From: dan Date: Thu, 2 Nov 2023 17:31:06 +0000 (+0000) Subject: Add the "remove_diacritics" option to the fts5 trigram tokenizer. X-Git-Tag: version-3.45.0~231^2~1 X-Git-Url: http://git.ipfire.org/?a=commitdiff_plain;h=e186fe20f5bddfd9f4d209f354dd6f5efa8e9c85;p=thirdparty%2Fsqlite.git Add the "remove_diacritics" option to the fts5 trigram tokenizer. FossilOrigin-Name: 83da80135b6105f47d1de560232449562ae8ac176c8011a6f75589f62bc9b1db --- diff --git a/ext/fts5/fts5_tokenize.c b/ext/fts5/fts5_tokenize.c index e61d6b1edd..098d0c86b7 100644 --- a/ext/fts5/fts5_tokenize.c +++ b/ext/fts5/fts5_tokenize.c @@ -227,6 +227,12 @@ static const unsigned char sqlite3Utf8Trans1[] = { } \ } +#define SKIP_UTF8(zIn) { \ + if( ((unsigned char)(*(zIn++)))>=0xc0 ){ \ + while( (((unsigned char)*zIn) & 0xc0)==0x80 ){ zIn++; } \ + } \ +} + #endif /* ifndef SQLITE_AMALGAMATION */ typedef struct Unicode61Tokenizer Unicode61Tokenizer; @@ -1264,6 +1270,7 @@ static int fts5PorterTokenize( typedef struct TrigramTokenizer TrigramTokenizer; struct TrigramTokenizer { int bFold; /* True to fold to lower-case */ + int iFoldParam; /* Parameter to pass to Fts5UnicodeFold() */ }; /* @@ -1298,10 +1305,21 @@ static int fts5TriCreate( }else{ pNew->bFold = (zArg[0]=='0'); } + }else if( 0==sqlite3_stricmp(azArg[i], "remove_diacritics") ){ + if( (zArg[0]!='0' && zArg[0]!='1' && zArg[0]!='2') || zArg[1] ){ + rc = SQLITE_ERROR; + }else{ + pNew->iFoldParam = (zArg[0]!='0') ? 2 : 0; + } }else{ rc = SQLITE_ERROR; } } + + if( pNew->iFoldParam!=0 && pNew->bFold==0 ){ + rc = SQLITE_ERROR; + } + if( rc!=SQLITE_OK ){ fts5TriDelete((Fts5Tokenizer*)pNew); pNew = 0; @@ -1324,40 +1342,62 @@ static int fts5TriTokenize( TrigramTokenizer *p = (TrigramTokenizer*)pTok; int rc = SQLITE_OK; char aBuf[32]; + char *zOut = aBuf; + int ii; const unsigned char *zIn = (const unsigned char*)pText; const unsigned char *zEof = &zIn[nText]; u32 iCode; + int aStart[3]; /* Input offset of each character in aBuf[] */ UNUSED_PARAM(unusedFlags); - while( 1 ){ - char *zOut = aBuf; - int iStart = zIn - (const unsigned char*)pText; - const unsigned char *zNext; - - READ_UTF8(zIn, zEof, iCode); - if( iCode==0 ) break; - zNext = zIn; - if( zInbFold ) iCode = sqlite3Fts5UnicodeFold(iCode, 0); - WRITE_UTF8(zOut, iCode); + + /* Populate aBuf[] with the characters for the first trigram. */ + for(ii=0; ii<3; ii++){ + do { + aStart[ii] = zIn - (const unsigned char*)pText; READ_UTF8(zIn, zEof, iCode); - if( iCode==0 ) break; - }else{ - break; - } - if( zInbFold ) iCode = sqlite3Fts5UnicodeFold(iCode, 0); - WRITE_UTF8(zOut, iCode); + if( iCode==0 ) return SQLITE_OK; + if( p->bFold ) iCode = sqlite3Fts5UnicodeFold(iCode, p->iFoldParam); + }while( iCode==0 ); + WRITE_UTF8(zOut, iCode); + } + + /* At the start of each iteration of this loop: + ** + ** aBuf: Contains 3 characters. The 3 characters of the next trigram. + ** zOut: Points to the byte following the last character in aBuf. + ** aStart[3]: Contains the byte offset in the input text corresponding + ** to the start of each of the three characters in the buffer. + */ + assert( zIn<=zEof ); + while( 1 ){ + int iNext; /* Start of character following current tri */ + const char *z1; + + /* Read characters from the input up until the first non-diacritic */ + do { + iNext = zIn - (const unsigned char*)pText; READ_UTF8(zIn, zEof, iCode); if( iCode==0 ) break; - if( p->bFold ) iCode = sqlite3Fts5UnicodeFold(iCode, 0); - WRITE_UTF8(zOut, iCode); - }else{ - break; - } - rc = xToken(pCtx, 0, aBuf, zOut-aBuf, iStart, iStart + zOut-aBuf); - if( rc!=SQLITE_OK ) break; - zIn = zNext; + if( p->bFold ) iCode = sqlite3Fts5UnicodeFold(iCode, p->iFoldParam); + }while( iCode==0 ); + + /* Pass the current trigram back to fts5 */ + rc = xToken(pCtx, 0, aBuf, zOut-aBuf, aStart[0], iNext); + if( iCode==0 || rc!=SQLITE_OK ) break; + + /* Remove the first character from buffer aBuf[]. Append the character + ** with codepoint iCode. */ + z1 = aBuf; + SKIP_UTF8(z1); + memmove(aBuf, z1, zOut - z1); + zOut -= (z1 - aBuf); + WRITE_UTF8(zOut, iCode); + + /* Update the aStart[] array */ + aStart[0] = aStart[1]; + aStart[1] = aStart[2]; + aStart[2] = iNext; } return rc; @@ -1380,7 +1420,9 @@ int sqlite3Fts5TokenizerPattern( ){ if( xCreate==fts5TriCreate ){ TrigramTokenizer *p = (TrigramTokenizer*)pTok; - return p->bFold ? FTS5_PATTERN_LIKE : FTS5_PATTERN_GLOB; + if( p->iFoldParam==0 ){ + return p->bFold ? FTS5_PATTERN_LIKE : FTS5_PATTERN_GLOB; + } } return FTS5_PATTERN_NONE; } diff --git a/ext/fts5/test/fts5trigram2.test b/ext/fts5/test/fts5trigram2.test new file mode 100644 index 0000000000..f5beae5b28 --- /dev/null +++ b/ext/fts5/test/fts5trigram2.test @@ -0,0 +1,109 @@ +# 2023 October 24 +# +# The author disclaims copyright to this source code. In place of +# a legal notice, here is a blessing: +# +# May you do good and not evil. +# May you find forgiveness for yourself and forgive others. +# May you share freely, never taking more than you give. +# +#************************************************************************* +# +# Tests for the fts5 "trigram" tokenizer. +# + +source [file join [file dirname [info script]] fts5_common.tcl] +ifcapable !fts5 { finish_test ; return } +set ::testprefix fts5trigram2 + +do_execsql_test 1.0 " + CREATE VIRTUAL TABLE t1 USING fts5(y, tokenize='trigram remove_diacritics 1'); + INSERT INTO t1 VALUES('abc\u0303defghijklm'); + INSERT INTO t1 VALUES('a\u0303b\u0303c\u0303defghijklm'); +" + +do_execsql_test 