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/net/yacy/search/query/SnippetProcess.java

648 lines
32 KiB

// SearchEvent.java
// (C) 2005 by Michael Peter Christen; mc@yacy.net, Frankfurt a. M., Germany
// first published 10.10.2005 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 net.yacy.search.query;
import java.util.ArrayList;
import java.util.Iterator;
import java.util.List;
import java.util.regex.Pattern;
import net.yacy.cora.document.ASCII;
import net.yacy.cora.document.Classification;
import net.yacy.cora.document.MultiProtocolURI;
import net.yacy.cora.protocol.ResponseHeader;
import net.yacy.cora.services.federated.solr.SolrConnector;
import net.yacy.cora.services.federated.yacy.CacheStrategy;
import net.yacy.cora.sorting.ScoreMap;
import net.yacy.cora.sorting.WeakPriorityBlockingQueue;
import net.yacy.cora.sorting.WeakPriorityBlockingQueue.Element;
import net.yacy.cora.sorting.WeakPriorityBlockingQueue.ReverseElement;
import net.yacy.document.Condenser;
import net.yacy.kelondro.data.meta.URIMetadataRow;
import net.yacy.kelondro.data.word.Word;
import net.yacy.kelondro.index.HandleSet;
import net.yacy.kelondro.index.RowSpaceExceededException;
import net.yacy.kelondro.logging.Log;
import net.yacy.kelondro.util.MemoryControl;
import net.yacy.peers.SeedDB;
import net.yacy.peers.graphics.ProfilingGraph;
import net.yacy.repository.LoaderDispatcher;
import net.yacy.search.EventTracker;
import net.yacy.search.Switchboard;
import net.yacy.search.index.Segment;
import net.yacy.search.index.SolrField;
import net.yacy.search.snippet.MediaSnippet;
import net.yacy.search.snippet.ResultEntry;
import net.yacy.search.snippet.TextSnippet;
import org.apache.solr.common.SolrDocument;
import org.apache.solr.common.SolrDocumentList;
import de.anomic.crawler.Cache;
import de.anomic.data.WorkTables;
public class SnippetProcess {
private final static int SNIPPET_WORKER_THREADS = Math.max(4, Runtime.getRuntime().availableProcessors() * 2);
// input values
final RWIProcess rankingProcess; // ordered search results, grows dynamically as all the query threads enrich this container
QueryParams query;
private final SeedDB peers;
private final WorkTables workTables;
// result values
protected final LoaderDispatcher loader;
protected Worker[] workerThreads;
protected final WeakPriorityBlockingQueue<ResultEntry> result;
protected final WeakPriorityBlockingQueue<MediaSnippet> images; // container to sort images by size
protected final HandleSet snippetFetchWordHashes; // a set of word hashes that are used to match with the snippets
long urlRetrievalAllTime;
long snippetComputationAllTime;
int taketimeout;
private final boolean deleteIfSnippetFail, remote;
private boolean cleanupState;
public SnippetProcess(
final LoaderDispatcher loader,
final RWIProcess rankedCache,
final QueryParams query,
final SeedDB peers,
final WorkTables workTables,
final int taketimeout,
final boolean deleteIfSnippetFail,
final boolean remote) {
assert query != null;
this.loader = loader;
this.rankingProcess = rankedCache;
this.query = query;
this.peers = peers;
this.workTables = workTables;
this.taketimeout = taketimeout;
this.deleteIfSnippetFail = deleteIfSnippetFail;
this.remote = remote;
this.cleanupState = false;
this.urlRetrievalAllTime = 0;
this.snippetComputationAllTime = 0;
this.result = new WeakPriorityBlockingQueue<ResultEntry>(Math.max(1000, 10 * query.itemsPerPage()), true); // this is the result, enriched with snippets, ranked and ordered by ranking
this.images = new WeakPriorityBlockingQueue<MediaSnippet>(Math.max(1000, 10 * query.itemsPerPage()), true);
// snippets do not need to match with the complete query hashes,
// only with the query minus the stopwords which had not been used for the search
HandleSet filtered;
try {
filtered = HandleSet.joinConstructive(query.queryHashes, Switchboard.stopwordHashes);
