paulhorn/sqlite
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Add the "remove_diacritics=2" option to the unicode61 tokenizer in both FTS5

Browse Source 
and FTS3/4.

FossilOrigin-Name: 06177f3f114b5d804b84c27ac843740282e2176fdf0f7a999feda0e1b624adec
This commit is contained in:
dan
2018-12-03 16:14:49 +00:00
parent 8c53b4e7f6
commit e89feee5c3
12 changed files with 320 additions and 94 deletions
Download Patch File Download Diff File Expand all files Collapse all files
ext/fts3/fts3_unicode.c
+8 -5
View File
@@ -82,7 +82,7 @@ typedef struct unicode_cursor unicode_cursor;
struct unicode_tokenizer {
sqlite3_tokenizer base;
int bRemoveDiacritic;
int eRemoveDiacritic;
int nException;
int *aiException;
};
@@ -227,17 +227,20 @@ static int unicodeCreate(
pNew = (unicode_tokenizer *) sqlite3_malloc(sizeof(unicode_tokenizer));
if( pNew==NULL ) return SQLITE_NOMEM;
memset(pNew, 0, sizeof(unicode_tokenizer));
pNew->bRemoveDiacritic = 1;
pNew->eRemoveDiacritic = 1;
for(i=0; rc==SQLITE_OK && i<nArg; i++){
const char *z = azArg[i];
int n = (int)strlen(z);
if( n==19 && memcmp("remove_diacritics=1", z, 19)==0 ){
pNew->bRemoveDiacritic = 1;
pNew->eRemoveDiacritic = 1;
}
else if( n==19 && memcmp("remove_diacritics=0", z, 19)==0 ){
pNew->bRemoveDiacritic = 0;
pNew->eRemoveDiacritic = 0;
}
else if( n==19 && memcmp("remove_diacritics=2", z, 19)==0 ){
pNew->eRemoveDiacritic = 2;
}
else if( n>=11 && memcmp("tokenchars=", z, 11)==0 ){
rc = unicodeAddExceptions(pNew, 1, &z[11], n-11);
@@ -350,7 +353,7 @@ static int unicodeNext(
/* Write the folded case of the last character read to the output */
zEnd = z;
iOut = sqlite3FtsUnicodeFold((int)iCode, p->bRemoveDiacritic);
iOut = sqlite3FtsUnicodeFold((int)iCode, p->eRemoveDiacritic);
if( iOut ){
WRITE_UTF8(zOut, iOut);
}
ext/fts3/fts3_unicode2.c
+40 -22
View File
@@ -159,32 +159,47 @@ int sqlite3FtsUnicodeIsalnum(int c){
** E"). The resuls of passing a codepoint that corresponds to an
** uppercase letter are undefined.
*/
static int remove_diacritic(int c){
static int remove_diacritic(int c, int bComplex){
unsigned short aDia[] = {
0, 1797, 1848, 1859, 1891, 1928, 1940, 1995,
2024, 2040, 2060, 2110, 2168, 2206, 2264, 2286,
2344, 2383, 2472, 2488, 2516, 2596, 2668, 2732,
2782, 2842, 2894, 2954, 2984, 3000, 3028, 3336,
3456, 3696, 3712, 3728, 3744, 3896, 3912, 3928,
3968, 4008, 4040, 4106, 4138, 4170, 4202, 4234,
4266, 4296, 4312, 4344, 4408, 4424, 4472, 4504,
6148, 6198, 6264, 6280, 6360, 6429, 6505, 6529,
61448, 61468, 61534, 61592, 61642, 61688, 61704, 61726,
61784, 61800, 61836, 61880, 61914, 61948, 61998, 62122,
62154, 62200, 62218, 62302, 62364, 62442, 62478, 62536,
62554, 62584, 62604, 62640, 62648, 62656, 62664, 62730,
62924, 63050, 63082, 63274, 63390,