1.1 { + SELECT highlight(t1, 0, '(', ')') FROM t1('abc'); +} [list \ + "(abc\u0303)defghijklm" \ + "(a\u0303b\u0303c\u0303)defghijklm" \ +] + +do_execsql_test 1.2 { + SELECT highlight(t1, 0, '(', ')') FROM t1('bcde'); +} [list \ + "a(bc\u0303de)fghijklm" \ + "a\u0303(b\u0303c\u0303de)fghijklm" \ +] + +do_execsql_test 1.3 { + SELECT highlight(t1, 0, '(', ')') FROM t1('cdef'); +} [list \ + "ab(c\u0303def)ghijklm" \ + "a\u0303b\u0303(c\u0303def)ghijklm" \ +] + +do_execsql_test 1.4 { + SELECT highlight(t1, 0, '(', ')') FROM t1('def'); +} [list \ + "abc\u0303(def)ghijklm" \ + "a\u0303b\u0303c\u0303(def)ghijklm" \ +] + + +#------------------------------------------------------------------------- +do_catchsql_test 2.0 { + CREATE VIRTUAL TABLE t2 USING fts5( + z, tokenize='trigram case_sensitive 1 remove_diacritics 1' + ); +} {1 {error in tokenizer constructor}} + +do_execsql_test 2.1 { + CREATE VIRTUAL TABLE t2 USING fts5( + z, tokenize='trigram case_sensitive 0 remove_diacritics 1' + ); +} +do_execsql_test 2.2 " + INSERT INTO t2 VALUES('\u00E3bcdef'); + INSERT INTO t2 VALUES('b\u00E3cdef'); + INSERT INTO t2 VALUES('bc\u00E3def'); + INSERT INTO t2 VALUES('bcd\u00E3ef'); +" + +do_execsql_test 2.3 { + SELECT highlight(t2, 0, '(', ')') FROM t2('abc'); +} "(\u00E3bc)def" +do_execsql_test 2.4 { + SELECT highlight(t2, 0, '(', ')') FROM t2('bac'); +} "(b\u00E3c)def" +do_execsql_test 2.5 { + SELECT highlight(t2, 0, '(', ')') FROM t2('bca'); +} "(bc\u00E3)def" +do_execsql_test 2.6 " + SELECT highlight(t2, 0, '(', ')') FROM t2('\u00E3bc'); +" "(\u00E3bc)def" + +#------------------------------------------------------------------------- +do_execsql_test 3.0 { + CREATE VIRTUAL TABLE t3 USING fts5( + z, tokenize='trigram remove_diacritics 1' + ); +} {} +do_execsql_test 3.1 " + INSERT INTO t3 VALUES ('\u0303abc\u0303'); +" +do_execsql_test 3.2 { + SELECT highlight(t3, 0, '(', ')') FROM t3('abc'); +} "\u0303(abc\u0303)" + +#------------------------------------------------------------------------- +do_execsql_test 4.0 { + CREATE VIRTUAL TABLE t4 USING fts5(z, tokenize=trigram); +} {} + +breakpoint +do_execsql_test 4.1 { + INSERT INTO t4 VALUES('ABCD'); +} {} + +finish_test diff --git a/manifest b/manifest index df7a5eb535..ad9b0ea28f 100644 --- a/manifest +++ b/manifest @@ -1,5 +1,5 @@ -C One\smore\stweak\sto\stool/srctree-check.tcl\sso\sthat\sa\scomplete\sbuild\scan\sbe\naccomplished\sfrom\sa\sread-only\ssource\stree. -D 2023-11-02T13:10:16.285 +C Add\sthe\s"remove_diacritics"\soption\sto\sthe\sfts5\strigram\stokenizer. +D 2023-11-02T17:31:06.153 F .fossil-settings/empty-dirs dbb81e8fc0401ac46a1491ab34a7f2c7c0452f2f06b54ebb845d024ca8283ef1 F .fossil-settings/ignore-glob 