Merge branch '195_utf8_substring' into 'master'

fixed #195 utf8 substring matching with min lengths defined

Closes #195

See merge request manticoresearch/dev!27

src/sphinx.cpp

@@ -17679,25 +17679,27 @@ bool sphExpandGetWords ( const char * sWord, const ExpansionContext_t & tCtx, IS
 
 		// compute non-wildcard prefix length
 		int iPrefix = 0;
-		for ( const char * s = sPrefix; *s && !sphIsWild ( *s ); s++ )
+		const char * sCodes = sPrefix;
+		for ( ; *sCodes && !sphIsWild ( *sCodes ); sCodes+=sphUtf8CharBytes ( *sCodes ) )
 			iPrefix++;
 
 		// do not expand prefixes under min length
 		int iMinLen = Max ( tCtx.m_iMinPrefixLen, tCtx.m_iMinInfixLen );
 		if ( iPrefix<iMinLen )
 			return false;
 
+		int iBytes = sCodes - sPrefix;
 		// prefix expansion should work on nonstemmed words only
 		char sFixed[MAX_KEYWORD_BYTES];
 		if ( tCtx.m_bHasMorphology )
 		{
 			sFixed[0] = MAGIC_WORD_HEAD_NONSTEMMED;
-			memcpy ( sFixed+1, sPrefix, iPrefix );
+			memcpy ( sFixed+1, sPrefix, iBytes );
 			sPrefix = sFixed;
-			iPrefix++;
+			iBytes++;
 		}
 
-		tCtx.m_pWordlist->GetPrefixedWords ( sPrefix, iPrefix, sWildcard, tWordlist );
+		tCtx.m_pWordlist->GetPrefixedWords ( sPrefix, iBytes, sWildcard, tWordlist );
 
 	} else
 	{
@@ -17706,28 +17708,40 @@ bool sphExpandGetWords ( const char * sWord, const ExpansionContext_t & tCtx, IS
 		assert ( tCtx.m_iMinInfixLen>0 );
 
 		// find the longest substring of non-wildcards
+		int iCodepoints = 0;
+		int iInfixCodepoints = 0;
+		int iInfixBytes = 0;
 		const char * sMaxInfix = NULL;
-		int iMaxInfix = 0;
-		int iCur = 0;
+		const char * sInfix = sWord;
 
-		for ( const char * s = sWord; *s; s++ )
+		for ( const char * s = sWord; *s; )
 		{
+			int iCodeLen = sphUtf8CharBytes ( *s );
+
 			if ( sphIsWild ( *s ) )
 			{
-				iCur = 0;
-			} else if ( ++iCur > iMaxInfix )
+				sInfix = s + 1;
+				iCodepoints = 0;
+			} else
 			{
-				sMaxInfix = s-iCur+1;
-				iMaxInfix = iCur;
+				iCodepoints++;
+				if ( s - sInfix + iCodeLen > iInfixBytes )
+				{
+					sMaxInfix = sInfix;
+					iInfixBytes = s - sInfix + iCodeLen;
+					iInfixCodepoints = iCodepoints;
+				}
 			}
+
+			s += iCodeLen;
 		}
 
 		// do not expand infixes under min_infix_len
-		if ( iMaxInfix < tCtx.m_iMinInfixLen )
+		if ( iInfixCodepoints < tCtx.m_iMinInfixLen )
 			return false;
 
 		// ignore heading star
-		tCtx.m_pWordlist->GetInfixedWords ( sMaxInfix, iMaxInfix, sWord, tWordlist );
+		tCtx.m_pWordlist->GetInfixedWords ( sMaxInfix, iInfixBytes, sWord, tWordlist );
 	}
 
 	return true;