} catch (final RowSpaceExceededException e) {
Log.logException(e);
filtered = new HandleSet(query.queryHashes.row().primaryKeyLength, query.queryHashes.comparator(), 0);
}
this.snippetFetchWordHashes = query.queryHashes.clone();
if (filtered != null && !filtered.isEmpty()) {
this.snippetFetchWordHashes.excludeDestructive(Switchboard.stopwordHashes);
}
// start worker threads to fetch urls and snippets
this.workerThreads = null;
deployWorker(Math.min(SNIPPET_WORKER_THREADS, query.itemsPerPage), query.neededResults());
EventTracker.update(EventTracker.EClass.SEARCH, new ProfilingGraph.EventSearch(query.id(true), SearchEvent.Type.SNIPPETFETCH_START, ((this.workerThreads == null) ? "no" : this.workerThreads.length) + " online snippet fetch threads started", 0, 0), false);
}
public void setCleanupState() {
this.cleanupState = true;
}
public long getURLRetrievalTime() {
return this.urlRetrievalAllTime;
}
public long getSnippetComputationTime() {
return this.snippetComputationAllTime;
}
public ResultEntry oneResult(final int item, final long timeout) {
// check if we already retrieved this item
// (happens if a search pages is accessed a second time)
final long finishTime = System.currentTimeMillis() + timeout;
EventTracker.update(EventTracker.EClass.SEARCH, new ProfilingGraph.EventSearch(this.query.id(true), SearchEvent.Type.ONERESULT, "started, item = " + item + ", available = " + this.result.sizeAvailable(), 0, 0), false);
//Log.logInfo("SnippetProcess", "*start method for item = " + item + "; anyWorkerAlive=" + anyWorkerAlive() + "; this.rankingProcess.isAlive() = " + this.rankingProcess.isAlive() + "; this.rankingProcess.feedingIsFinished() = " + this.rankingProcess.feedingIsFinished() + "; this.result.sizeAvailable() = " + this.result.sizeAvailable() + ", this.rankingProcess.sizeQueue() = " + this.rankingProcess.sizeQueue());
// we must wait some time until the first result page is full to get enough elements for ranking
final long waittimeout = System.currentTimeMillis() + 300;
if (this.remote && item < 10 && !this.rankingProcess.feedingIsFinished()) {
// the first 10 results have a very special timing to get most of the remote results ordered
// before they are presented on the first lines .. yes sleeps seem to be bad. but how shall we predict how long other
// peers will take until they respond?
long sleep = item == 0 ? 600 : (10 - item) * 12; // the first result takes the longest time
//Log.logInfo("SnippetProcess", "SLEEP = " + sleep);
try { Thread.sleep(sleep); } catch (final InterruptedException e1) { Log.logException(e1); }
}
int thisRankingQueueSize, lastRankingQueueSize = 0;
if (item < 10) {
while (
((thisRankingQueueSize = this.rankingProcess.sizeQueue()) > 0 || !this.rankingProcess.feedingIsFinished()) &&
(thisRankingQueueSize > lastRankingQueueSize || this.result.sizeAvailable() < item + 1) &&
System.currentTimeMillis() < waittimeout &&
anyWorkerAlive()
) {
// wait a little time to get first results in the search
lastRankingQueueSize = thisRankingQueueSize;
try { Thread.sleep(20); } catch (final InterruptedException e1) {}
}
}
if (this.result.sizeAvailable() > item) {
// we have the wanted result already in the result array .. return that
final ResultEntry re = this.result.element(item).getElement();
EventTracker.update(EventTracker.EClass.SEARCH, new ProfilingGraph.EventSearch(this.query.id(true), SearchEvent.Type.ONERESULT, "prefetched, item = " + item + ", available = " + this.result.sizeAvailable() + ": " + re.urlstring(), 0, 0), false);
return re;
}
// finally wait until enough results are there produced from the snippet fetch process
WeakPriorityBlockingQueue.Element<ResultEntry> entry = null;
while (System.currentTimeMillis() < finishTime) {
