742 lines
24 KiB
C++
742 lines
24 KiB
C++
// © 2016 and later: Unicode, Inc. and others.
|
|
// License & terms of use: http://www.unicode.org/copyright.html
|
|
/*
|
|
******************************************************************************
|
|
*
|
|
* Copyright (C) 2007-2012, International Business Machines
|
|
* Corporation and others. All Rights Reserved.
|
|
*
|
|
******************************************************************************
|
|
* file name: bmpset.cpp
|
|
* encoding: UTF-8
|
|
* tab size: 8 (not used)
|
|
* indentation:4
|
|
*
|
|
* created on: 2007jan29
|
|
* created by: Markus W. Scherer
|
|
*/
|
|
|
|
#include "unicode/utypes.h"
|
|
#include "unicode/uniset.h"
|
|
#include "unicode/utf8.h"
|
|
#include "unicode/utf16.h"
|
|
#include "cmemory.h"
|
|
#include "bmpset.h"
|
|
#include "uassert.h"
|
|
|
|
U_NAMESPACE_BEGIN
|
|
|
|
BMPSet::BMPSet(const int32_t *parentList, int32_t parentListLength) :
|
|
list(parentList), listLength(parentListLength) {
|
|
uprv_memset(latin1Contains, 0, sizeof(latin1Contains));
|
|
uprv_memset(table7FF, 0, sizeof(table7FF));
|
|
uprv_memset(bmpBlockBits, 0, sizeof(bmpBlockBits));
|
|
|
|
/*
|
|
* Set the list indexes for binary searches for
|
|
* U+0800, U+1000, U+2000, .., U+F000, U+10000.
|
|
* U+0800 is the first 3-byte-UTF-8 code point. Lower code points are
|
|
* looked up in the bit tables.
|
|
* The last pair of indexes is for finding supplementary code points.
|
|
*/
|
|
list4kStarts[0]=findCodePoint(0x800, 0, listLength-1);
|
|
int32_t i;
|
|
for(i=1; i<=0x10; ++i) {
|
|
list4kStarts[i]=findCodePoint(i<<12, list4kStarts[i-1], listLength-1);
|
|
}
|
|
list4kStarts[0x11]=listLength-1;
|
|
containsFFFD=containsSlow(0xfffd, list4kStarts[0xf], list4kStarts[0x10]);
|
|
|
|
initBits();
|
|
overrideIllegal();
|
|
}
|
|
|
|
BMPSet::BMPSet(const BMPSet &otherBMPSet, const int32_t *newParentList, int32_t newParentListLength) :
|
|
containsFFFD(otherBMPSet.containsFFFD),
|
|
list(newParentList), listLength(newParentListLength) {
|
|
uprv_memcpy(latin1Contains, otherBMPSet.latin1Contains, sizeof(latin1Contains));
|
|
uprv_memcpy(table7FF, otherBMPSet.table7FF, sizeof(table7FF));
|
|
uprv_memcpy(bmpBlockBits, otherBMPSet.bmpBlockBits, sizeof(bmpBlockBits));
|
|
uprv_memcpy(list4kStarts, otherBMPSet.list4kStarts, sizeof(list4kStarts));
|
|
}
|
|
|
|
BMPSet::~BMPSet() {
|
|
}
|
|
|
|
/*
|
|
* Set bits in a bit rectangle in "vertical" bit organization.
|
|
* start<limit<=0x800
|
|
*/
|
|
static void set32x64Bits(uint32_t table[64], int32_t start, int32_t limit) {
|
|
U_ASSERT(start<limit);
|
|
U_ASSERT(limit<=0x800);
|
|
|
|
int32_t lead=start>>6; // Named for UTF-8 2-byte lead byte with upper 5 bits.
|
|
int32_t trail=start&0x3f; // Named for UTF-8 2-byte trail byte with lower 6 bits.
|
|
|
|
// Set one bit indicating an all-one block.
|
|
uint32_t bits=(uint32_t)1<<lead;
|
|
if((start+1)==limit) { // Single-character shortcut.
