bool CISpellDict::Load ( const char * szFilename )
{
if ( !szFilename )
return false;
m_dEntries.Reset ();
m_dEntries.Reserve ( 131072 );
FILE * pFile = fopen ( szFilename, "rt" );
if ( !pFile )
return false;
char szWordBuffer [MAX_STR_LENGTH];
while ( !feof ( pFile ) )
{
char * szResult = fgets ( szWordBuffer, MAX_STR_LENGTH, pFile );
if ( !szResult )
break;
int iPos = strlen ( szWordBuffer ) - 1;
while ( iPos>=0 && isspace ( (unsigned char)szWordBuffer[iPos] ) )
szWordBuffer [iPos--] = '\0';
CISpellDictWord Word;
char * szPosition = strchr ( szWordBuffer, '/' );
if ( !szPosition )
{
szPosition = szWordBuffer;
while ( *szPosition && !isspace ( (unsigned char)*szPosition ) )
++szPosition;
*szPosition = '\0';
Word.m_sWord = szWordBuffer;
} else
{
*szPosition = '\0';
Word.m_sWord = szWordBuffer;
++szPosition;
char * szFlags = szPosition;
while ( *szPosition && !isspace ( (unsigned char)*szPosition ) )
++szPosition;
*szPosition = '\0';
Word.m_sFlags = szFlags;
}
m_dEntries.Add ( Word );
}
fclose ( pFile );
return true;
}
void CSphStopwordBuilderDict::Save ( const char * sOutput, int iTop, bool bFreqs )
{
FILE * fp = fopen ( sOutput, "w+" );
if ( !fp )
return;
CSphVector<Word_t> dTop;
dTop.Reserve ( 1024 );
const CSphMTFHashEntry<int> * it;
HASH_FOREACH ( it, m_hWords )
{
Word_t t;
t.m_sWord = it->m_sKey.cstr();
t.m_iCount = it->m_tValue;
dTop.Add ( t );
}
void ExcerptGen_c::TokenizeQuery ( const ExcerptQuery_t & tQuery, CSphDict * pDict, ISphTokenizer * pTokenizer )
{
const bool bUtf8 = pTokenizer->IsUtf8();
// tokenize query words
int iWordsLength = strlen ( tQuery.m_sWords.cstr() );
m_dKeywords.Reserve ( MAX_HIGHLIGHT_WORDS );
BYTE * sWord;
int iKwIndex = 0;
pTokenizer->SetBuffer ( (BYTE *)tQuery.m_sWords.cstr(), iWordsLength );
while ( ( sWord = pTokenizer->GetToken() )!=NULL )
{
SphWordID_t iWord = pDict->GetWordID ( sWord );
if ( iWord )
{
Token_t & tLast = m_dWords.Add();
tLast.m_eType = TOK_WORD;
tLast.m_iWordID = iWord;
tLast.m_iLengthBytes = strlen ( (const char *)sWord );
tLast.m_iLengthCP = bUtf8 ? sphUTF8Len ( (const char *)sWord ) : tLast.m_iLengthBytes;
// store keyword
Keyword_t & kwLast = m_dKeywords.Add();
kwLast.m_iLength = tLast.m_iLengthCP;
// find stars
bool bStarBack = ( *pTokenizer->GetTokenEnd()=='*' );
bool bStarFront = ( pTokenizer->GetTokenStart()!=pTokenizer->GetBufferPtr() ) &&
( pTokenizer->GetTokenStart()[-1]=='*' );
kwLast.m_uStar = ( bStarFront ? STAR_FRONT : 0 ) | ( bStarBack ? STAR_BACK : 0 );
// store token
const int iEndIndex = iKwIndex + tLast.m_iLengthBytes + 1;
m_dKeywordsBuffer.Resize ( iEndIndex );
kwLast.m_iWord = iKwIndex;
strcpy ( &m_dKeywordsBuffer [ iKwIndex ], (const char *)sWord ); // NOLINT
iKwIndex = iEndIndex;
if ( m_dWords.GetLength()==MAX_HIGHLIGHT_WORDS )
break;
}
}
}
char * ExcerptGen_c::BuildExcerpt ( const ExcerptQuery_t & q, CSphDict * pDict, ISphTokenizer * pTokenizer )
{
m_dTokens.Reserve ( 1024 );
m_sBuffer = q.m_sSource;
const bool bUtf8 = pTokenizer->IsUtf8();
m_bUtf8 = bUtf8;
// tokenize query words
int iWordsLength = strlen ( q.m_sWords.cstr() );
CSphVector<char> dKwBuffer ( iWordsLength );
CSphVector<Keyword_t> dKeywords;
dKeywords.Reserve ( MAX_HIGHLIGHT_WORDS );
BYTE * sWord;
