okular/generators/chm/lib/helper_search_index.cpp

/*
 *  Kchmviewer - a CHM and EPUB file viewer with broad language support
 *  Copyright (C) 2004-2014 George Yunaev, gyunaev@ulduzsoft.com
 *
 *  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 3 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, see <http://www.gnu.org/licenses/>.
 */

#include <QApplication>
#include <QTextCodec>

#include "ebook.h"
#include "ebook_search.h"
#include "helper_search_index.h"

static const int DICT_VERSION = 4;

namespace QtAs {

// Those characters are splitters (i.e. split the word), but added themselves into dictionary too.
// This makes the dictionary MUCH larger, but ensure that for the piece of "window->print" both 
// search for "print" and "->print" will find it.
static const char SPLIT_CHARACTERS[] = "!()*&^%#@[]{}':;,.?/|/?<>\\-+=~`";
	
// Those characters are parts of word - for example, '_' is here, and search for _debug will find only _debug.
static const char WORD_CHARACTERS[] = "$_";


struct Term
{
	Term() : frequency(-1) {}
	Term( const QString &t, int f, QVector<Document> l ) : term( t ), frequency( f ), documents( l ) {}
	QString term;
	int frequency;
	QVector<Document>documents;
	bool operator<( const Term &i2 ) const { return frequency < i2.frequency; }
};

	
QDataStream &operator>>( QDataStream &s, Document &l )
{
	s >> l.docNumber;
	s >> l.frequency;
	return s;
}

QDataStream &operator<<( QDataStream &s, const Document &l )
{
	s << (short)l.docNumber;
	s << (short)l.frequency;
	return s;
}

Index::Index()
	: QObject( 0 )
{
	lastWindowClosed = false;
	connect( qApp, SIGNAL( lastWindowClosed() ), this, SLOT( setLastWinClosed() ) );
}

void Index::setLastWinClosed()
{
	lastWindowClosed = true;
}


bool Index::makeIndex(const QList< QUrl >& docs, EBook *chmFile )
{
	if ( docs.isEmpty() )
		return false;
	
	docList = docs;

	if ( chmFile->hasFeature( EBook::FEATURE_ENCODING ) )
		entityDecoder.changeEncoding( QTextCodec::codecForName( chmFile->currentEncoding().toUtf8() ) );
	
	QList< QUrl >::ConstIterator it = docList.constBegin();
	int steps = docList.count() / 100;
	
	if ( !steps )
		steps++;
	
	int prog = 0;
	
	for ( int i = 0; it != docList.constEnd(); ++it, ++i )
	{
		if ( lastWindowClosed )
			return false;

		QUrl filename = *it;
		QStringList terms;
		
		if ( parseDocumentToStringlist( chmFile, filename, terms ) )
		{
			for ( QStringList::ConstIterator tit = terms.constBegin(); tit != terms.constEnd(); ++tit )
				insertInDict( *tit, i );
		}
		
		if ( i%steps == 0 )
		{
			prog++;
			prog = qMin( prog, 99 );
			emit indexingProgress( prog, tr("Processing document %1") .arg( (*it).path() ) );
		}
	}
	
	emit indexingProgress( 100, tr("Processing completed") );
	return true;
}


void Index::insertInDict( const QString &str, int docNum )
{
	Entry *e = 0;
	if ( dict.count() )
		e = dict[ str ];

	if ( e )
	{
		if ( e->documents.last().docNumber != docNum )
			e->documents.append( Document(docNum, 1 ) );
		else
			e->documents.last().frequency++;
	}
	else
	{
		dict.insert( str, new Entry( docNum ) );
	}
}


bool Index::parseDocumentToStringlist(EBook *chmFile, const QUrl& filename, QStringList& tokenlist )
{
	QString parsedbuf, parseentity, text;
	
	if ( !chmFile->getFileContentAsString( text, filename )
	|| text.isEmpty() )
	{
		qWarning( "Search index generator: could not retrieve the document content for %s", qPrintable( filename.toString() ) );
		return false;
	}

	m_charssplit = SPLIT_CHARACTERS;
	m_charsword = WORD_CHARACTERS;
	
	tokenlist.clear();
	