35175cdfcf539b2318cb04a9901442804be81cd677d8b889fcc9149c21f239ea F LICENSE.md df5091916dbb40e6e9686186587125e1b2ff51f022cc334e886c19a0e9982724 @@ -100,7 +100,7 @@ F ext/fts5/fts5_storage.c 5d10b9bdcce5b90656cad13c7d12ad4148677d4b9e3fca0481fca5 F ext/fts5/fts5_tcl.c b1445cbe69908c411df8084a10b2485500ac70a9c747cdc8cda175a3da59d8ae F ext/fts5/fts5_test_mi.c 08c11ec968148d4cb4119d96d819f8c1f329812c568bac3684f5464be177d3ee F ext/fts5/fts5_test_tok.c 3cb0a9b508b30d17ef025ccddd26ae3dc8ddffbe76c057616e59a9aa85d36f3b -F ext/fts5/fts5_tokenize.c 5e251efb0f1af99a25ed50010ba6b1ad1250aca5921af1988fdcabe5ebc3cb43 +F ext/fts5/fts5_tokenize.c 5a895f3bf366a87cb8c0812445736248c236e7ea664a10125021662acd6b0ba4 F ext/fts5/fts5_unicode2.c eca63dbc797f8ff0572e97caf4631389c0ab900d6364861b915bdd4735973f00 F ext/fts5/fts5_varint.c e64d2113f6e1bfee0032972cffc1207b77af63319746951bf1d09885d1dadf80 F ext/fts5/fts5_vocab.c aed56169ae5c1aa9b8189c779ffeef04ed516d3c712c06914e6d91a6759f4e4a @@ -217,6 +217,7 @@ F ext/fts5/test/fts5tok1.test 1f7817499f5971450d8c4a652114b3d833393c8134e32422d0 F ext/fts5/test/fts5tok2.test dcacb32d4a2a3f0dd3215d4a3987f78ae4be21a2 F ext/fts5/test/fts5tokenizer.test ac3c9112b263a639fb0508ae73a3ee886bf4866d2153771a8e8a20c721305a43 F ext/fts5/test/fts5trigram.test 6c4e37864f3e7d90673db5563d9736d7e40080ab94d10ebdffa94c1b77941da0 +F ext/fts5/test/fts5trigram2.test 9fe4207f8a4241747aff1005258b564958588d21bfd240d6cd4c2e955d31c156 F ext/fts5/test/fts5ubsan.test 783d5a8d13ebfa169e634940228db54540780e3ba7a87ad1e4510e61440bf64b F ext/fts5/test/fts5umlaut.test a42fe2fe6387c40c49ab27ccbd070e1ae38e07f38d05926482cc0bccac9ad602 F ext/fts5/test/fts5unicode.test 17056f4efe6b0a5d4f41fdf7a7dc9af2873004562eaa899d40633b93dc95f5a9 @@ -2139,8 +2140,8 @@ F vsixtest/vsixtest.tcl 6a9a6ab600c25a91a7acc6293828957a386a8a93 F vsixtest/vsixtest.vcxproj.data 2ed517e100c66dc455b492e1a33350c1b20fbcdc F vsixtest/vsixtest.vcxproj.filters 37e51ffedcdb064aad6ff33b6148725226cd608e F vsixtest/vsixtest_TemporaryKey.pfx e5b1b036facdb453873e7084e1cae9102ccc67a0 -P 58eb5440d75fb13c1a089cb935de6fb94b4322e10e349db8f11f37b9a2fda2fc -R 18054cea9bc23d801f864db646273e7d -U drh -Z 72328c40506425c913ea7fab75e23b19 +P cba9f0601ca995ac6952aa3a83f7264f6d25aaaa2ea36b19e90cbf591077de72 +R 91c6db43050bc1c48384df29a062813f +U dan +Z e9e46724e46ba97d1a7a1bb71106fcd0 # Remove this line to create a well-formed Fossil manifest. diff --git a/manifest.uuid b/manifest.uuid index 7142949e45..c9ea7d6610 100644 --- a/manifest.uuid +++ b/manifest.uuid @@ -1 +1 @@ -cba9f0601ca995ac6952aa3a83f7264f6d25aaaa2ea36b19e90cbf591077de72 \ No newline at end of file +83da80135b6105f47d1de560232449562ae8ac176c8011a6f75589f62bc9b1db \ No newline at end of file