int iKwIndex = 0;
pTokenizer->SetBuffer ( (BYTE*)q.m_sWords.cstr(), iWordsLength );
while ( ( sWord = pTokenizer->GetToken() ) != NULL )
{
SphWordID_t iWord = pDict->GetWordID ( sWord );
if ( iWord )
{
m_dWords.Resize ( m_dWords.GetLength () + 1 );
Token_t & tLast = m_dWords.Last ();
tLast.m_eType = TOK_WORD;
tLast.m_iWordID = iWord;
tLast.m_iLengthBytes = strlen ( (const char *)sWord );
tLast.m_iLengthCP = bUtf8 ? sphUTF8Len ( (const char *)sWord ) : tLast.m_iLengthBytes;
// store keyword
dKeywords.Resize( dKeywords.GetLength() + 1 );
Keyword_t & kwLast = dKeywords.Last ();
// find stars
bool bStarBack = *pTokenizer->GetTokenEnd() == '*';
bool bStarFront = ( pTokenizer->GetTokenStart() != pTokenizer->GetBufferPtr() ) &&
pTokenizer->GetTokenStart()[-1] == '*';
kwLast.m_uStar = ( bStarFront ? STAR_FRONT : 0 ) | ( bStarBack ? STAR_BACK : 0 );
// store token
const int iEndIndex = iKwIndex + tLast.m_iLengthBytes + 1;
dKwBuffer.Resize ( iEndIndex );
kwLast.m_iWord = iKwIndex;
strcpy ( &dKwBuffer [ iKwIndex ], (const char *)sWord );
iKwIndex = iEndIndex;
if ( m_dWords.GetLength() == MAX_HIGHLIGHT_WORDS )
break;
}
}
// tokenize document
pTokenizer->SetBuffer ( (BYTE*)q.m_sSource.cstr (), strlen ( q.m_sSource.cstr () ) );
const char * pStartPtr = pTokenizer->GetBufferPtr ();
const char * pLastTokenEnd = pStartPtr;
//assign utf-8
m_sBufferUTF8 = pStartPtr;
while ( ( sWord = pTokenizer->GetToken() ) != NULL )
{
const char * pTokenStart = pTokenizer->GetTokenStart ();
if ( pTokenStart != pStartPtr )
AddJunk ( pLastTokenEnd - pStartPtr,
pTokenStart - pLastTokenEnd,
pTokenizer->GetBoundary() ? pTokenizer->GetBoundaryOffset() : -1 );
SphWordID_t iWord = pDict->GetWordID ( sWord );
pLastTokenEnd = pTokenizer->GetTokenEnd ();
m_dTokens.Resize ( m_dTokens.GetLength () + 1 );
Token_t & tLast = m_dTokens.Last ();
tLast.m_eType = iWord ? TOK_WORD : TOK_SPACE;
tLast.m_iStart = pTokenStart - pStartPtr;
tLast.m_iLengthBytes = pLastTokenEnd - pTokenStart;
tLast.m_iWordID = iWord;
tLast.m_uWords = 0;
// fill word mask
if ( iWord )
{
bool bMatch = false;
int iOffset;
ARRAY_FOREACH ( nWord, m_dWords )
{
const char * keyword = &dKwBuffer [ dKeywords[nWord].m_iWord ];
const Token_t & token = m_dWords[nWord];
switch ( dKeywords[nWord].m_uStar )
{
case STAR_NONE:
bMatch = iWord == token.m_iWordID;
break;
case STAR_FRONT:
iOffset = tLast.m_iLengthBytes - token.m_iLengthBytes;
bMatch = (iOffset >= 0) &&
( memcmp( keyword, sWord + iOffset, token.m_iLengthBytes ) == 0 );
break;
case STAR_BACK:
bMatch = ( tLast.m_iLengthBytes >= token.m_iLengthBytes ) &&
( memcmp( keyword, sWord, token.m_iLengthBytes ) == 0 );
break;
case STAR_BOTH:
bMatch = strstr( (const char *)sWord, keyword ) != NULL;
break;
}
if ( bMatch )
tLast.m_uWords |= (1UL << nWord);
}
}
}
// last space if any
if ( pLastTokenEnd != pTokenizer->GetBufferEnd () )
{
int iOffset = pTokenizer->GetBoundary() ? pTokenizer->GetBoundaryOffset() : -1;
AddJunk ( pLastTokenEnd - pStartPtr, pTokenizer->GetBufferEnd () - pLastTokenEnd, iOffset );
}
m_dTokens.Resize ( m_dTokens.GetLength () + 1 );
Token_t & tLast = m_dTokens.Last ();
tLast.m_eType = TOK_NONE;
tLast.m_iStart = 0;
tLast.m_iLengthBytes = 0;
tLast.m_iWordID = 0;