//Log.logInfo("SnippetProcess", "item = " + item + "; anyWorkerAlive=" + anyWorkerAlive() + "; this.rankingProcess.isAlive() = " + this.rankingProcess.isAlive() + "; this.rankingProcess.feedingIsFinished() = " + this.rankingProcess.feedingIsFinished() + "; this.result.sizeAvailable() = " + this.result.sizeAvailable() + ", this.rankingProcess.sizeQueue() = " + this.rankingProcess.sizeQueue());
if (!anyWorkerAlive() && !this.rankingProcess.isAlive() && this.result.sizeAvailable() + this.rankingProcess.sizeQueue() <= item && this.rankingProcess.feedingIsFinished()) {
//Log.logInfo("SnippetProcess", "interrupted result fetching; item = " + item + "; this.result.sizeAvailable() = " + this.result.sizeAvailable() + ", this.rankingProcess.sizeQueue() = " + this.rankingProcess.sizeQueue() + "; this.rankingProcess.feedingIsFinished() = " + this.rankingProcess.feedingIsFinished());
break; // the fail case
}
// deploy worker to get more results
if (!anyWorkerAlive()) {
final int neededInclPrefetch = this.query.neededResults() + ((MemoryControl.available() > 100 * 1024 * 1024 && SNIPPET_WORKER_THREADS >= 8) ? this.query.itemsPerPage : 0);
deployWorker(Math.min(SNIPPET_WORKER_THREADS, this.query.itemsPerPage), neededInclPrefetch);
}
try {entry = this.result.element(item, 50);} catch (final InterruptedException e) {break;}
if (entry != null) {
break;
}
}
// finally, if there is something, return the result
if (entry == null) {
EventTracker.update(EventTracker.EClass.SEARCH, new ProfilingGraph.EventSearch(this.query.id(true), SearchEvent.Type.ONERESULT, "not found, item = " + item + ", available = " + this.result.sizeAvailable(), 0, 0), false);
//Log.logInfo("SnippetProcess", "NO ENTRY computed (possible timeout); anyWorkerAlive=" + anyWorkerAlive() + "; this.rankingProcess.isAlive() = " + this.rankingProcess.isAlive() + "; this.rankingProcess.feedingIsFinished() = " + this.rankingProcess.feedingIsFinished() + "; this.result.sizeAvailable() = " + this.result.sizeAvailable() + ", this.rankingProcess.sizeQueue() = " + this.rankingProcess.sizeQueue());
return null;
}
final ResultEntry re = entry.getElement();
EventTracker.update(EventTracker.EClass.SEARCH, new ProfilingGraph.EventSearch(this.query.id(true), SearchEvent.Type.ONERESULT, "retrieved, item = " + item + ", available = " + this.result.sizeAvailable() + ": " + re.urlstring(), 0, 0), false);
if (item == this.query.offset + this.query.itemsPerPage - 1) {
stopAllWorker(); // we don't need more
}
return re;
}
private int resultCounter = 0;
public ResultEntry nextResult() {
final ResultEntry re = oneResult(this.resultCounter, Math.max(3000, this.query.timeout - System.currentTimeMillis()));
this.resultCounter++;
return re;
}
public MediaSnippet oneImage(final int item) {
// always look for a next object if there are way too less
if (this.images.sizeAvailable() <= item + 10) {
fillImagesCache();
}
// check if we already retrieved the item
if (this.images.sizeDrained() > item) {
return this.images.element(item).getElement();
}
// look again if there are not enough for presentation
while (this.images.sizeAvailable() <= item) {
if (fillImagesCache() == 0) {
break;
}
}
if (this.images.sizeAvailable() <= item) {
return null;
}
// now take the specific item from the image stack
return this.images.element(item).getElement();
}
private int fillImagesCache() {
final ResultEntry result = nextResult();
int c = 0;
if (result == null) {
return c;
}
// iterate over all images in the result
final List<MediaSnippet> imagemedia = result.mediaSnippets();
if (imagemedia != null) {
ResponseHeader header;
feedloop: for (final MediaSnippet ms: imagemedia) {
// check cache to see if the mime type of the image url is correct
header = Cache.getResponseHeader(ms.href.hash());
if (header != null) {
// this does not work for all urls since some of them may not be in the cache
if (header.mime().startsWith("text") || header.mime().startsWith("application")) {
continue feedloop;
}
}
