goldendict-ng/wordfinder.cc

473 lines
14 KiB
C++
Raw Normal View History

2012-02-20 21:47:14 +00:00
/* This file is (c) 2008-2012 Konstantin Isakov <ikm@goldendict.org>
2009-01-29 19:16:25 +00:00
* Part of GoldenDict. Licensed under GPLv3 or later, see the LICENSE file */
#include "wordfinder.hh"
#include "folding.hh"
#include "wstring_qt.hh"
#include <QThreadPool>
2009-01-29 19:16:25 +00:00
#include <map>
#include "gddebug.hh"
2009-01-29 19:16:25 +00:00
using std::vector;
using std::list;
using gd::wstring;
using gd::wchar;
2009-01-29 19:16:25 +00:00
using std::map;
using std::pair;
WordFinder::WordFinder( QObject * parent ):
QObject( parent ), searchInProgress( false ),
updateResultsTimer( this ),
searchQueued( false )
2009-01-29 19:16:25 +00:00
{
updateResultsTimer.setInterval( 1000 ); // We use a one second update timer
updateResultsTimer.setSingleShot( true );
connect( &updateResultsTimer, SIGNAL( timeout() ),
2011-12-12 16:52:07 +00:00
this, SLOT( updateResults() ), Qt::QueuedConnection );
2009-01-29 19:16:25 +00:00
}
WordFinder::~WordFinder()
2009-01-29 19:16:25 +00:00
{
clear();
2009-01-29 19:16:25 +00:00
}
void WordFinder::prefixMatch( QString const & str,
std::vector< sptr< Dictionary::Class > > const & dicts,
unsigned long maxResults,
Dictionary::Features features )
2009-01-29 19:16:25 +00:00
{
cancel();
2009-01-29 19:16:25 +00:00
searchQueued = true;
searchType = PrefixMatch;
inputWord = str;
inputDicts = &dicts;
requestedMaxResults = maxResults;
requestedFeatures = features;
2009-01-29 19:16:25 +00:00
resultsArray.clear();
resultsIndex.clear();
searchResults.clear();
2009-01-29 19:16:25 +00:00
if ( queuedRequests.empty() )
{
// No requests are queued, no need to wait for them to finish.
startSearch();
}
2009-01-29 19:16:25 +00:00
// Else some requests are still queued, last one to finish would trigger
// new search. This shouldn't take a lot of time, since they were all
// cancelled, but still it could take some time.
2009-01-29 19:16:25 +00:00
}
void WordFinder::stemmedMatch( QString const & str,
std::vector< sptr< Dictionary::Class > > const & dicts,
unsigned minLength,
unsigned maxSuffixVariation,
unsigned long maxResults,
Dictionary::Features features )
{
cancel();
searchQueued = true;
searchType = StemmedMatch;
inputWord = str;
inputDicts = &dicts;
requestedMaxResults = maxResults;
requestedFeatures = features;
stemmedMinLength = minLength;
stemmedMaxSuffixVariation = maxSuffixVariation;
resultsArray.clear();
resultsIndex.clear();
searchResults.clear();
if ( queuedRequests.empty() )
startSearch();
}
2009-01-29 19:16:25 +00:00
void WordFinder::startSearch()
2009-01-29 19:16:25 +00:00
{
if ( !searchQueued )
return; // Search was probably cancelled
// Clear the requests just in case
queuedRequests.clear();
finishedRequests.clear();
searchErrorString.clear();
searchResultsUncertain = false;
searchQueued = false;
searchInProgress = true;
// Gather all writings of the word
if ( allWordWritings.size() != 1 )
allWordWritings.resize( 1 );
allWordWritings[ 0 ] = gd::toWString( inputWord );
for( size_t x = 0; x < inputDicts->size(); ++x )
{
vector< wstring > writings = (*inputDicts)[ x ]->getAlternateWritings( allWordWritings[ 0 ] );
allWordWritings.insert( allWordWritings.end(), writings.begin(), writings.end() );
}
// Query each dictionary for all word writings
for( size_t x = 0; x < inputDicts->size(); ++x )
{
if ( ( (*inputDicts)[ x ]->getFeatures() & requestedFeatures ) != requestedFeatures )
continue;
for( size_t y = 0; y < allWordWritings.size(); ++y )
{
2013-09-19 19:43:16 +00:00
try
{
sptr< Dictionary::WordSearchRequest > sr =
( searchType == PrefixMatch ) ?
