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.
560 lines
29 KiB
560 lines
29 KiB
// plasmaCRProcess.java
|
|
// -----------------------
|
|
// part of YaCy
|
|
// (C) by Michael Peter Christen; mc@anomic.de
|
|
// first published on http://www.anomic.de
|
|
// Frankfurt, Germany, 2005
|
|
// Created 15.11.2005
|
|
//
|
|
// $LastChangedDate$
|
|
// $LastChangedRevision$
|
|
// $LastChangedBy$
|
|
//
|
|
// 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.
|
|
|
|
package de.anomic.plasma;
|
|
|
|
import java.io.File;
|
|
import java.io.IOException;
|
|
import java.util.Iterator;
|
|
|
|
import de.anomic.kelondro.kelondroAttrSeq;
|
|
import de.anomic.kelondro.kelondroBase64Order;
|
|
import de.anomic.kelondro.kelondroBitfield;
|
|
import de.anomic.kelondro.kelondroCollectionIndex;
|
|
import de.anomic.kelondro.kelondroFlexTable;
|
|
import de.anomic.kelondro.kelondroIndex;
|
|
import de.anomic.kelondro.kelondroRow;
|
|
import de.anomic.kelondro.kelondroRowSet;
|
|
import de.anomic.server.serverDate;
|
|
import de.anomic.server.serverFileUtils;
|
|
import de.anomic.server.serverMemory;
|
|
|
|
public class plasmaRankingCRProcess {
|
|
|
|
/*
|
|
header.append("# Name=YaCy " + ((type.equals("crl")) ? "Local" : "Global") + " Citation Reference Ticket"); header.append((char) 13); header.append((char) 10);
|
|
header.append("# Created=" + System.currentTimeMillis()); header.append((char) 13); header.append((char) 10);
|
|
header.append("# Structure=<Referee-12>,'=',<UDate-3>,<VDate-3>,<LCount-2>,<GCount-2>,<ICount-2>,<DCount-2>,<TLength-3>,<WACount-3>,<WUCount-3>,<Flags-1>,'|',*<Anchor-" + ((type.equals("crl")) ? "6" : "12") + ">"); header.append((char) 13); header.append((char) 10);
|
|
header.append("# ---"); header.append((char) 13); header.append((char) 10);
|
|
*/
|
|
|
|
public static final kelondroRow CRG_accrow = new kelondroRow(
|
|
"byte[] Referee-12," +
|
|
"Cardinal UDate-3 {b64e}, Cardinal VDate-3 {b64e}, " +
|
|
"Cardinal LCount-2 {b64e}, Cardinal GCount-2 {b64e}, Cardinal ICount-2 {b64e}, Cardinal DCount-2 {b64e}, Cardinal TLength-3 {b64e}, " +
|
|
"Cardinal WACount-3 {b64e}, Cardinal WUCount-3 {b64e}, Cardinal Flags-1 {b64e}, " +
|
|
"Cardinal FUDate-3 {b64e}, Cardinal FDDate-3 {b64e}, Cardinal LUDate-3 {b64e}, " +
|
|
"Cardinal UCount-2 {b64e}, Cardinal PCount-2 {b64e}, Cardinal ACount-2 {b64e}, Cardinal VCount-2 {b64e}, Cardinal Vita-2 {b64e}",
|
|
kelondroBase64Order.enhancedCoder, 0);
|
|
public static final kelondroRow CRG_colrow = new kelondroRow("byte[] Anchor-12", kelondroBase64Order.enhancedCoder, 0);
|
|
public static final String CRG_accname = "CRG-a-attr";
|
|
public static final String CRG_seqname = "CRG-a-coli";
|
|
public static final kelondroRow RCI_coli = new kelondroRow("byte[] RefereeDom-6", kelondroBase64Order.enhancedCoder, 0);
|
|
public static final String RCI_colname = "RCI-a-coli";
|