3456, 3696, 3712, 3728, 3744, 3766, 3832, 3896,
3912, 3928, 3944, 3968, 4008, 4040, 4056, 4106,
4138, 4170, 4202, 4234, 4266, 4296, 4312, 4344,
4408, 4424, 4442, 4472, 4488, 4504, 6148, 6198,
6264, 6280, 6360, 6429, 6505, 6529, 61448, 61468,
61512, 61534, 61592, 61610, 61642, 61672, 61688, 61704,
61726, 61784, 61800, 61816, 61836, 61880, 61896, 61914,
61948, 61998, 62062, 62122, 62154, 62184, 62200, 62218,
62252, 62302, 62364, 62410, 62442, 62478, 62536, 62554,
62584, 62604, 62640, 62648, 62656, 62664, 62730, 62766,
62830, 62890, 62924, 62974, 63032, 63050, 63082, 63118,
63182, 63242, 63274, 63310, 63368, 63390,
};
char aChar[] = {
'\0', 'a', 'c', 'e', 'i', 'n', 'o', 'u', 'y', 'y', 'a', 'c',
'd', 'e', 'e', 'g', 'h', 'i', 'j', 'k', 'l', 'n', 'o', 'r',
's', 't', 'u', 'u', 'w', 'y', 'z', 'o', 'u', 'a', 'i', 'o',
'u', 'g', 'k', 'o', 'j', 'g', 'n', 'a', 'e', 'i', 'o', 'r',
'u', 's', 't', 'h', 'a', 'e', 'o', 'y', '\0', '\0', '\0', '\0',
'\0', '\0', '\0', '\0', 'a', 'b', 'd', 'd', 'e', 'f', 'g', 'h',
'h', 'i', 'k', 'l', 'l', 'm', 'n', 'p', 'r', 'r', 's', 't',
'u', 'v', 'w', 'w', 'x', 'y', 'z', 'h', 't', 'w', 'y', 'a',
'e', 'i', 'o', 'u', 'y',
'\0', 'a'|0x00, 'c'|0x00, 'e'|0x00, 'i'|0x00, 'n'|0x00,
'o'|0x00, 'u'|0x00, 'y'|0x00, 'y'|0x00, 'a'|0x00, 'c'|0x00,
'd'|0x00, 'e'|0x00, 'e'|0x00, 'g'|0x00, 'h'|0x00, 'i'|0x00,
'j'|0x00, 'k'|0x00, 'l'|0x00, 'n'|0x00, 'o'|0x00, 'r'|0x00,
's'|0x00, 't'|0x00, 'u'|0x00, 'u'|0x00, 'w'|0x00, 'y'|0x00,
'z'|0x00, 'o'|0x00, 'u'|0x00, 'a'|0x00, 'i'|0x00, 'o'|0x00,
'u'|0x00, 'u'|0x80, 'a'|0x80, 'g'|0x00, 'k'|0x00, 'o'|0x00,
'o'|0x80, 'j'|0x00, 'g'|0x00, 'n'|0x00, 'a'|0x80, 'a'|0x00,
'e'|0x00, 'i'|0x00, 'o'|0x00, 'r'|0x00, 'u'|0x00, 's'|0x00,
't'|0x00, 'h'|0x00, 'a'|0x00, 'e'|0x00, 'o'|0x80, 'o'|0x00,
'o'|0x80, 'y'|0x00, '\0', '\0', '\0', '\0',
'\0', '\0', '\0', '\0', 'a'|0x00, 'b'|0x00,
'c'|0x80, 'd'|0x00, 'd'|0x00, 'e'|0x80, 'e'|0x00, 'e'|0x80,
'f'|0x00, 'g'|0x00, 'h'|0x00, 'h'|0x00, 'i'|0x00, 'i'|0x80,
'k'|0x00, 'l'|0x00, 'l'|0x80, 'l'|0x00, 'm'|0x00, 'n'|0x00,
'o'|0x80, 'p'|0x00, 'r'|0x00, 'r'|0x80, 'r'|0x00, 's'|0x00,
's'|0x80, 't'|0x00, 'u'|0x00, 'u'|0x80, 'v'|0x00, 'w'|0x00,
'w'|0x00, 'x'|0x00, 'y'|0x00, 'z'|0x00, 'h'|0x00, 't'|0x00,
'w'|0x00, 'y'|0x00, 'a'|0x00, 'a'|0x80, 'a'|0x80, 'a'|0x80,
'e'|0x00, 'e'|0x80, 'e'|0x80, 'i'|0x00, 'o'|0x00, 'o'|0x80,
'o'|0x80, 'o'|0x80, 'u'|0x00, 'u'|0x80, 'u'|0x80, 'y'|0x00,
};
unsigned int key = (((unsigned int)c)<<3) | 0x00000007;
@@ -201,7 +216,8 @@ static int remove_diacritic(int c){
}
}
assert( key>=aDia[iRes] );
return ((c > (aDia[iRes]>>3) + (aDia[iRes]&0x07)) ? c : (int)aChar[iRes]);
if( bComplex==0 && (aChar[iRes] & 0x80) ) return c;
return (c > (aDia[iRes]>>3) + (aDia[iRes]&0x07)) ? c : ((int)aChar[iRes] & 0x7F);