tLast.m_uWords = 0;
// sum token lengths
int iSourceCodes = 0;
ARRAY_FOREACH ( i, m_dTokens )
{
m_dTokens [i].m_iWeight = 0;
if ( m_dTokens [i].m_iLengthBytes )
{
if ( bUtf8 )
{
//int iLen = sphUTF8Len ( m_sBuffer.SubString ( m_dTokens[i].m_iStart, m_dTokens[i].m_iLengthBytes ).cstr() );
int iLen = sphUTF8Len ( m_sBufferUTF8.SubString ( m_dTokens[i].m_iStart, m_dTokens[i].m_iLengthBytes ).cstr() );
m_dTokens[i].m_iLengthCP = iLen;
}
else
m_dTokens[i].m_iLengthCP = m_dTokens[i].m_iLengthBytes;
iSourceCodes += m_dTokens[i].m_iLengthCP;
}
else
m_dTokens [i].m_iLengthCP = 0;
}
void ExcerptGen_c::TokenizeDocument ( char * pData, CSphDict * pDict, ISphTokenizer * pTokenizer, bool bFillMasks, bool bRetainHtml )
{
m_iDocumentWords = 0;
m_dTokens.Reserve ( 1024 );
m_sBuffer = pData;
pTokenizer->SetBuffer ( (BYTE*)pData, strlen(pData) );
const char * pStartPtr = pTokenizer->GetBufferPtr ();
const char * pLastTokenEnd = pStartPtr;
if ( bRetainHtml )
pTokenizer->AddSpecials ( "<" );
BYTE * sWord;
DWORD uPosition = 0; // hit position in document
while ( ( sWord = pTokenizer->GetToken() )!=NULL )
{
if ( pTokenizer->TokenIsBlended() )
continue;
const char * pTokenStart = pTokenizer->GetTokenStart ();
if ( pTokenStart!=pStartPtr && pTokenStart>pLastTokenEnd )
AddJunk ( pLastTokenEnd - pStartPtr,
pTokenStart - pLastTokenEnd,
pTokenizer->GetBoundary() ? pTokenizer->GetBoundaryOffset() : -1 );
if ( bRetainHtml && *pTokenStart=='<' )
{
int iTagEnd = FindTagEnd ( pTokenStart );
if ( iTagEnd!=-1 )
{
assert ( pTokenStart+iTagEnd<pTokenizer->GetBufferEnd() );
AddJunk ( pTokenStart-pStartPtr, iTagEnd+1, pTokenizer->GetBoundary() ? pTokenizer->GetBoundaryOffset() : -1 );
pTokenizer->SetBufferPtr ( pTokenStart+iTagEnd+1 );
pLastTokenEnd = pTokenStart+iTagEnd+1; // fix it up to prevent adding last chunk on exit
continue;
}
}
SphWordID_t iWord = iWord = pDict->GetWordID ( sWord );
pLastTokenEnd = pTokenizer->GetTokenEnd ();
if ( pTokenizer->GetBoundary() )
uPosition += 100; // FIXME: this should be taken from index settings
Token_t & tLast = m_dTokens.Add();
tLast.m_eType = iWord ? TOK_WORD : TOK_SPACE;
tLast.m_uPosition = iWord ? ++uPosition : 0;
tLast.m_iStart = pTokenStart - pStartPtr;
tLast.m_iLengthBytes = pLastTokenEnd - pTokenStart;
tLast.m_iWordID = iWord;
tLast.m_uWords = 0;
if ( iWord )
m_iDocumentWords++;
m_iLastWord = iWord ? m_dTokens.GetLength() - 1 : m_iLastWord;
// fill word mask
if ( bFillMasks && iWord )
{
bool bMatch = false;
int iOffset;
ARRAY_FOREACH ( nWord, m_dWords )
{
const char * sKeyword = &m_dKeywordsBuffer [ m_dKeywords[nWord].m_iWord ];
const Token_t & tToken = m_dWords[nWord];
switch ( m_dKeywords[nWord].m_uStar )
{
case STAR_NONE:
bMatch = ( iWord==tToken.m_iWordID );
break;
case STAR_FRONT:
iOffset = tLast.m_iLengthBytes - tToken.m_iLengthBytes;
bMatch = ( iOffset>=0 ) &&
( memcmp ( sKeyword, sWord + iOffset, tToken.m_iLengthBytes )==0 );
break;
case STAR_BACK:
bMatch = ( tLast.m_iLengthBytes>=tToken.m_iLengthBytes ) &&
( memcmp ( sKeyword, sWord, tToken.m_iLengthBytes )==0 );
break;
case STAR_BOTH:
bMatch = strstr ( (const char *)sWord, sKeyword )!=NULL;
break;
}
if ( bMatch )
{
tLast.m_uWords |= 1UL<<nWord;
m_uFoundWords |= 1UL<<nWord;
}
}
}
}