|
|
|
private static boolean accumulate_upd(File f, kelondroAttrSeq acc) {
|
|
// open file
|
|
kelondroAttrSeq source_cr = null;
|
|
try {
|
|
source_cr = new kelondroAttrSeq(f, false);
|
|
} catch (IOException e) {
|
|
return false;
|
|
}
|
|
|
|
// put elements in accumulator file
|
|
Iterator<String> el = source_cr.keys();
|
|
String key;
|
|
kelondroAttrSeq.Entry new_entry, acc_entry;
|
|
int FUDate, FDDate, LUDate, UCount, PCount, ACount, VCount, Vita;
|
|
kelondroBitfield acc_flags, new_flags;
|
|
while (el.hasNext()) {
|
|
key = (String) el.next();
|
|
new_entry = source_cr.getEntry(key);
|
|
new_flags = new kelondroBitfield(kelondroBase64Order.enhancedCoder.encodeLong((long) new_entry.getAttr("Flags", 0), 1).getBytes());
|
|
// enrich information with additional values
|
|
if ((acc_entry = acc.getEntry(key)) != null) {
|
|
FUDate = (int) acc_entry.getAttr("FUDate", 0);
|
|
FDDate = (int) acc_entry.getAttr("FDDate", 0);
|
|
LUDate = (int) acc_entry.getAttr("LUDate", 0);
|
|
UCount = (int) acc_entry.getAttr("UCount", 0);
|
|
PCount = (int) acc_entry.getAttr("PCount", 0);
|
|
ACount = (int) acc_entry.getAttr("ACount", 0);
|
|
VCount = (int) acc_entry.getAttr("VCount", 0);
|
|
Vita = (int) acc_entry.getAttr("Vita", 0);
|
|
|
|
// update counters and dates
|
|
acc_entry.setSeq(new_entry.getSeqSet()); // need to be checked
|
|
|
|
UCount++; // increase update counter
|
|
PCount += (new_flags.get(1)) ? 1 : 0;
|
|
ACount += (new_flags.get(2)) ? 1 : 0;
|
|
VCount += (new_flags.get(3)) ? 1 : 0;
|
|
|
|
// 'OR' the flags
|
|
acc_flags = new kelondroBitfield(kelondroBase64Order.enhancedCoder.encodeLong((long) acc_entry.getAttr("Flags", 0), 1).getBytes());
|
|
for (int i = 0; i < 6; i++) {
|
|
if (new_flags.get(i)) acc_flags.set(i, true);
|
|
}
|
|
acc_entry.setAttr("Flags", (int) kelondroBase64Order.enhancedCoder.decodeLong(acc_flags.exportB64()));
|
|
} else {
|
|
// initialize counters and dates
|
|
acc_entry = acc.newEntry(key, new_entry.getAttrs(), new_entry.getSeqSet());
|
|
FUDate = plasmaWordIndex.microDateHoursInt(System.currentTimeMillis()); // first update date
|
|
FDDate = plasmaWordIndex.microDateHoursInt(System.currentTimeMillis()); // very difficult to compute; this is only a quick-hack
|
|
LUDate = (int) new_entry.getAttr("VDate", 0);
|
|
UCount = 0;
|
|
PCount = (new_flags.get(1)) ? 1 : 0;
|
|
ACount = (new_flags.get(2)) ? 1 : 0;
|
|
VCount = (new_flags.get(3)) ? 1 : 0;
|
|
Vita = 0;
|
|
}
|
|
// make plausibility check?
|
|
|
|
// insert into accumulator
|
|
acc_entry.setAttr("FUDate", (long) FUDate);
|
|
acc_entry.setAttr("FDDate", (long) FDDate);
|
|
acc_entry.setAttr("LUDate", (long) LUDate);
|
|
acc_entry.setAttr("UCount", (long) UCount);
|
|
acc_entry.setAttr("PCount", (long) PCount);
|
|
acc_entry.setAttr("ACount", (long) ACount);
|
|
acc_entry.setAttr("VCount", (long) VCount);
|
|
acc_entry.setAttr("Vita", (long) Vita);