}
@@ -228,7 +244,7 @@ int sqlite3FtsUnicodeIsdiacritic(int c){
** The results are undefined if the value passed to this function
** is less than zero.
*/
int sqlite3FtsUnicodeFold(int c, int bRemoveDiacritic){
int sqlite3FtsUnicodeFold(int c, int eRemoveDiacritic){
/* Each entry in the following array defines a rule for folding a range
** of codepoints to lower case. The rule applies to a range of nRange
** codepoints starting at codepoint iCode.
@@ -351,7 +367,9 @@ int sqlite3FtsUnicodeFold(int c, int bRemoveDiacritic){
assert( ret>0 );
}
if( bRemoveDiacritic ) ret = remove_diacritic(ret);
if( eRemoveDiacritic ){
ret = remove_diacritic(ret, eRemoveDiacritic==2);
}
}
else if( c>=66560 && c<66600 ){
ext/fts3/unicode/mkunicode.tcl
+24 -12
View File
@@ -9,11 +9,12 @@ proc print_rd {map} {
set nRange 1
set iFirst [lindex $map 0 0]
set cPrev [lindex $map 0 1]
set fPrev [lindex $map 0 2]
foreach m [lrange $map 1 end] {
foreach {i c} $m {}
foreach {i c f} $m {}
if {$cPrev == $c} {
if {$cPrev == $c && $fPrev==$f} {
for {set j [expr $iFirst+$nRange]} {$j<$i} {incr j} {
if {[info exists tl_lookup_table($j)]==0} break
}
@@ -29,13 +30,16 @@ proc print_rd {map} {
lappend lRange [list $iFirst $nRange]
lappend aChar $cPrev
lappend aFlag $fPrev
set iFirst $i
set cPrev $c
set fPrev $f
set nRange 1
}
lappend lRange [list $iFirst $nRange]
lappend aChar $cPrev
lappend aFlag $fPrev
puts "/*"
puts "** If the argument is a codepoint corresponding to a lowercase letter"
@@ -45,7 +49,7 @@ proc print_rd {map} {
puts "** E\"). The resuls of passing a codepoint that corresponds to an"
puts "** uppercase letter are undefined."
puts "*/"
puts "static int ${::remove_diacritic}(int c)\{"
puts "static int ${::remove_diacritic}(int c, int bComplex)\{"
puts " unsigned short aDia\[\] = \{"
puts -nonewline " 0, "
set i 1
@@ -60,13 +64,17 @@ proc print_rd {map} {
puts ""
puts " \};"
puts " char aChar\[\] = \{"
puts -nonewline " '\\0', "
puts -nonewline " '\\0', "
set i 1
foreach c $aChar {
set str "'$c', "
if {$c == ""} { set str "'\\0', " }
foreach c $aChar f $aFlag {
if { $f } {
set str "'$c'|0x80, "
} else {
set str "'$c'|0x00, "
}
if {$c == ""} { set str "'\\0', " }
if {($i % 12)==0} {puts "" ; puts -nonewline " " }
if {($i % 6)==0} {puts "" ; puts -nonewline " " }
incr i
puts -nonewline "$str"
}
@@ -87,7 +95,8 @@ proc print_rd {map} {
}
}
assert( key>=aDia[iRes] );
return ((c > (aDia[iRes]>>3) + (aDia[iRes]&0x07)) ? c : (int)aChar[iRes]);}
if( bComplex==0 && (aChar[iRes] & 0x80) ) return c;
return (c > (aDia[iRes]>>3) + (aDia[iRes]&0x07)) ? c : ((int)aChar[iRes] & 0x7F);}
puts "\}"
}
@@ -95,7 +104,8 @@ proc print_isdiacritic {zFunc map} {
set lCode [list]
foreach m $map {
foreach {code char} $m {}
foreach {code char flag} $m {}
if {$flag} continue
if {$code && $char == ""} { lappend lCode $code }
}
set lCode [lsort -integer $lCode]
@@ -472,7 +482,7 @@ proc print_fold {zFunc} {
puts "** The results are undefined if the value passed to this function"
puts "** is less than zero."
