMillis() - timeout))) {
|
|
this.delete(TABLE_SEARCH_FAILURE_NAME, row.getPK());
|
|
}
|
|
}
|
|
} catch (IOException e) {
|
|
Log.logException(e);
|
|
}
|
|
}
|
|
}
|
|
|
|
public static Map<byte[], String> commentCache(Switchboard sb) {
|
|
Map<byte[], String> comments = new TreeMap<byte[], String>(Base64Order.enhancedCoder);
|
|
Iterator<Tables.Row> i;
|
|
try {
|
|
i = sb.tables.iterator(WorkTables.TABLE_API_NAME);
|
|
Tables.Row row;
|
|
while (i.hasNext()) {
|
|
row = i.next();
|
|
comments.put(row.getPK(), UTF8.String(row.get(WorkTables.TABLE_API_COL_COMMENT)));
|
|
}
|
|
} catch (IOException e) {
|
|
Log.logException(e);
|
|
}
|
|
return comments;
|
|
}
|
|
}
|