|
|
acc.putEntrySmall(acc_entry);
|
|
}
|
|
|
|
return true;
|
|
}
|
|
|
|
private static boolean accumulate_upd(File f, kelondroIndex acc, kelondroCollectionIndex seq) throws IOException {
|
|
// open file
|
|
kelondroAttrSeq source_cr = null;
|
|
try {
|
|
source_cr = new kelondroAttrSeq(f, false);
|
|
} catch (IOException e) {
|
|
return false;
|
|
}
|
|
|
|
// put elements in accumulator file
|
|
Iterator<String> el = source_cr.keys();
|
|
String key;
|
|
kelondroAttrSeq.Entry new_entry;
|
|
kelondroRow.Entry acc_entry;
|
|
int FUDate, FDDate, LUDate, UCount, PCount, ACount, VCount, Vita;
|
|
kelondroBitfield acc_flags, new_flags;
|
|
while (el.hasNext()) {
|
|
key = (String) el.next();
|
|
new_entry = source_cr.getEntry(key);
|
|
new_flags = new kelondroBitfield(kelondroBase64Order.enhancedCoder.encodeLong((long) new_entry.getAttr("Flags", 0), 1).getBytes());
|
|
// enrich information with additional values
|
|
if ((acc_entry = acc.get(key.getBytes())) != null) {
|
|
FUDate = (int) acc_entry.getColLong("FUDate", 0);
|
|
FDDate = (int) acc_entry.getColLong("FDDate", 0);
|
|
LUDate = (int) acc_entry.getColLong("LUDate", 0);
|
|
UCount = (int) acc_entry.getColLong("UCount", 0);
|
|
PCount = (int) acc_entry.getColLong("PCount", 0);
|
|
ACount = (int) acc_entry.getColLong("ACount", 0);
|
|
VCount = (int) acc_entry.getColLong("VCount", 0);
|
|
Vita = (int) acc_entry.getColLong("Vita", 0);
|
|
|
|
// update counters and dates
|
|
seq.put(key.getBytes(), new_entry.getSeqCollection()); // FIXME: old and new collection must be joined
|
|
|
|
UCount++; // increase update counter
|
|
PCount += (new_flags.get(1)) ? 1 : 0;
|
|
ACount += (new_flags.get(2)) ? 1 : 0;
|
|
VCount += (new_flags.get(3)) ? 1 : 0;
|
|
|
|
// 'OR' the flags
|
|
acc_flags = new kelondroBitfield(kelondroBase64Order.enhancedCoder.encodeLong(acc_entry.getColLong("Flags", 0), 1).getBytes());
|
|
for (int i = 0; i < 6; i++) {
|
|
if (new_flags.get(i)) acc_flags.set(i, true);
|
|
}
|
|
acc_entry.setCol("Flags", (int) kelondroBase64Order.enhancedCoder.decodeLong(acc_flags.exportB64()));
|
|
} else {
|
|
// initialize counters and dates
|
|
acc_entry = acc.row().newEntry();
|
|
acc_entry.setCol("Referee", key, null);
|
|
for (int i = 1; i < acc.row().columns(); i++) {
|
|
acc_entry.setCol(i, new_entry.getAttr(acc.row().column(i).nickname, 0));
|
|
}
|
|
seq.put(key.getBytes(), new_entry.getSeqCollection());
|
|
FUDate = plasmaWordIndex.microDateHoursInt(System.currentTimeMillis()); // first update date
|
|
FDDate = plasmaWordIndex.microDateHoursInt(System.currentTimeMillis()); // very difficult to compute; this is only a quick-hack
|
|
LUDate = (int) new_entry.getAttr("VDate", 0);
|
|
UCount = 0;
|
|
PCount = (new_flags.get(1)) ? 1 : 0;
|
|
ACount = (new_flags.get(2)) ? 1 : 0;
|
|
VCount = (new_flags.get(3)) ? 1 : 0;
|
|
Vita = 0;
|
|
}
|
|
// make plausibility check?
