mirror of
https://github.com/xiaoyifang/goldendict-ng.git
synced 2024-11-27 19:24:08 +00:00
fix: code smells
🎨 apply clang-format changes
This commit is contained in:
parent
9bbb23e5d4
commit
5290d0ceb9
|
@ -34,7 +34,8 @@ enum
|
|||
};
|
||||
|
||||
BtreeIndex::BtreeIndex():
|
||||
idxFile( 0 ), rootNodeLoaded( false )
|
||||
idxFile( nullptr ),
|
||||
rootNodeLoaded( false )
|
||||
{
|
||||
}
|
||||
|
||||
|
@ -175,10 +176,7 @@ void BtreeWordSearchRequest::findMatches()
|
|||
|
||||
bool insideSet = false;
|
||||
bool escaped = false;
|
||||
for( wstring::size_type x = 0; x < foldedWithWildcards.size(); x++ )
|
||||
{
|
||||
wchar ch = foldedWithWildcards[ x ];
|
||||
|
||||
for ( char32_t ch : foldedWithWildcards ) {
|
||||
if( ch == L'\\' && !escaped )
|
||||
{
|
||||
escaped = true;
|
||||
|
@ -216,10 +214,7 @@ void BtreeWordSearchRequest::findMatches()
|
|||
folded.clear();
|
||||
folded.reserve( foldedWithWildcards.size() );
|
||||
escaped = false;
|
||||
for( wstring::size_type x = 0; x < foldedWithWildcards.size(); x++ )
|
||||
{
|
||||
wchar ch = foldedWithWildcards[ x ];
|
||||
|
||||
for ( char32_t ch : foldedWithWildcards ) {
|
||||
if( escaped )
|
||||
{
|
||||
if( bNoLetters || ( ch != L'*' && ch != L'?' && ch != L'[' && ch != L']' ) )
|
||||
|
@ -303,11 +298,10 @@ void BtreeWordSearchRequest::findMatches()
|
|||
|
||||
QMutexLocker _( &dataMutex );
|
||||
|
||||
for( unsigned x = 0; x < chain.size(); ++x )
|
||||
{
|
||||
for ( auto & x : chain ) {
|
||||
if( useWildcards )
|
||||
{
|
||||
wstring word = Utf8::decode( chain[ x ].prefix + chain[ x ].word );
|
||||
wstring word = Utf8::decode( x.prefix + x.word );
|
||||
wstring result = Folding::applyDiacriticsOnly( word );
|
||||
if( result.size() >= (wstring::size_type)minMatchLength )
|
||||
{
|
||||
|
@ -322,9 +316,9 @@ void BtreeWordSearchRequest::findMatches()
|
|||
{
|
||||
// Skip middle matches, if requested. If suffix variation is specified,
|
||||
// make sure the string isn't larger than requested.
|
||||
if ( ( allowMiddleMatches || Folding::apply( Utf8::decode( chain[ x ].prefix ) ).empty() ) &&
|
||||
( maxSuffixVariation < 0 || (int)resultFolded.size() - initialFoldedSize <= maxSuffixVariation ) )
|
||||
addMatch( Utf8::decode( chain[ x ].prefix + chain[ x ].word ) );
|
||||
if ( ( allowMiddleMatches || Folding::apply( Utf8::decode( x.prefix ) ).empty() )
|
||||
&& ( maxSuffixVariation < 0 || (int)resultFolded.size() - initialFoldedSize <= maxSuffixVariation ) )
|
||||
addMatch( Utf8::decode( x.prefix + x.word ) );
|
||||
}
|
||||
}
|
||||
|
||||
|
@ -517,7 +511,7 @@ char const * BtreeIndex::findChainOffsetExactOrPrefix( wstring const & target,
|
|||
nextLeaf = 0;
|
||||
}
|
||||
if( !leafEntries )
|
||||
return 0;
|
||||
return nullptr;
|
||||
|
||||
return leaf + sizeof( uint32_t );
|
||||
}
|
||||
|
@ -645,7 +639,7 @@ char const * BtreeIndex::findChainOffsetExactOrPrefix( wstring const & target,
|
|||
if ( currentNodeOffset != rootOffset )
|
||||
throw exCorruptedChainData();
|
||||
else
|
||||
return 0; // No match
|
||||
return nullptr; // No match
|
||||
}
|
||||
|
||||