|
|
table[trail]|=bits;
|
|
return;
|
|
}
|
|
|
|
int32_t limitLead=limit>>6;
|
|
int32_t limitTrail=limit&0x3f;
|
|
|
|
if(lead==limitLead) {
|
|
// Partial vertical bit column.
|
|
while(trail<limitTrail) {
|
|
table[trail++]|=bits;
|
|
}
|
|
} else {
|
|
// Partial vertical bit column,
|
|
// followed by a bit rectangle,
|
|
// followed by another partial vertical bit column.
|
|
if(trail>0) {
|
|
do {
|
|
table[trail++]|=bits;
|
|
} while(trail<64);
|
|
++lead;
|
|
}
|
|
if(lead<limitLead) {
|
|
bits=~(((unsigned)1<<lead)-1);
|
|
if(limitLead<0x20) {
|
|
bits&=((unsigned)1<<limitLead)-1;
|
|
}
|
|
for(trail=0; trail<64; ++trail) {
|
|
table[trail]|=bits;
|
|
}
|
|
}
|
|
// limit<=0x800. If limit==0x800 then limitLead=32 and limitTrail=0.
|
|
// In that case, bits=1<<limitLead is undefined but the bits value
|
|
// is not used because trail<limitTrail is already false.
|
|
bits=(uint32_t)1<<((limitLead == 0x20) ? (limitLead - 1) : limitLead);
|
|
for(trail=0; trail<limitTrail; ++trail) {
|
|
table[trail]|=bits;
|
|
}
|
|
}
|
|
}
|
|
|
|
void BMPSet::initBits() {
|
|
UChar32 start, limit;
|
|
int32_t listIndex=0;
|
|
|
|
// Set latin1Contains[].
|
|
do {
|
|
start=list[listIndex++];
|
|
if(listIndex<listLength) {
|
|
limit=list[listIndex++];
|
|
} else {
|
|
limit=0x110000;
|
|
}
|
|
if(start>=0x100) {
|
|
break;
|
|
}
|
|
do {
|
|
latin1Contains[start++]=1;
|
|
} while(start<limit && start<0x100);
|
|
} while(limit<=0x100);
|
|
|
|
// Find the first range overlapping with (or after) 80..FF again,
|
|
// to include them in table7FF as well.
|
|
for(listIndex=0;;) {
|
|
start=list[listIndex++];
|
|
if(listIndex<listLength) {
|
|
limit=list[listIndex++];
|
|
} else {
|
|
limit=0x110000;
|
|
}
|
|
if(limit>0x80) {
|
|
if(start<0x80) {
|
|
start=0x80;
|
|
}
|
|
break;
|
|
}
|
|
}
|
|
|
|
// Set table7FF[].
|
|
while(start<0x800) {
|
|
set32x64Bits(table7FF, start, limit<=0x800 ? limit : 0x800);
|
|
if(limit>0x800) {
|
|
start=0x800;
|
|
break;
|
|
}
|
|
|
|
start=list[listIndex++];
|
|
if(listIndex<listLength) {
|
|
limit=list[listIndex++];
|
|
} else {
|
|
limit=0x110000;
|
|
}
|
|
}
|
|
|
|
// Set bmpBlockBits[].
|
|
int32_t minStart=0x800;
|
|
while(start<0x10000) {
|
|
if(limit>0x10000) {
|
|
limit=0x10000;
|
|
}
|
|
|
|
if(start<minStart) {
|
|
start=minStart;
|
|
}
|
|
if(start<limit) { // Else: Another range entirely in a known mixed-value block.
|
|
if(start&0x3f) {
|
|
// Mixed-value block of 64 code points.
|
|
start>>=6;
|
|
bmpBlockBits[start&0x3f]|=0x10001<<(start>>6);
|
|
start=(start+1)<<6; // Round up to the next block boundary.
|
|
minStart=start; // Ignore further ranges in this block.
|
|
}
|
|
if(start<limit) {
|
|
if(start<(limit&~0x3f)) {
|
|
// Multiple all-ones blocks of 64 code points each.