|
|
|
|
// insert into accumulator
|
|
acc_entry.setCol("FUDate", (long) FUDate);
|
|
acc_entry.setCol("FDDate", (long) FDDate);
|
|
acc_entry.setCol("LUDate", (long) LUDate);
|
|
acc_entry.setCol("UCount", (long) UCount);
|
|
acc_entry.setCol("PCount", (long) PCount);
|
|
acc_entry.setCol("ACount", (long) ACount);
|
|
acc_entry.setCol("VCount", (long) VCount);
|
|
acc_entry.setCol("Vita", (long) Vita);
|
|
acc.put(acc_entry);
|
|
}
|
|
|
|
return true;
|
|
}
|
|
|
|
public static void accumulate(File from_dir, File tmp_dir, File err_dir, File bkp_dir, File to_file, int max_files, boolean newdb) throws IOException {
|
|
if (!(from_dir.isDirectory())) {
|
|
System.out.println("source path " + from_dir + " is not a directory.");
|
|
return;
|
|
}
|
|
if (!(tmp_dir.isDirectory())) {
|
|
System.out.println("temporary path " + tmp_dir + " is not a directory.");
|
|
return;
|
|
}
|
|
if (!(err_dir.isDirectory())) {
|
|
System.out.println("error path " + err_dir + " is not a directory.");
|
|
return;
|
|
}
|
|
if (!(bkp_dir.isDirectory())) {
|
|
System.out.println("back-up path " + bkp_dir + " is not a directory.");
|
|
return;
|
|
}
|
|
|
|
// open target file
|
|
kelondroAttrSeq acc = null;
|
|
kelondroIndex newacc = null;
|
|
kelondroCollectionIndex newseq = null;
|
|
if (newdb) {
|
|
File path = to_file.getParentFile(); // path to storage place
|
|
newacc = new kelondroFlexTable(path, CRG_accname, CRG_accrow, 0, false);
|
|
newseq = new kelondroCollectionIndex(path, CRG_seqname, 12, kelondroBase64Order.enhancedCoder, 2, 9, CRG_colrow);
|
|
} else {
|
|
if (!(to_file.exists())) {
|
|
acc = new kelondroAttrSeq("Global Ranking Accumulator File",
|
|
"<Referee-12>,'='," +
|
|
"<UDate-3>,<VDate-3>,<LCount-2>,<GCount-2>,<ICount-2>,<DCount-2>,<TLength-3>,<WACount-3>,<WUCount-3>,<Flags-1>," +
|
|
"<FUDate-3>,<FDDate-3>,<LUDate-3>,<UCount-2>,<PCount-2>,<ACount-2>,<VCount-2>,<Vita-2>," +
|
|
"'|',*<Anchor-12>", false);
|
|
acc.toFile(to_file);
|
|
}
|
|
acc = new kelondroAttrSeq(to_file, false);
|
|
}
|
|
// collect source files
|
|
File source_file = null;
|
|
String[] files = from_dir.list();
|
|
if (files.length < max_files) max_files = files.length;
|
|
for (int i = 0; i < max_files; i++) {
|
|
// open file
|
|
source_file = new File(from_dir, files[i]);
|
|
if (newdb) {
|
|
if (accumulate_upd(source_file, newacc, newseq)) {
|
|
// move CR file to temporary folder
|
|
source_file.renameTo(new File(tmp_dir, files[i]));
|
|
} else {
|
|
// error case: the CR-file is not valid; move to error path
|
|
source_file.renameTo(new File(err_dir, files[i]));
|
|
}
|
|
} else {
|
|
if (accumulate_upd(source_file, acc)) {
|
|
// move CR file to temporary folder
|
|
source_file.renameTo(new File(tmp_dir, files[i]));
|
|
} else {
|
|
// error case: the CR-file is not valid; move to error path
|
|
source_file.renameTo(new File(err_dir, files[i]));
|
|
}
|
|
}
|
|
}
|
|
|
|
try {
|
|
if (newdb) {
|
|
newacc.close();
|
|
newseq.close();
|
|
} else {
|
|
// save accumulator to temporary file
|
|
File tmp_file;
|
|
if (to_file.toString().endsWith(".gz")) {
|
|
tmp_file = new File(to_file.toString() + "." + (System.currentTimeMillis() % 1000) + ".tmp.gz");
|
|
} else {
|
|
tmp_file = new File(to_file.toString() + "." + (System.currentTimeMillis() % 1000) + ".tmp");
|
|
}
|
|
// store the file
|
|
acc.toFile(tmp_file);
|
|
// since this was successful, we remove the old file and move the new file to it
|
|
to_file.delete();
|
|
tmp_file.renameTo(to_file);
|
|
}
|
|
serverFileUtils.moveAll(tmp_dir, bkp_dir);
|
|
} catch (IOException e) {