this.images.put(new ReverseElement<MediaSnippet>(ms, ms.ranking)); // remove smallest in case of overflow
c++;
//System.out.println("*** image " + UTF8.String(ms.href.hash()) + " images.size = " + images.size() + "/" + images.size());
}
}
return c;
}
public ArrayList<WeakPriorityBlockingQueue.Element<ResultEntry>> completeResults(final long waitingtime) {
final long timeout = System.currentTimeMillis() + waitingtime;
while ( this.result.sizeAvailable() < this.query.neededResults() &&
anyWorkerAlive() &&
System.currentTimeMillis() < timeout) {
try {Thread.sleep(10);} catch (final InterruptedException e) {}
//System.out.println("+++DEBUG-completeResults+++ sleeping " + 200);
}
return this.result.list(Math.min(this.query.neededResults(), this.result.sizeAvailable()));
}
public long postRanking(
final ResultEntry rentry,
final ScoreMap<String> topwords) {
long r = 0;
// for media search: prefer pages with many links
r += rentry.limage() << this.query.ranking.coeff_cathasimage;
r += rentry.laudio() << this.query.ranking.coeff_cathasaudio;
r += rentry.lvideo() << this.query.ranking.coeff_cathasvideo;
r += rentry.lapp() << this.query.ranking.coeff_cathasapp;
// apply citation count
//System.out.println("POSTRANKING CITATION: references = " + rentry.referencesCount() + ", inbound = " + rentry.llocal() + ", outbound = " + rentry.lother());
r += (128 * rentry.referencesCount() / (1 + 2 * rentry.llocal() + rentry.lother())) << this.query.ranking.coeff_citation;
// prefer hit with 'prefer' pattern
if (this.query.prefer.matcher(rentry.url().toNormalform(true, true)).matches()) {
r += 256 << this.query.ranking.coeff_prefer;
}
if (this.query.prefer.matcher(rentry.title()).matches()) {
r += 256 << this.query.ranking.coeff_prefer;
}
// apply 'common-sense' heuristic using references
final String urlstring = rentry.url().toNormalform(true, true);
final String[] urlcomps = MultiProtocolURI.urlComps(urlstring);
final String[] descrcomps = MultiProtocolURI.splitpattern.split(rentry.title().toLowerCase());
int tc;
for (final String urlcomp : urlcomps) {
tc = topwords.get(urlcomp);
if (tc > 0) {
r += Math.max(1, tc) << this.query.ranking.coeff_urlcompintoplist;
}
}
for (final String descrcomp : descrcomps) {
tc = topwords.get(descrcomp);
if (tc > 0) {
r += Math.max(1, tc) << this.query.ranking.coeff_descrcompintoplist;
}
}
// apply query-in-result matching
final HandleSet urlcomph = Word.words2hashesHandles(urlcomps);
final HandleSet descrcomph = Word.words2hashesHandles(descrcomps);
final Iterator<byte[]> shi = this.query.queryHashes.iterator();
byte[] queryhash;
while (shi.hasNext()) {
queryhash = shi.next();
if (urlcomph.has(queryhash)) {
r += 256 << this.query.ranking.coeff_appurl;
}
if (descrcomph.has(queryhash)) {
r += 256 << this.query.ranking.coeff_app_dc_title;
}
}
return r;
}
public void deployWorker(int deployCount, final int neededResults) {
if (this.cleanupState ||
(this.rankingProcess.feedingIsFinished() && this.rankingProcess.sizeQueue() == 0) ||
this.result.sizeAvailable() >= neededResults) {
return;
}
Worker worker;
if (this.workerThreads == null) {
this.workerThreads = new Worker[deployCount];
synchronized(this.workerThreads) {try {
for (int i = 0; i < this.workerThreads.length; i++) {
if (this.result.sizeAvailable() >= neededResults ||
(this.rankingProcess.feedingIsFinished() && this.rankingProcess.sizeQueue() == 0)) {
break;
}
worker = new Worker(i, this.query.maxtime, this.query.snippetCacheStrategy, this.query.snippetMatcher, neededResults);
worker.start();
this.workerThreads[i] = worker;
if (this.rankingProcess.expectMoreRemoteReferences()) {
long wait = this.rankingProcess.waitTimeRecommendation();
if (wait > 0) {
try {Thread.sleep(wait);} catch ( InterruptedException e ) {}
}
}
}
} catch (OutOfMemoryError e) {}}
} else {
// there are still worker threads running, but some may be dead.