(*inputDicts)[ x ]->prefixMatch( allWordWritings[ y ], requestedMaxResults ) :
(*inputDicts)[ x ]->stemmedMatch( allWordWritings[ y ], stemmedMinLength, stemmedMaxSuffixVariation, requestedMaxResults );
connect( sr.get(), SIGNAL( finished() ),
this, SLOT( requestFinished() ), Qt::QueuedConnection );
queuedRequests.push_back( sr );
}
catch( std::exception & e )
{
gdWarning( "Word \"%s\" search error (%s) in \"%s\"\n",
inputWord.toUtf8().data(), e.what(), (*inputDicts)[ x ]->getName().c_str() );
2013-09-19 19:43:16 +00:00
}
}
}
2009-01-29 19:16:25 +00:00
// Handle any requests finished already
2009-01-29 19:16:25 +00:00
requestFinished();
}
2009-01-29 19:16:25 +00:00
void WordFinder::cancel()
{
searchQueued = false;
searchInProgress = false;
cancelSearches();
2009-01-29 19:16:25 +00:00
}
void WordFinder::clear()
2009-01-29 19:16:25 +00:00
{
cancel();
queuedRequests.clear();
finishedRequests.clear();
}
2009-01-29 19:16:25 +00:00
void WordFinder::requestFinished()
{
bool newResults = false;
2009-01-29 19:16:25 +00:00
// See how many new requests have finished, and if we have any new results
for( list< sptr< Dictionary::WordSearchRequest > >::iterator i =
queuedRequests.begin(); i != queuedRequests.end(); )
{
if ( (*i)->isFinished() )
2009-01-29 19:16:25 +00:00
{
if ( searchInProgress && !(*i)->getErrorString().isEmpty() )
searchErrorString = tr( "Failed to query some dictionaries." );
2009-01-29 19:16:25 +00:00
if ( (*i)->isUncertain() )
searchResultsUncertain = true;
if ( (*i)->matchesCount() )
{
newResults = true;
2009-01-29 19:16:25 +00:00
// This list is handled by updateResults()
finishedRequests.splice( finishedRequests.end(), queuedRequests, i++ );
}
else // We won't do anything with it anymore, so we erase it
queuedRequests.erase( i++ );
}
else
++i;
}
2009-01-29 19:16:25 +00:00
if ( !searchInProgress )
{
// There is no search in progress, so we just wait until there's
// no requests left
if ( queuedRequests.empty() )
{
// We got rid of all queries, queued search can now start
finishedRequests.clear();
if ( searchQueued )
startSearch();
}
2009-01-29 19:16:25 +00:00
return;
}
2009-01-29 19:16:25 +00:00
if ( newResults && queuedRequests.size() && !updateResultsTimer.isActive() )
{
// If we have got some new results, but not all of them, we would start a
// timer to update a user some time in the future
updateResultsTimer.start();
}
2009-01-29 19:16:25 +00:00
if ( queuedRequests.empty() )
{
// Search is finished.
updateResults();
}
}
namespace {
unsigned saturated( unsigned x )
{
return x < 255 ? x : 255;
}
/// Checks whether the first string has the second one inside, surrounded from
/// both sides by either whitespace, punctuation or begin/end of string.