|
|
// move previously processed files back
|
|
e.printStackTrace();
|
|
serverFileUtils.moveAll(tmp_dir, from_dir);
|
|
}
|
|
|
|
}
|
|
|
|
public static int genrci(File cr_in, File rci_out) throws IOException {
|
|
if (!(cr_in.exists())) return 0;
|
|
kelondroAttrSeq cr = new kelondroAttrSeq(cr_in, false);
|
|
//if (rci_out.exists()) rci_out.delete(); // we want only fresh rci here (during testing)
|
|
if (!(rci_out.exists())) {
|
|
kelondroAttrSeq rcix = new kelondroAttrSeq("Global Ranking Reverse Citation Index",
|
|
"<AnchorDom-6>,'='," +
|
|
"<UDate-3>," +
|
|
"'|',*<Referee-12>", false);
|
|
rcix.toFile(rci_out);
|
|
}
|
|
final kelondroAttrSeq rci = new kelondroAttrSeq(rci_out, false);
|
|
|
|
// loop over all referees
|
|
int count = 0;
|
|
int size = cr.size();
|
|
long start = System.currentTimeMillis();
|
|
long l;
|
|
final Iterator<String> i = cr.keys();
|
|
String referee, anchor, anchorDom;
|
|
kelondroAttrSeq.Entry cr_entry, rci_entry;
|
|
long cr_UDate, rci_UDate;
|
|
while (i.hasNext()) {
|
|
referee = i.next();
|
|
cr_entry = cr.getEntry(referee);
|
|
cr_UDate = cr_entry.getAttr("UDate", 0);
|
|
|
|
// loop over all anchors
|
|
Iterator<String> j = cr_entry.getSeqSet().iterator();
|
|
while (j.hasNext()) {
|
|
// get domain of anchors
|
|
anchor = j.next();
|
|
if (anchor.length() == 6) anchorDom = anchor; else anchorDom = anchor.substring(6);
|
|
|
|
// update domain-specific entry
|
|
rci_entry = rci.getEntry(anchorDom);
|
|
if (rci_entry == null) rci_entry = rci.newEntry(anchorDom, false);
|
|
rci_entry.addSeq(referee);
|
|
|
|
// update Update-Date
|
|
rci_UDate = rci_entry.getAttr("UDate", 0);
|
|
if (cr_UDate > rci_UDate) rci_entry.setAttr("UDate", cr_UDate);
|
|
|
|
// insert entry
|
|
rci.putEntry(rci_entry);
|
|
}
|
|
count++;
|
|
if ((count % 1000) == 0) {
|
|
l = java.lang.Math.max(1, (System.currentTimeMillis() - start) / 1000);
|
|
System.out.println("processed " + count + " citations, " + (count / l) + " per second, rci.size = " + rci.size() + ", " + ((size - count) / (count / l)) + " seconds remaining; mem = " + serverMemory.available());
|
|
}
|
|
i.remove();
|
|
}
|
|
|
|
// finished. write to file
|
|
cr = null;
|
|
cr_in = null;
|
|
serverMemory.gc(1000, "plasmaRankingCRProcess.genrci(...)"); // thq
|
|
rci.toFile(rci_out);
|
|
return count;
|
|
}
|
|
|
|
public static int genrcix(File cr_path_in, File rci_path_out) throws IOException {
|
|
//kelondroFlexTable acc = new kelondroFlexTable(cr_path_in, CRG_accname, kelondroBase64Order.enhancedCoder, 128 * 1024 * 1024, -1, CRG_accrow, true);
|
|
kelondroCollectionIndex seq = new kelondroCollectionIndex(cr_path_in, CRG_seqname, 12, kelondroBase64Order.enhancedCoder, 2, 9, CRG_colrow);
|
|
kelondroCollectionIndex rci = new kelondroCollectionIndex(rci_path_out, RCI_colname, 6, kelondroBase64Order.enhancedCoder, 2, 9, RCI_coli);
|
|
|
|
// loop over all referees
|
|
int count = 0;
|
|
int size = seq.size();
|
|
long start = System.currentTimeMillis();
|
|
long l;
|
|
final Iterator<Object[]> i = seq.keycollections(null, null, false);
|
|
Object[] keycollection;
|
|
String referee, refereeDom, anchor, anchorDom;
|
|
kelondroRowSet cr_entry, rci_entry;
|
|
while (i.hasNext()) {
|
|
keycollection = i.next();
|
|
referee = new String((byte[]) keycollection[0]);
|
|
if (referee.length() == 6) refereeDom = referee; else refereeDom = referee.substring(6);
|
|
cr_entry = (kelondroRowSet) keycollection[1];
|
|
|
|
// loop over all anchors
|
|
Iterator<kelondroRow.Entry> j = cr_entry.rows();
|
|
kelondroRow.Entry entry;
|
|