// if we find dead workers, reanimate them
synchronized(this.workerThreads) {
for (int i = 0; i < this.workerThreads.length; i++) {
if (deployCount <= 0 ||
this.result.sizeAvailable() >= neededResults ||
(this.rankingProcess.feedingIsFinished() && this.rankingProcess.sizeQueue() == 0)) {
break;
}
if (this.workerThreads[i] == null || !this.workerThreads[i].isAlive()) {
worker = new Worker(i, this.query.maxtime, this.query.snippetCacheStrategy, this.query.snippetMatcher, neededResults);
worker.start();
this.workerThreads[i] = worker;
deployCount--;
}
if (this.rankingProcess.expectMoreRemoteReferences()) {
long wait = this.rankingProcess.waitTimeRecommendation();
if (wait > 0) {
try {Thread.sleep(wait);} catch ( InterruptedException e ) {}
}
}
}
}
}
}
public void stopAllWorker() {
synchronized(this.workerThreads) {
for (int i = 0; i < this.workerThreads.length; i++) {
if (this.workerThreads[i] == null || !this.workerThreads[i].isAlive()) {
continue;
}
this.workerThreads[i].pleaseStop();
this.workerThreads[i].interrupt();
}
}
}
private boolean anyWorkerAlive() {
if (this.workerThreads == null || this.workerThreads.length == 0) {
return false;
}
synchronized(this.workerThreads) {
for (final Worker workerThread : this.workerThreads) {
if ((workerThread != null) &&
(workerThread.isAlive()) &&
(workerThread.busytime() < 10000)) {
return true;
}
}
}
return false;
}
protected class Worker extends Thread {
private final long timeout; // the date until this thread should try to work
private long lastLifeSign; // when the last time the run()-loop was executed
private final CacheStrategy cacheStrategy;
private final int neededResults;
private final Pattern snippetPattern;
private boolean shallrun;
private final SolrConnector solr;
public Worker(final int id, final long maxlifetime, final CacheStrategy cacheStrategy, final Pattern snippetPattern, final int neededResults) {
this.cacheStrategy = cacheStrategy;
this.lastLifeSign = System.currentTimeMillis();
this.snippetPattern = snippetPattern;
this.timeout = System.currentTimeMillis() + Math.max(1000, maxlifetime);
this.neededResults = neededResults;
this.shallrun = true;
this.solr = SnippetProcess.this.rankingProcess.getQuery().getSegment().getSolr();
}
@Override
public void run() {
// start fetching urls and snippets
URIMetadataRow page;
ResultEntry resultEntry;
//final int fetchAhead = snippetMode == 0 ? 0 : 10;
final boolean nav_topics = SnippetProcess.this.query.navigators.equals("all") || SnippetProcess.this.query.navigators.indexOf("topics",0) >= 0;
try {
//System.out.println("DEPLOYED WORKER " + id + " FOR " + this.neededResults + " RESULTS, timeoutd = " + (this.timeout - System.currentTimeMillis()));
while (this.shallrun && System.currentTimeMillis() < this.timeout) {
//Log.logInfo("SnippetProcess", "***** timeleft = " + (this.timeout - System.currentTimeMillis()));
this.lastLifeSign = System.currentTimeMillis();
if (MemoryControl.shortStatus()) {
break;
}
// check if we have enough; we stop only if we can fetch online; otherwise its better to run this to get better navigation
if ((this.cacheStrategy == null || this.cacheStrategy.isAllowedToFetchOnline()) && SnippetProcess.this.result.sizeAvailable() >= this.neededResults) {
//Log.logWarning("ResultFetcher", SnippetProcess.this.result.sizeAvailable() + " = result.sizeAvailable() >= this.neededResults = " + this.neededResults);
break;
}