// Build an array containing all chain pointers
|
||||
|
@ -745,7 +739,7 @@ char const * BtreeIndex::findChainOffsetExactOrPrefix( wstring const & target,
|
|||
return &extLeaf.front() + sizeof( uint32_t );
|
||||
}
|
||||
else
|
||||
return 0; // This was the last leaf
|
||||
return nullptr; // This was the last leaf
|
||||
}
|
||||
else
|
||||
return chainToCheck[ 1 ];
|
||||
|
@ -1117,7 +1111,7 @@ void BtreeIndex::getAllHeadwords( QSet< QString > & headwords )
|
|||
if ( !idxFile )
|
||||
throw exIndexWasNotOpened();
|
||||
|
||||
findArticleLinks( NULL, NULL, &headwords );
|
||||
findArticleLinks( nullptr, nullptr, &headwords );
|
||||
}
|
||||
|
||||
void BtreeIndex::findAllArticleLinks( QVector< WordArticleLink > & articleLinks )
|
||||
|
@ -1127,7 +1121,7 @@ void BtreeIndex::findAllArticleLinks( QVector< WordArticleLink > & articleLinks
|
|||
|
||||
QSet< uint32_t > offsets;
|
||||
|
||||
findArticleLinks( &articleLinks, &offsets, NULL );
|
||||
findArticleLinks( &articleLinks, &offsets, nullptr );
|
||||
}
|
||||
|
||||
void BtreeIndex::findArticleLinks( QVector< WordArticleLink > * articleLinks,
|
||||
|
@ -1150,7 +1144,7 @@ void BtreeIndex::findArticleLinks( QVector< WordArticleLink > * articleLinks,
|
|||
|
||||
char const * leaf = &rootNode.front();
|
||||
char const * leafEnd = leaf + rootNode.size();
|
||||
char const * chainPtr = 0;
|
||||
char const * chainPtr = nullptr;
|
||||
|
||||
vector< char > extLeaf;
|
||||
|
||||
|
@ -1213,22 +1207,21 @@ void BtreeIndex::findArticleLinks( QVector< WordArticleLink > * articleLinks,
|
|||
articleLinks->reserve( n + n / 10 );
|
||||
}
|
||||
|
||||
for( unsigned i = 0; i < result.size(); i++ )
|
||||
{
|
||||
for ( auto & i : result ) {
|
||||
if( isCancelled && Utils::AtomicInt::loadAcquire( *isCancelled ) )
|
||||
return;
|
||||
|
||||
if( headwords )
|
||||
headwords->insert( QString::fromUtf8( ( result[ i ].prefix + result[ i ].word ).c_str() ) );
|
||||
headwords->insert( QString::fromUtf8( ( i.prefix + i.word ).c_str() ) );
|
||||
|
||||
if( offsets && offsets->contains( result[ i ].articleOffset ) )
|
||||
if ( offsets && offsets->contains( i.articleOffset ) )
|
||||
continue;
|
||||
|
||||
if( offsets )
|
||||
offsets->insert( result[ i ].articleOffset );
|
||||
offsets->insert( i.articleOffset );
|
||||
|
||||
if( articleLinks )
|
||||
articleLinks->push_back( WordArticleLink( result[ i ].prefix + result[ i ].word, result[ i ].articleOffset ) );
|
||||
articleLinks->push_back( WordArticleLink( i.prefix + i.word, i.articleOffset ) );
|
||||
}
|
||||
|
||||
if ( chainPtr >= leafEnd )
|
||||
|
@ -1279,9 +1272,9 @@ void BtreeIndex::findSingleNodeHeadwords( uint32_t offsets,
|
|||
|
||||
QMutexLocker _( idxFileMutex );
|
||||
|
||||
char const * leaf = 0;
|
||||
char const * leafEnd = 0;
|
||||
char const * chainPtr = 0;
|
||||
char const * leaf = nullptr;
|
||||
char const * leafEnd = nullptr;
|
||||
char const * chainPtr = nullptr;
|
||||
|
||||
vector< char > extLeaf;
|
||||
|
||||
|
@ -1299,9 +1292,8 @@ void BtreeIndex::findSingleNodeHeadwords( uint32_t offsets,
|
|||
|
||||
if( headwords )
|
||||
{