while (j.hasNext()) {
|
|
// get domain of anchors
|
|
entry = j.next();
|
|
anchor = (String) entry.getColString(0, null);
|
|
if (anchor.length() == 6) anchorDom = anchor; else anchorDom = anchor.substring(6);
|
|
|
|
// update domain-specific entry
|
|
rci_entry = rci.get(anchorDom.getBytes());
|
|
if (rci_entry == null) rci_entry = new kelondroRowSet(RCI_coli, 0);
|
|
rci_entry.add(refereeDom.getBytes());
|
|
|
|
// insert entry
|
|
rci.put(anchorDom.getBytes(), rci_entry);
|
|
}
|
|
count++;
|
|
if ((count % 1000) == 0) {
|
|
l = java.lang.Math.max(1, (System.currentTimeMillis() - start) / 1000);
|
|
System.out.println("processed " + count + " citations, " + (count / l) + " per second, rci.size = " + rci.size() + ", " + ((size - count) / (count / l) / 60) + " minutes remaining; mem = " + serverMemory.free());
|
|
}
|
|
}
|
|
|
|
// finished. write to file
|
|
seq.close();
|
|
rci.close();
|
|
return count;
|
|
}
|
|
|
|
public static void main(String[] args) {
|
|
// java -classpath source de.anomic.plasma.kelondroPropFile -transcode DATA/RANKING/GLOBAL/CRG-test-unsorted-original.cr DATA/RANKING/GLOBAL/CRG-test-generated.cr
|
|
try {
|
|
if ((args.length == 5) && (args[0].equals("-accumulate"))) {
|
|
accumulate(new File(args[1]), new File(args[2]), new File(args[3]), new File(args[4]), new File(args[5]), Integer.parseInt(args[6]), true);
|
|
}
|
|
if ((args.length == 2) && (args[0].equals("-accumulate"))) {
|
|
File root_path = new File(args[1]);
|
|
File from_dir = new File(root_path, "DATA/RANKING/GLOBAL/014_othercr");
|
|
File ready_dir = new File(root_path, "DATA/RANKING/GLOBAL/015_ready");
|
|
File tmp_dir = new File(root_path, "DATA/RANKING/GLOBAL/016_tmp");
|
|
File err_dir = new File(root_path, "DATA/RANKING/GLOBAL/017_err");
|
|
File acc_dir = new File(root_path, "DATA/RANKING/GLOBAL/018_acc");
|
|
String filename = "CRG-a-" + new serverDate().toShortString(true) + ".cr.gz";
|
|
File to_file = new File(root_path, "DATA/RANKING/GLOBAL/020_con0/" + filename);
|
|
if (!(ready_dir.exists())) ready_dir.mkdirs();
|
|
if (!(tmp_dir.exists())) tmp_dir.mkdirs();
|
|
if (!(err_dir.exists())) err_dir.mkdirs();
|
|
if (!(acc_dir.exists())) acc_dir.mkdirs();
|
|
if (!(to_file.getParentFile().exists())) to_file.getParentFile().mkdirs();
|
|
serverFileUtils.moveAll(from_dir, ready_dir);
|
|
long start = System.currentTimeMillis();
|
|
int files = ready_dir.list().length;
|
|
accumulate(ready_dir, tmp_dir, err_dir, acc_dir, to_file, 1000, true);
|
|
long seconds = java.lang.Math.max(1, (System.currentTimeMillis() - start) / 1000);
|
|
System.out.println("Finished accumulate for " + files + " files in " + seconds + " seconds (" + (files / seconds) + " files/second)");
|
|
}
|
|
if ((args.length == 3) && (args[0].equals("-recycle"))) {
|
|
File root_path = new File(args[1]);
|
|
int max_age_hours = Integer.parseInt(args[2]);
|
|
File own_dir = new File(root_path, "DATA/RANKING/GLOBAL/010_owncr");
|
|
File acc_dir = new File(root_path, "DATA/RANKING/GLOBAL/018_acc");
|
|
File bkp_dir = new File(root_path, "DATA/RANKING/GLOBAL/019_bkp");
|
|
if (!(own_dir.exists())) return;
|
|
if (!(acc_dir.exists())) return;
|
|
if (!(bkp_dir.exists())) bkp_dir.mkdirs();
|
|
String[] list = acc_dir.list();
|
|
long start = System.currentTimeMillis();
|
|
int files = list.length;
|
|
long d;
|
|
File f;
|
|
for (int i = 0; i < list.length; i++) {
|
|
f = new File(acc_dir, list[i]);
|
|
try {
|
|
d = (System.currentTimeMillis() - (new kelondroAttrSeq(f, false)).created()) / 3600000;
|
|
if (d > max_age_hours) {
|
|