// check if we can succeed if we try to take another url
if (SnippetProcess.this.rankingProcess.feedingIsFinished() && SnippetProcess.this.rankingProcess.sizeQueue() == 0) {
//Log.logWarning("ResultFetcher", "rankingProcess.feedingIsFinished() && rankingProcess.sizeQueue() == 0");
break;
}
// get next entry
page = SnippetProcess.this.rankingProcess.takeURL(true, Math.min(500, Math.max(20, this.timeout - System.currentTimeMillis())));
//if (page != null) Log.logInfo("ResultFetcher", "got one page: " + page.metadata().url().toNormalform(true, false));
//if (page == null) page = rankedCache.takeURL(false, this.timeout - System.currentTimeMillis());
if (page == null) {
//Log.logWarning("ResultFetcher", "page == null");
break; // no more available
}
this.setName(page.url().toNormalform(true, false)); // to support debugging
if (SnippetProcess.this.query.filterfailurls && SnippetProcess.this.workTables.failURLsContains(page.hash())) {
continue;
}
// in case that we have an attached solr, we load also the solr document
String solrContent = null;
if (this.solr != null) {
SolrDocument sd = null;
final SolrDocumentList sdl = this.solr.get(SolrField.id.getSolrFieldName()+ ":" + ASCII.String(page.hash()), 0, 1);
if (sdl.size() > 0) {
sd = sdl.get(0);
}
if (sd != null) {
solrContent = Switchboard.getSwitchboard().solrScheme.solrGetText(sd);
}
}
resultEntry = fetchSnippet(page, solrContent, this.cacheStrategy); // does not fetch snippets if snippetMode == 0
if (resultEntry == null)
{
continue; // the entry had some problems, cannot be used
//final String rawLine = resultEntry.textSnippet() == null ? null : resultEntry.textSnippet().getLineRaw();
//System.out.println("***SNIPPET*** raw='" + rawLine + "', pattern='" + this.snippetPattern.toString() + "'");
//if (rawLine != null && !this.snippetPattern.matcher(rawLine).matches()) continue;
}
//if (result.contains(resultEntry)) continue;
SnippetProcess.this.urlRetrievalAllTime += resultEntry.dbRetrievalTime;
SnippetProcess.this.snippetComputationAllTime += resultEntry.snippetComputationTime;
// place the result to the result vector
// apply post-ranking
long ranking = Long.valueOf(SnippetProcess.this.rankingProcess.getOrder().cardinal(resultEntry.word()));
ranking += postRanking(resultEntry, SnippetProcess.this.rankingProcess.getTopicNavigator(10));
resultEntry.ranking = ranking;
SnippetProcess.this.result.put(new ReverseElement<ResultEntry>(resultEntry, ranking)); // remove smallest in case of overflow
if (nav_topics) {
SnippetProcess.this.rankingProcess.addTopics(resultEntry);
}
}
if (System.currentTimeMillis() >= this.timeout) {
Log.logWarning("SnippetProcess", "worker ended with timoeout");
}
//System.out.println("FINISHED WORKER " + id + " FOR " + this.neededResults + " RESULTS, loops = " + loops);
} catch (final Exception e) {
Log.logException(e);
}
//Log.logInfo("SEARCH", "resultWorker thread " + this.id + " terminated");
}
public void pleaseStop() {
this.shallrun = false;
}
/**
* calculate the time since the worker has had the latest activity
* @return time in milliseconds lasted since latest activity
*/
public long busytime() {
return System.currentTimeMillis() - this.lastLifeSign;
}
}
protected ResultEntry fetchSnippet(final URIMetadataRow page, final String solrText, final CacheStrategy cacheStrategy) {
// Snippet Fetching can has 3 modes:
// 0 - do not fetch snippets
// 1 - fetch snippets offline only