	// State machine states
	enum state_t
	{
		STATE_OUTSIDE_TAGS,		// outside HTML tags; parse text
		STATE_IN_HTML_TAG,		// inside HTML tags; wait for end tag
		STATE_IN_QUOTES,		// inside HTML tags and inside quotes; wait for end quote (in var QuoteChar)
		STATE_IN_HTML_ENTITY	// inside HTML entity; parse the entity
	};
	
	state_t state = STATE_OUTSIDE_TAGS;
	QChar QuoteChar; // used in STATE_IN_QUOTES
	
	for ( int j = 0; j < text.length(); j++ )
	{
		QChar ch = text[j];
		
		if ( (j % 20000) == 0 )
			qApp->processEvents( QEventLoop::ExcludeUserInputEvents );
		
		if ( state == STATE_IN_HTML_TAG )
		{
			// We are inside HTML tag.
			// Ignore everything until we see '>' (end of HTML tag) or quote char (quote start)
			if ( ch == '"' || ch == '\'' )
			{
				state = STATE_IN_QUOTES;
				QuoteChar = ch;
			}
			else if ( ch == '>' )
				state = STATE_OUTSIDE_TAGS;
				
			continue;
		}
		else if ( state == STATE_IN_QUOTES )
		{
			// We are inside quoted text inside HTML tag. 
			// Ignore everything until we see the quote character again
			if ( ch == QuoteChar )
				state = STATE_IN_HTML_TAG;
				
			continue;
		}
		else if ( state == STATE_IN_HTML_ENTITY )
		{
			// We are inside encoded HTML entity (like &nbsp;).
			// Collect to parsedbuf everything until we see ;
			if ( ch.isLetterOrNumber() )
			{
				// get next character of this entity
				parseentity.append( ch );
				continue;
			}
				
			// The entity ended
			state = STATE_OUTSIDE_TAGS;
			
			// Some shitty HTML does not terminate entities correctly. Screw it.			
			if ( ch != ';' && ch != '<' )
			{
				if ( parseentity.isEmpty() )
				{
					// straight '&' symbol. Add and continue.
					parsedbuf += "&";
				}
				else
					qWarning( "Index::parseDocument: incorrectly terminated HTML entity '&%s%c', ignoring", qPrintable( parseentity ), ch.toLatin1() );
				
				j--; // parse this character again, but in different state
				continue;
			}
			
			// Don't we have a space?
			if ( parseentity.toLower() != "nbsp" )
			{
				QString entity = entityDecoder.decode( parseentity );
			
				if ( entity.isNull() )
				{
					// decodeEntity() already printed error message
					//qWarning( "Index::parseDocument: failed to decode entity &%s;", parsedbuf.ascii() );
					continue;
				}
			
				parsedbuf += entity;
				continue;
			}
			else
				ch = ' '; // We got a space, so treat it like it, and not add it to parsebuf
		}
		
		// 
		// Now process STATE_OUTSIDE_TAGS
		//
		
		// Check for start of HTML tag, and switch to STATE_IN_HTML_TAG if it is
		if ( ch == '<' )
		{
			state = STATE_IN_HTML_TAG;
			goto tokenize_buf;
		}
		
		// Check for start of HTML entity
		if ( ch == '&' )
		{
			state = STATE_IN_HTML_ENTITY;
			parseentity = QString::null;
			continue;
		}
		
		// Replace quote by ' - quotes are used in search window to set the phrase
		if ( ch == '"' )
			ch = '\'';
		
		// Ok, we have a valid character outside HTML tags, and probably some in buffer already.
		// If it is char or letter, add it and continue
		if ( ch.isLetterOrNumber() || m_charsword.indexOf( ch ) != -1 )
		{
			parsedbuf.append( ch );
			continue;
		}
		
		// If it is a split char, add the word to the dictionary, and then add the char itself.
		if ( m_charssplit.indexOf( ch ) != -1 )
		{
			if ( !parsedbuf.isEmpty() )
				tokenlist.push_back( parsedbuf.toLower() );
			
			tokenlist.push_back( ch.toLower() );
			parsedbuf = QString::null;
			continue;
		}
		
tokenize_buf:		
		// Just add the word; it is most likely a space or terminated by tokenizer.
		if ( !parsedbuf.isEmpty() )
		{
			tokenlist.push_back( parsedbuf.toLower() );
			parsedbuf = QString::null;
		}
	}
	
	// Add the last word if still here - for broken htmls.
	if ( !parsedbuf.isEmpty() )
		tokenlist.push_back( parsedbuf.toLower() );
	
	return true;
}


void Index::writeDict( QDataStream& stream )
{
	stream << DICT_VERSION;
	stream << m_charssplit;
	stream << m_charsword;
	
	// Document list
	stream << docList;
	
	// Dictionary
	for( QHash<QString, Entry *>::ConstIterator it = dict.constBegin(); it != dict.constEnd(); ++it )
	{
		stream << it.key();
		stream << (int) it.value()->documents.count();
		stream << it.value()->documents;
	}
}


bool Index::readDict( QDataStream& stream )
{
	dict.clear();
	docList.clear();
	
	QString key;
	int version, numOfDocs;
	
	stream >> version;
	
	if ( version < 2 )
		return false;
	
	stream >> m_charssplit;
	stream >> m_charsword;
	