/// If true is returned, pos holds the offset in the haystack. If the offset
/// is larger than 255, it is set to 255.
bool hasSurroundedWithWs( wstring const & haystack, wstring const & needle,
wstring::size_type & pos )
{
if ( haystack.size() < needle.size() )
return false; // Needle won't even fit into a haystack
for( pos = 0; ; ++pos )
{
pos = haystack.find( needle, pos );
if ( pos == wstring::npos )
return false; // Not found
if ( ( !pos || Folding::isWhitespace( haystack[ pos - 1 ] ) ||
Folding::isPunct( haystack[ pos - 1 ] ) ) &&
( ( pos + needle.size() == haystack.size() ) ||
Folding::isWhitespace( haystack[ pos + needle.size() ] ) ||
Folding::isPunct( haystack[ pos + needle.size() ] ) ) )
{
pos = saturated( pos );
return true;
}
}
}
}
void WordFinder::updateResults()
{
if ( !searchInProgress )
return; // Old queued signal
2009-01-29 19:16:25 +00:00
if ( updateResultsTimer.isActive() )
updateResultsTimer.stop(); // Can happen when we were done before it'd expire
2009-01-29 19:16:25 +00:00
for( list< sptr< Dictionary::WordSearchRequest > >::iterator i =
finishedRequests.begin(); i != finishedRequests.end(); )
{
for( size_t count = (*i)->matchesCount(), x = 0; x < count; ++x )
2009-01-29 19:16:25 +00:00
{
wstring match = (**i)[ x ].word;
int weight = (**i)[ x ].weight;
wstring lowerCased = Folding::applySimpleCaseOnly( match );
2009-01-29 19:16:25 +00:00
pair< ResultsIndex::iterator, bool > insertResult =
resultsIndex.insert( pair< wstring, ResultsArray::iterator >( lowerCased,
resultsArray.end() ) );
if ( !insertResult.second )
2009-01-29 19:16:25 +00:00
{
// Wasn't inserted since there was already an item -- check the case
if ( insertResult.first->second->word != match )
2009-01-29 19:16:25 +00:00
{
// The case is different -- agree on a lowercase version
insertResult.first->second->word = lowerCased;
2009-01-29 19:16:25 +00:00
}
if ( !weight && insertResult.first->second->wasSuggested )
insertResult.first->second->wasSuggested = false;
2009-01-29 19:16:25 +00:00
}
else
{
resultsArray.push_back( OneResult() );
resultsArray.back().word = match;
resultsArray.back().rank = INT_MAX;
resultsArray.back().wasSuggested = ( weight != 0 );
insertResult.first->second = --resultsArray.end();
}
2009-01-29 19:16:25 +00:00
}
finishedRequests.erase( i++ );
}
2009-01-29 19:16:25 +00:00
size_t maxSearchResults = 500;
if ( resultsArray.size() )
{
if ( searchType == PrefixMatch )
{
/// Assign each result a category, storing it in the rank's field
enum Category
{
ExactMatch,
ExactNoFullCaseMatch,
ExactNoDiaMatch,
ExactNoPunctMatch,
ExactNoWsMatch,
ExactInsideMatch,
ExactNoDiaInsideMatch,
ExactNoPunctInsideMatch,
PrefixMatch,
PrefixNoDiaMatch,
PrefixNoPunctMatch,
PrefixNoWsMatch,
WorstMatch,
Multiplier = 256 // Categories should be multiplied by Multiplier
};
for( unsigned wr = 0; wr < allWordWritings.size(); ++wr )
{
wstring target = Folding::applySimpleCaseOnly( allWordWritings[ wr ] );
wstring targetNoFullCase = Folding::applyFullCaseOnly( target );
wstring targetNoDia = Folding::applyDiacriticsOnly( targetNoFullCase );
wstring targetNoPunct = Folding::applyPunctOnly( targetNoDia );
wstring targetNoWs = Folding::applyWhitespaceOnly( targetNoPunct );
wstring::size_type matchPos = 0;
for( ResultsIndex::const_iterator i = resultsIndex.begin(), j = resultsIndex.end();
i != j; ++i )
{
wstring resultNoFullCase, resultNoDia, resultNoPunct, resultNoWs;
int rank;
if ( i->first == target )
rank = ExactMatch * Multiplier;
else
if ( ( resultNoFullCase = Folding::applyFullCaseOnly( i->first ) ) == targetNoFullCase )
rank = ExactNoFullCaseMatch * Multiplier;