puts "*/"
puts "int ${zFunc}\(int c, int bRemoveDiacritic)\{"
puts "int ${zFunc}\(int c, int eRemoveDiacritic)\{"
set liOff [tl_generate_ioff_table $lRecord]
tl_print_table_header
@@ -516,7 +526,9 @@ proc print_fold {zFunc} {
assert( ret>0 );
}
if( bRemoveDiacritic ) ret = ${::remove_diacritic}(ret);
if( eRemoveDiacritic ){
ret = ${::remove_diacritic}(ret, eRemoveDiacritic==2);
}
}
}]
ext/fts3/unicode/parseunicode.tcl
+27 -4
View File
@@ -7,12 +7,24 @@
# character that it should be replaced with, or an empty string if the
# codepoint should simply be removed from the input. Examples:
#
# { 224 a } (replace codepoint 224 to "a")
# { 769 "" } (remove codepoint 769 from input)
# { 224 a 0 } (replace codepoint 224 to "a")
# { 769 "" 0 } (remove codepoint 769 from input)
#
# Mappings are only returned for non-upper case codepoints. It is assumed
# that the input has already been folded to lower case.
#
# The third value in the list is always either 0 or 1. 0 if the
# UnicodeData.txt file maps the codepoint to a single ASCII character and
# a diacritic, or 1 if the mapping is indirect. For example, consider the
# two entries:
#
# 1ECD;LATIN SMALL LETTER O WITH DOT BELOW;Ll;0;L;006F 0323;;;;N;;;1ECC;;1ECC
# 1ED9;LATIN SMALL LETTER O WITH CIRCUMFLEX AND DOT BELOW;Ll;0;L;1ECD 0302;;;;N;;;1ED8;;1ED8
#
# The first codepoint is a direct mapping (as 006F is ASCII and 0323 is a
# diacritic). The second is an indirect mapping, as it maps to the
# first codepoint plus 0302 (a diacritic).
#
proc rd_load_unicodedata_text {zName} {
global tl_lookup_table
@@ -53,18 +65,29 @@ proc rd_load_unicodedata_text {zName} {
set iAscii [expr "0x[lindex $character_decomposition_mapping 0]"]
set iDia [expr "0x[lindex $character_decomposition_mapping 1]"]
# Filter out upper-case characters, as they will be mapped to their
# lower-case equivalents before this data is used.
if {[info exists tl_lookup_table($iCode)]} continue
# Check if this is an indirect mapping. If so, set bIndirect to true
# and change $iAscii to the indirectly mappped ASCII character.
set bIndirect 0
if {[info exists dia($iDia)] && [info exists mapping($iAscii)]} {
set iAscii $mapping($iAscii)
set bIndirect 1
}
if { ($iAscii >= 97 && $iAscii <= 122)
|| ($iAscii >= 65 && $iAscii <= 90)
} {
lappend lRet [list $iCode [string tolower [format %c $iAscii]]]
lappend lRet [list $iCode [string tolower [format %c $iAscii]] $bIndirect]
set mapping($iCode) $iAscii
set dia($iDia) 1
}
}
foreach d [array names dia] {
lappend lRet [list $d ""]
lappend lRet [list $d "" 0]
}
set lRet [lsort -integer -index 0 $lRet]