	// Read the document list
	stream >> docList;
	
	while ( !stream.atEnd() )
	{
		stream >> key;
		stream >> numOfDocs;
		
		QVector<Document> docs( numOfDocs );
		
		stream >> docs;
		dict.insert( key, new Entry( docs ) );
	}
	
	return dict.size() > 0;
}


QList< QUrl > Index::query(const QStringList &terms, const QStringList &termSeq, const QStringList &seqWords, EBook *chmFile )
{
	QList<Term> termList;

	QStringList::ConstIterator it = terms.begin();
	for ( it = terms.begin(); it != terms.end(); ++it )
	{
		Entry *e = 0;
		
		if ( dict[ *it ] )
		{
			e = dict[ *it ];
			termList.append( Term( *it, e->documents.count(), e->documents ) );
		}
		else
		{
			return QList< QUrl >();
		}
	}
	
	if ( !termList.count() )
		return QList< QUrl >();
	
	qSort( termList );

	QVector<Document> minDocs = termList.takeFirst().documents;
	for(QList<Term>::Iterator it = termList.begin(); it != termList.end(); ++it) {
		Term *t = &(*it);
		QVector<Document> docs = t->documents;
		for(QVector<Document>::Iterator minDoc_it = minDocs.begin(); minDoc_it != minDocs.end(); ) {
			bool found = false;
			for (QVector<Document>::ConstIterator doc_it = docs.constBegin(); doc_it != docs.constEnd(); ++doc_it ) {
				if ( (*minDoc_it).docNumber == (*doc_it).docNumber ) {
					(*minDoc_it).frequency += (*doc_it).frequency;
					found = true;
					break;
				}
			}
			if ( !found )
				minDoc_it = minDocs.erase( minDoc_it );
			else
				++minDoc_it;
		}
	}

	QList< QUrl > results;
	qSort( minDocs );
	if ( termSeq.isEmpty() ) {
		for(QVector<Document>::Iterator it = minDocs.begin(); it != minDocs.end(); ++it)
			results << docList.at((int)(*it).docNumber);
		return results;
	}

	QUrl fileName;
	for(QVector<Document>::Iterator it = minDocs.begin(); it != minDocs.end(); ++it) {
		fileName =  docList[ (int)(*it).docNumber ];
		if ( searchForPhrases( termSeq, seqWords, fileName, chmFile ) )
			results << fileName;
	}
	
	return results;
}


bool Index::searchForPhrases( const QStringList &phrases, const QStringList &words, const QUrl &filename, EBook * chmFile )
{
	QStringList parsed_document;

	if ( !parseDocumentToStringlist( chmFile, filename, parsed_document ) )
		return false;

	miniDict.clear();
	
	// Initialize the dictionary with the words in phrase(s)
	for ( QStringList::ConstIterator cIt = words.begin(); cIt != words.end(); ++cIt )
		miniDict.insert( *cIt, new PosEntry( 0 ) );

	// Fill the dictionary with the words from the document
	unsigned int word_offset = 3;
	for ( QStringList::ConstIterator it = parsed_document.constBegin(); it != parsed_document.constEnd(); it++, word_offset++ )
	{
		PosEntry * entry = miniDict[ *it ];
		
		if ( entry )
			entry->positions.append( word_offset );
	}
	
	// Dump it
/*	
	QDictIterator<PosEntry> it( miniDict );
	for( ; it.current(); ++it )
	{
		QString text( it.currentKey() );
		QValueList<uint> pos = miniDict[text]->positions;
		for ( unsigned int i = 1; i < pos.size(); i++ )
			text += " " + QString::number( pos[i] );
		
		qDebug( "%s", text.ascii());
	}
*/				
	
	QList<uint> first_word_positions;
	
	for ( QStringList::ConstIterator phrase_it = phrases.constBegin(); phrase_it != phrases.constEnd(); phrase_it++ )
	{
		QStringList phrasewords = phrase_it->split( ' ' );
		first_word_positions = miniDict[ phrasewords[0] ]->positions;
		
		for ( int j = 1; j < phrasewords.count(); ++j )
		{
			QList<uint> next_word_it = miniDict[ phrasewords[j] ]->positions;
			QList<uint>::iterator dict_it = first_word_positions.begin();
			
			while ( dict_it != first_word_positions.end() )
			{
				if ( next_word_it.indexOf( *dict_it + 1 ) != -1 )
				{
					(*dict_it)++;
					++dict_it;
				}
				else
					dict_it = first_word_positions.erase( dict_it );
			}
		}
	}
	
	if ( first_word_positions.count() )
		return true;
	
	return false;
}


};