|
|
set32x64Bits(bmpBlockBits, start>>6, limit>>6);
|
|
}
|
|
|
|
if(limit&0x3f) {
|
|
// Mixed-value block of 64 code points.
|
|
limit>>=6;
|
|
bmpBlockBits[limit&0x3f]|=0x10001<<(limit>>6);
|
|
limit=(limit+1)<<6; // Round up to the next block boundary.
|
|
minStart=limit; // Ignore further ranges in this block.
|
|
}
|
|
}
|
|
}
|
|
|
|
if(limit==0x10000) {
|
|
break;
|
|
}
|
|
|
|
start=list[listIndex++];
|
|
if(listIndex<listLength) {
|
|
limit=list[listIndex++];
|
|
} else {
|
|
limit=0x110000;
|
|
}
|
|
}
|
|
}
|
|
|
|
/*
|
|
* Override some bits and bytes to the result of contains(FFFD)
|
|
* for faster validity checking at runtime.
|
|
* No need to set 0 values where they were reset to 0 in the constructor
|
|
* and not modified by initBits().
|
|
* (table7FF[] 0..7F, bmpBlockBits[] 0..7FF)
|
|
* Need to set 0 values for surrogates D800..DFFF.
|
|
*/
|
|
void BMPSet::overrideIllegal() {
|
|
uint32_t bits, mask;
|
|
int32_t i;
|
|
|
|
if(containsFFFD) {
|
|
bits=3; // Lead bytes 0xC0 and 0xC1.
|
|
for(i=0; i<64; ++i) {
|
|
table7FF[i]|=bits;
|
|
}
|
|
|
|
bits=1; // Lead byte 0xE0.
|
|
for(i=0; i<32; ++i) { // First half of 4k block.
|
|
bmpBlockBits[i]|=bits;
|
|
}
|
|
|
|
mask= static_cast<uint32_t>(~(0x10001<<0xd)); // Lead byte 0xED.
|
|
bits=1<<0xd;
|
|
for(i=32; i<64; ++i) { // Second half of 4k block.
|
|
bmpBlockBits[i]=(bmpBlockBits[i]&mask)|bits;
|
|
}
|
|
} else {
|
|
mask= static_cast<uint32_t>(~(0x10001<<0xd)); // Lead byte 0xED.
|
|
for(i=32; i<64; ++i) { // Second half of 4k block.
|
|
bmpBlockBits[i]&=mask;
|
|
}
|
|
}
|
|
}
|
|
|
|
int32_t BMPSet::findCodePoint(UChar32 c, int32_t lo, int32_t hi) const {
|
|
/* Examples:
|
|
findCodePoint(c)
|
|
set list[] c=0 1 3 4 7 8
|
|
=== ============== ===========
|
|
[] [110000] 0 0 0 0 0 0
|
|
[\u0000-\u0003] [0, 4, 110000] 1 1 1 2 2 2
|
|
[\u0004-\u0007] [4, 8, 110000] 0 0 0 1 1 2
|
|
[:Any:] [0, 110000] 1 1 1 1 1 1
|
|
*/
|
|
|
|
// Return the smallest i such that c < list[i]. Assume
|
|
// list[len - 1] == HIGH and that c is legal (0..HIGH-1).
|
|
if (c < list[lo])
|
|
return lo;
|
|
// High runner test. c is often after the last range, so an
|
|
// initial check for this condition pays off.
|
|
if (lo >= hi || c >= list[hi-1])
|
|
return hi;
|
|
// invariant: c >= list[lo]
|
|
// invariant: c < list[hi]
|
|
for (;;) {
|
|
int32_t i = (lo + hi) >> 1;
|
|
if (i == lo) {
|
|
break; // Found!