test/test_350/model.bin

@@ -0,0 +1 @@
+a:1:{i:0;a:8:{i:0;a:2:{s:8:"sphinxql";s:42:"select * from test where match ('как*')";s:10:"total_rows";i:0;}i:1;a:3:{s:8:"sphinxql";s:13:"show warnings";s:10:"total_rows";i:1;s:4:"rows";a:1:{i:0;a:3:{s:5:"Level";s:7:"warning";s:4:"Code";s:4:"1000";s:7:"Message";s:65:"Query word length is less than min infix length. word: 'как*' ";}}}i:2;a:3:{s:8:"sphinxql";s:44:"select * from test where match ('какт*')";s:10:"total_rows";i:2;s:4:"rows";a:2:{i:0;a:3:{s:2:"id";s:2:"11";s:3:"idd";s:2:"11";s:4:"body";s:25:"както неочень";}i:1;a:3:{s:2:"id";s:2:"12";s:3:"idd";s:2:"12";s:4:"body";s:27:"кактотак очень";}}}i:3;a:2:{s:8:"sphinxql";s:13:"show warnings";s:14:"total_affected";i:0;}i:4;a:2:{s:8:"sphinxql";s:43:"select * from test where match ('*акт*')";s:10:"total_rows";i:0;}i:5;a:3:{s:8:"sphinxql";s:13:"show warnings";s:10:"total_rows";i:1;s:4:"rows";a:1:{i:0;a:3:{s:5:"Level";s:7:"warning";s:4:"Code";s:4:"1000";s:7:"Message";s:66:"Query word length is less than min infix length. word: '*акт*' ";}}}i:6;a:3:{s:8:"sphinxql";s:45:"select * from test where match ('*акто*')";s:10:"total_rows";i:2;s:4:"rows";a:2:{i:0;a:3:{s:2:"id";s:2:"11";s:3:"idd";s:2:"11";s:4:"body";s:25:"както неочень";}i:1;a:3:{s:2:"id";s:2:"12";s:3:"idd";s:2:"12";s:4:"body";s:27:"кактотак очень";}}}i:7;a:2:{s:8:"sphinxql";s:13:"show warnings";s:14:"total_affected";i:0;}}}

test/test_350/test.xml

@@ -0,0 +1,68 @@
+<?xml version="1.0" encoding="utf-8"?>
+<test>
+
+<name>warnings from short utf8 prefix and infix</name>
+
+<config>
+
+indexer
+{
+	mem_limit = 16M
+}
+
+searchd
+{
+	<searchd_settings/>
+}
+
+source test
+{
+	type = mysql
+	<sql_settings/>
+	sql_query_pre 		= SET NAMES utf8
+	sql_query 			= SELECT * FROM test_table
+	sql_attr_uint		= idd
+    sql_field_string    = body
+}
+
+index test
+{
+	source			= test
+	path = <data_path/>/plain1
+	docinfo			= extern
+	dict = keywords
+    charset_table = a..z, _, A..Z->a..z, russian
+	min_infix_len = 4
+}
+</config>
+
+<db_create>
+CREATE TABLE test_table
+(
+	id INTEGER PRIMARY KEY NOT NULL,
+	idd INTEGER NOT NULL,
+	body VARCHAR(255) NOT NULL
+) CHARACTER SET utf8;
+</db_create>
+<db_drop>DROP TABLE IF EXISTS test_table</db_drop>
+<db_insert>SET NAMES utf8</db_insert>
+<db_insert>INSERT INTO test_table VALUES
+( 1, 1, 'test' ), (11, 11, 'както неочень') , (12, 12, 'кактотак очень')
+</db_insert>
+
+
+<sphqueries>
+
+<!-- regression: warning from utf8 term -->
+<sphinxql>select * from test where match ('как*')</sphinxql>
+<sphinxql>show warnings</sphinxql>
+<sphinxql>select * from test where match ('какт*')</sphinxql>
+<sphinxql>show warnings</sphinxql>
+<sphinxql>select * from test where match ('*акт*')</sphinxql>
+<sphinxql>show warnings</sphinxql>
+<sphinxql>select * from test where match ('*акто*')</sphinxql>
+<sphinxql>show warnings</sphinxql>
+
+</sphqueries>
+
+</test>