// file is considered to be too old, it is not recycled
|
|
System.out.println("file " + f.getName() + " is old (" + d + " hours) and not recycled, only moved to backup");
|
|
f.renameTo(new File(bkp_dir, list[i]));
|
|
} else {
|
|
// file is fresh, it is duplicated and moved to be transferred to other peers again
|
|
System.out.println("file " + f.getName() + " is fresh (" + d + " hours old), recycled and moved to backup");
|
|
serverFileUtils.copy(f, new File(own_dir, list[i]));
|
|
f.renameTo(new File(bkp_dir, list[i]));
|
|
}
|
|
} catch (IOException e) {
|
|
// there is something wrong with this file; delete it
|
|
System.out.println("file " + f.getName() + " is corrupted and deleted");
|
|
f.delete();
|
|
}
|
|
}
|
|
long seconds = java.lang.Math.max(1, (System.currentTimeMillis() - start) / 1000);
|
|
System.out.println("Finished recycling of " + files + " files in " + seconds + " seconds (" + (files / seconds) + " files/second)");
|
|
}
|
|
if ((args.length == 2) && (args[0].equals("-genrci"))) {
|
|
File root_path = new File(args[1]);
|
|
File cr_filedir = new File(root_path, "DATA/RANKING/GLOBAL/020_con0");
|
|
File rci_filedir = new File(root_path, "DATA/RANKING/GLOBAL/030_rci0");
|
|
rci_filedir.mkdirs();
|
|
long start = System.currentTimeMillis();
|
|
int count = genrcix(cr_filedir, rci_filedir);
|
|
long seconds = java.lang.Math.max(1, (System.currentTimeMillis() - start) / 1000);
|
|
System.out.println("Completed RCI generation: " + count + " citation references in " + seconds + " seconds (" + (count / seconds) + " CR-records/second)");
|
|
}
|
|
/*
|
|
if ((args.length == 2) && (args[0].equals("-genrci"))) {
|
|
File root_path = new File(args[1]);
|
|
File cr_filedir = new File(root_path, "DATA/RANKING/GLOBAL/020_con0");
|
|
File rci_file = new File(root_path, "DATA/RANKING/GLOBAL/030_rci0/RCI-0.rci.gz");
|
|
rci_file.getParentFile().mkdirs();
|
|
String[] cr_filenames = cr_filedir.list();
|
|
for (int i = 0; i < cr_filenames.length; i++) {
|
|
long start = System.currentTimeMillis();
|
|
int count = genrci(new File(cr_filedir, cr_filenames[i]), rci_file);
|
|
long seconds = java.lang.Math.max(1, (System.currentTimeMillis() - start) / 1000);
|
|
System.out.println("Completed RCI generation for input file " + cr_filenames[i] + ": " + count + " citation references in " + seconds + " seconds (" + (count / seconds) + " CR-records/second)");
|
|
}
|
|
}
|
|
*/
|
|
} catch (IOException e) {
|
|
e.printStackTrace();
|
|
}
|
|
}
|
|
|
|
/*
|
|
Class-A File format:
|
|
|
|
UDate : latest update timestamp of the URL (as virtual date, hours since epoch)
|
|
VDate : last visit timestamp of the URL (as virtual date, hours since epoch)
|
|
LCount : count of links to local resources
|
|
GCount : count of links to global resources
|
|
ICount : count of links to images (in document)
|
|
DCount : count of links to other documents
|
|
TLength: length of the plain text content (bytes)
|
|
WACount: total number of all words in content
|
|
WUCount: number of unique words in content (removed doubles)
|
|
Flags : Flags (0=update, 1=popularity, 2=attention, 3=vote)
|
|
|
|
Class-a File format is an extension of Class-A plus the following attributes
|
|
FUDate : first update timestamp of the URL
|
|
FDDate : first update timestamp of the domain
|
|
LUDate : latest update timestamp of the URL
|
|
UCount : Update Counter (of 'latest update timestamp')
|
|
PCount : Popularity Counter (proxy clicks)
|
|
ACount : Attention Counter (search result clicks)
|
|
VCount : Votes
|
|
Vita : Vitality (normed number of updates per time)
|
|
*/
|
|
}
|