|
|
} else if (c < list[i]) {
|
|
hi = i;
|
|
} else {
|
|
lo = i;
|
|
}
|
|
}
|
|
return hi;
|
|
}
|
|
|
|
UBool
|
|
BMPSet::contains(UChar32 c) const {
|
|
if((uint32_t)c<=0xff) {
|
|
return (UBool)latin1Contains[c];
|
|
} else if((uint32_t)c<=0x7ff) {
|
|
return (UBool)((table7FF[c&0x3f]&((uint32_t)1<<(c>>6)))!=0);
|
|
} else if((uint32_t)c<0xd800 || (c>=0xe000 && c<=0xffff)) {
|
|
int lead=c>>12;
|
|
uint32_t twoBits=(bmpBlockBits[(c>>6)&0x3f]>>lead)&0x10001;
|
|
if(twoBits<=1) {
|
|
// All 64 code points with the same bits 15..6
|
|
// are either in the set or not.
|
|
return (UBool)twoBits;
|
|
} else {
|
|
// Look up the code point in its 4k block of code points.
|
|
return containsSlow(c, list4kStarts[lead], list4kStarts[lead+1]);
|
|
}
|
|
} else if((uint32_t)c<=0x10ffff) {
|
|
// surrogate or supplementary code point
|
|
return containsSlow(c, list4kStarts[0xd], list4kStarts[0x11]);
|
|
} else {
|
|
// Out-of-range code points get false, consistent with long-standing
|
|
// behavior of UnicodeSet::contains(c).
|
|
return false;
|
|
}
|
|
}
|
|
|
|
/*
|
|
* Check for sufficient length for trail unit for each surrogate pair.
|
|
* Handle single surrogates as surrogate code points as usual in ICU.
|
|
*/
|
|
const char16_t *
|
|
BMPSet::span(const char16_t *s, const char16_t *limit, USetSpanCondition spanCondition) const {
|
|
char16_t c, c2;
|
|
|
|
if(spanCondition) {
|
|
// span
|
|
do {
|
|
c=*s;
|
|
if(c<=0xff) {
|
|
if(!latin1Contains[c]) {
|
|
break;
|
|
}
|
|
} else if(c<=0x7ff) {
|
|
if((table7FF[c&0x3f]&((uint32_t)1<<(c>>6)))==0) {
|
|
break;
|
|
}
|
|
} else if(c<0xd800 || c>=0xe000) {
|
|
int lead=c>>12;
|
|
uint32_t twoBits=(bmpBlockBits[(c>>6)&0x3f]>>lead)&0x10001;
|
|
if(twoBits<=1) {
|
|
// All 64 code points with the same bits 15..6
|
|
// are either in the set or not.
|
|
if(twoBits==0) {
|
|
break;
|
|
}
|
|
} else {
|
|
// Look up the code point in its 4k block of code points.
|
|
if(!containsSlow(c, list4kStarts[lead], list4kStarts[lead+1])) {
|
|
break;
|
|
}
|
|
}
|
|
} else if(c>=0xdc00 || (s+1)==limit || (c2=s[1])<0xdc00 || c2>=0xe000) {
|
|
// surrogate code point
|
|
if(!containsSlow(c, list4kStarts[0xd], list4kStarts[0xe])) {
|
|
break;
|
|
}
|
|
} else {
|
|
// surrogate pair
|
|
if(!containsSlow(U16_GET_SUPPLEMENTARY(c, c2), list4kStarts[0x10], list4kStarts[0x11])) {
|
|
break;
|
|
}
|
|
++s;
|
|
}
|
|
} while(++s<limit);
|
|
} else {
|
|
// span not
|
|
do {
|
|
c=*s;
|
|
if(c<=0xff) {
|
|
if(latin1Contains[c]) {
|
|
break;
|
|
}
|
|
} else if(c<=0x7ff) {
|
|
if((table7FF[c&0x3f]&((uint32_t)1<<(c>>6)))!=0) {
|
|
break;
|
|
}
|
|
} else if(c<0xd800 || c>=0xe000) {
|
|
int lead=c>>12;
|
|
uint32_t twoBits=(bmpBlockBits[(c>>6)&0x3f]>>lead)&0x10001;
|
|
if(twoBits<=1) {
|
|
// All 64 code points with the same bits 15..6
|
|
// are either in the set or not.