|
||||
for( unsigned i = 0; i < result.size(); i++ )
|
||||
{
|
||||
headwords->insert( QString::fromUtf8( ( result[ i ].prefix + result[ i ].word ).c_str() ) );
|
||||
for ( auto & i : result ) {
|
||||
headwords->insert( QString::fromUtf8( ( i.prefix + i.word ).c_str() ) );
|
||||
}
|
||||
}
|
||||
|
||||
|
@ -1368,7 +1360,7 @@ void BtreeIndex::getHeadwordsFromOffsets( QList<uint32_t> & offsets,
|
|||
|
||||
char const * leaf = &rootNode.front();
|
||||
char const * leafEnd = leaf + rootNode.size();
|
||||
char const * chainPtr = 0;
|
||||
char const * chainPtr = nullptr;
|
||||
|
||||
vector< char > extLeaf;
|
||||
|
||||
|
@ -1416,9 +1408,8 @@ void BtreeIndex::getHeadwordsFromOffsets( QList<uint32_t> & offsets,
|
|||
{
|
||||
vector< WordArticleLink > result = readChain( chainPtr );
|
||||
|
||||
for( unsigned i = 0; i < result.size(); i++ )
|
||||
{
|
||||
uint32_t articleOffset = result.at(i).articleOffset;
|
||||
for ( auto & i : result ) {
|
||||
uint32_t articleOffset = i.articleOffset;
|
||||
|
||||
QList<uint32_t>::Iterator it = std::lower_bound( begOffsets, endOffsets,
|
||||
articleOffset );
|
||||
|
@ -1428,7 +1419,7 @@ void BtreeIndex::getHeadwordsFromOffsets( QList<uint32_t> & offsets,
|
|||
if( isCancelled && Utils::AtomicInt::loadAcquire( *isCancelled ) )
|
||||
return;
|
||||
|
||||
auto word = QString::fromUtf8( ( result[ i ].prefix + result[ i ].word ).c_str() );
|
||||
auto word = QString::fromUtf8( ( i.prefix + i.word ).c_str() );
|
||||
|
||||
if ( headwords.indexOf( word ) == -1 ) {
|
||||
headwords.append( word );
|
||||
|
|
|
@ -9,7 +9,6 @@
|
|||
|
||||
namespace Folding {
|
||||
|
||||
|
||||
/// Tests if the given char is one of the Unicode combining marks. Some are
|
||||
/// caught by the diacritics folding table, but they are only handled there
|
||||
/// when they come with their main characters, not by themselves. The rest
|
||||
|
@ -154,9 +153,10 @@ wstring applyWhitespaceAndPunctOnly( wstring const & in )
|
|||
|
||||
out.reserve( in.size() );
|
||||
|
||||
for( size_t left = in.size(); left--; ++nextChar )
|
||||
for ( size_t left = in.size(); left--; ++nextChar ) {
|
||||
if ( !isWhitespace( *nextChar ) && !isPunct( *nextChar ) )
|
||||
out.push_back( *nextChar );
|
||||
}
|
||||
|
||||
return out;
|
||||
}
|
||||
|
@ -246,7 +246,7 @@ QString trimWhitespace( QString const & in )
|
|||
QString escapeWildcardSymbols( const QString & str )
|
||||
{
|
||||
QString escaped( str );
|
||||
escaped.replace( QRegularExpression( R"(([\[\]\?\*]))" ), "\\\\1" );
|
||||
escaped.replace( QRegularExpression( R"(([\[\]\?\*]))" ), R"(\\1)" );
|
||||
|
||||
return escaped;
|
||||
}
|
||||
|
@ -254,8 +254,8 @@ QString escapeWildcardSymbols( const QString & str )
|
|||
QString unescapeWildcardSymbols( const QString & str )
|
||||
{
|
||||
QString unescaped( str );
|
||||
unescaped.replace( QRegularExpression( R"(\\([\[\]\?\*]))" ), "\\1" );
|
||||
unescaped.replace( QRegularExpression( R"(\\([\[\]\?\*]))" ), R"(\1)" );
|
||||
|
||||
return unescaped;
|
||||
}
|
||||
}
|
||||
} // namespace Folding
|
||||
|
|
Loading…
Reference in a new issue