// 2 - online snippet fetch
// load only urls if there was not yet a root url of that hash
// find the url entry
long startTime = System.currentTimeMillis();
if (page == null) {
return null;
}
final long dbRetrievalTime = System.currentTimeMillis() - startTime;
if (cacheStrategy == null) {
final TextSnippet snippet = new TextSnippet(
null,
solrText,
page,
this.snippetFetchWordHashes,
null,
((this.query.constraint != null) && (this.query.constraint.get(Condenser.flag_cat_indexof))),
220,
Integer.MAX_VALUE,
!this.query.isLocal());
return new ResultEntry(page, this.query.getSegment(), this.peers, snippet, null, dbRetrievalTime, 0); // result without snippet
}
// load snippet
if (page.url().getContentDomain() == Classification.ContentDomain.TEXT || page.url().getContentDomain() == Classification.ContentDomain.ALL) {
// attach text snippet
startTime = System.currentTimeMillis();
final TextSnippet snippet = new TextSnippet(
this.loader,
solrText,
page,
this.snippetFetchWordHashes,
cacheStrategy,
((this.query.constraint != null) && (this.query.constraint.get(Condenser.flag_cat_indexof))),
180,
Integer.MAX_VALUE,
!this.query.isLocal());
final long snippetComputationTime = System.currentTimeMillis() - startTime;
Log.logInfo("SEARCH", "text snippet load time for " + page.url() + ": " + snippetComputationTime + ", " + (!snippet.getErrorCode().fail() ? "snippet found" : ("no snippet found (" + snippet.getError() + ")")));
if (!snippet.getErrorCode().fail()) {
// we loaded the file and found the snippet
return new ResultEntry(page, this.query.getSegment(), this.peers, snippet, null, dbRetrievalTime, snippetComputationTime); // result with snippet attached
} else if (cacheStrategy.mustBeOffline()) {
// we did not demand online loading, therefore a failure does not mean that the missing snippet causes a rejection of this result
// this may happen during a remote search, because snippet loading is omitted to retrieve results faster
return new ResultEntry(page, this.query.getSegment(), this.peers, null, null, dbRetrievalTime, snippetComputationTime); // result without snippet
} else {
// problems with snippet fetch
if (this.snippetFetchWordHashes.has(Segment.catchallHash)) {
// we accept that because the word cannot be on the page
return new ResultEntry(page, this.query.getSegment(), this.peers, null, null, dbRetrievalTime, 0);
}
final String reason = "no text snippet; errorCode = " + snippet.getErrorCode();
if (this.deleteIfSnippetFail) {
this.workTables.failURLsRegisterMissingWord(this.query.getSegment().termIndex(), page.url(), this.query.queryHashes, reason);
}
Log.logInfo("SEARCH", "sorted out url " + page.url().toNormalform(true, false) + " during search: " + reason);
return null;
}
} else {
return new ResultEntry(page, this.query.getSegment(), this.peers, null, null, dbRetrievalTime, 0); // result without snippet
}
// finished, no more actions possible here
}
/**
* delete a specific entry from the search results
* this is used if the user clicks on a '-' sign beside the search result
* @param urlhash
* @return true if an entry was deleted, false otherwise
*/
public boolean delete(final String urlhash) {
final Iterator<Element<ResultEntry>> i = this.result.iterator();
Element<ResultEntry> entry;
while (i.hasNext()) {
entry = i.next();
if (urlhash.equals(ASCII.String(entry.getElement().url().hash()))) {
i.remove();
return true;
}
}
return false;
}
}