else
if ( ( resultNoDia = Folding::applyDiacriticsOnly( resultNoFullCase ) ) == targetNoDia )
rank = ExactNoDiaMatch * Multiplier;
else
if ( ( resultNoPunct = Folding::applyPunctOnly( resultNoDia ) ) == targetNoPunct )
rank = ExactNoPunctMatch * Multiplier;
else
if ( ( resultNoWs = Folding::applyWhitespaceOnly( resultNoPunct ) ) == targetNoWs )
rank = ExactNoWsMatch * Multiplier;
else
if ( hasSurroundedWithWs( i->first, target, matchPos ) )
rank = ExactInsideMatch * Multiplier + matchPos;
else
if ( hasSurroundedWithWs( resultNoDia, targetNoDia, matchPos ) )
rank = ExactNoDiaInsideMatch * Multiplier + matchPos;
else
if ( hasSurroundedWithWs( resultNoPunct, targetNoPunct, matchPos ) )
rank = ExactNoPunctInsideMatch * Multiplier + matchPos;
else
if ( i->first.size() > target.size() && i->first.compare( 0, target.size(), target ) == 0 )
rank = PrefixMatch * Multiplier + saturated( i->first.size() );
else
if ( resultNoDia.size() > targetNoDia.size() && resultNoDia.compare( 0, targetNoDia.size(), targetNoDia ) == 0 )
rank = PrefixNoDiaMatch * Multiplier + saturated( i->first.size() );
else
if ( resultNoPunct.size() > targetNoPunct.size() && resultNoPunct.compare( 0, targetNoPunct.size(), targetNoPunct ) == 0 )
rank = PrefixNoPunctMatch * Multiplier + saturated( i->first.size() );
else
if ( resultNoWs.size() > targetNoWs.size() && resultNoWs.compare( 0, targetNoWs.size(), targetNoWs ) == 0 )
rank = PrefixNoWsMatch * Multiplier + saturated( i->first.size() );
else
rank = WorstMatch * Multiplier;
if ( i->second->rank > rank )
i->second->rank = rank; // We store the best rank of any writing
}
}
resultsArray.sort( SortByRank() );
}
else
{
// Handling stemmed matches
// We use two factors -- first is the number of characters strings share
// in their beginnings, and second, the length of the strings. Here we assign
// only the first one, storing it in rank. Then we sort the results using
// SortByRankAndLength.
for( unsigned wr = 0; wr < allWordWritings.size(); ++wr )
{
wstring target = Folding::apply( allWordWritings[ wr ] );
for( ResultsIndex::const_iterator i = resultsIndex.begin(), j = resultsIndex.end();
i != j; ++i )
{
wstring resultFolded = Folding::apply( i->first );
int charsInCommon = 0;
for( wchar const * t = target.c_str(), * r = resultFolded.c_str();
*t && *t == *r; ++t, ++r, ++charsInCommon ) ;
int rank = -charsInCommon; // Negated so the lesser-than
// comparison would yield right
// results.
if ( i->second->rank > rank )
i->second->rank = rank; // We store the best rank of any writing
}
}
resultsArray.sort( SortByRankAndLength() );
maxSearchResults = 15;
}
}
2009-01-29 19:16:25 +00:00
searchResults.clear();
searchResults.reserve( resultsArray.size() < maxSearchResults ? resultsArray.size() : maxSearchResults );
2009-01-29 19:16:25 +00:00
for( ResultsArray::const_iterator i = resultsArray.begin(), j = resultsArray.end();
i != j; ++i )
{
//DPRINTF( "%d: %ls\n", i->second, i->first.c_str() );
if ( searchResults.size() < maxSearchResults )
searchResults.push_back( std::pair< QString, bool >( gd::toQString( i->word ), i->wasSuggested ) );
else
break;
}
2009-01-29 19:16:25 +00:00
if ( queuedRequests.size() )
{
// There are still some unhandled results.
emit updated();
2009-01-29 19:16:25 +00:00
}
else
{
// That were all of them.
searchInProgress = false;
emit finished();
}
}
2009-01-29 19:16:25 +00:00
void WordFinder::cancelSearches()
{
for( list< sptr< Dictionary::WordSearchRequest > >::iterator i =
queuedRequests.begin(); i != queuedRequests.end(); ++i )
(*i)->cancel();
2009-01-29 19:16:25 +00:00
}