|
|
if(twoBits!=0) {
|
|
break;
|
|
}
|
|
} else {
|
|
// Look up the code point in its 4k block of code points.
|
|
if(containsSlow(c, list4kStarts[lead], list4kStarts[lead+1])) {
|
|
break;
|
|
}
|
|
}
|
|
} else if(c>=0xdc00 || (s+1)==limit || (c2=s[1])<0xdc00 || c2>=0xe000) {
|
|
// surrogate code point
|
|
if(containsSlow(c, list4kStarts[0xd], list4kStarts[0xe])) {
|
|
break;
|
|
}
|
|
} else {
|
|
// surrogate pair
|
|
if(containsSlow(U16_GET_SUPPLEMENTARY(c, c2), list4kStarts[0x10], list4kStarts[0x11])) {
|
|
break;
|
|
}
|
|
++s;
|
|
}
|
|
} while(++s<limit);
|
|
}
|
|
return s;
|
|
}
|
|
|
|
/* Symmetrical with span(). */
|
|
const char16_t *
|
|
BMPSet::spanBack(const char16_t *s, const char16_t *limit, USetSpanCondition spanCondition) const {
|
|
char16_t c, c2;
|
|
|
|
if(spanCondition) {
|
|
// span
|
|
for(;;) {
|
|
c=*(--limit);
|
|
if(c<=0xff) {
|
|
if(!latin1Contains[c]) {
|
|
break;
|
|
}
|
|
} else if(c<=0x7ff) {
|
|
if((table7FF[c&0x3f]&((uint32_t)1<<(c>>6)))==0) {
|
|
break;
|
|
}
|
|
} else if(c<0xd800 || c>=0xe000) {
|
|
int lead=c>>12;
|
|
uint32_t twoBits=(bmpBlockBits[(c>>6)&0x3f]>>lead)&0x10001;
|
|
if(twoBits<=1) {
|
|
// All 64 code points with the same bits 15..6
|
|
// are either in the set or not.
|
|
if(twoBits==0) {
|
|
break;
|
|
}
|
|
} else {
|
|
// Look up the code point in its 4k block of code points.
|
|
if(!containsSlow(c, list4kStarts[lead], list4kStarts[lead+1])) {
|
|
break;
|
|
}
|
|
}
|
|
} else if(c<0xdc00 || s==limit || (c2=*(limit-1))<0xd800 || c2>=0xdc00) {
|
|
// surrogate code point
|
|
if(!containsSlow(c, list4kStarts[0xd], list4kStarts[0xe])) {
|
|
break;
|
|
}
|
|
} else {
|
|
// surrogate pair
|
|
if(!containsSlow(U16_GET_SUPPLEMENTARY(c2, c), list4kStarts[0x10], list4kStarts[0x11])) {
|
|
break;
|
|
}
|
|
--limit;
|
|
}
|
|
if(s==limit) {
|
|
return s;
|
|
}
|
|
}
|
|
} else {
|
|
// span not
|
|
for(;;) {
|
|
c=*(--limit);
|
|
if(c<=0xff) {
|
|
if(latin1Contains[c]) {
|
|
break;
|
|
}
|
|
} else if(c<=0x7ff) {
|
|
if((table7FF[c&0x3f]&((uint32_t)1<<(c>>6)))!=0) {
|
|
break;
|
|
}
|
|
} else if(c<0xd800 || c>=0xe000) {
|
|
int lead=c>>12;
|
|
uint32_t twoBits=(bmpBlockBits[(c>>6)&0x3f]>>lead)&0x10001;
|
|
if(twoBits<=1) {
|
|
// All 64 code points with the same bits 15..6
|
|
// are either in the set or not.
|
|
if(twoBits!=0) {
|
|
break;
|
|
}
|
|
} else {
|
|
// Look up the code point in its 4k block of code points.
|
|
if(containsSlow(c, list4kStarts[lead], list4kStarts[lead+1])) {
|
|
break;
|
|
}
|
|
}
|
|
} else if(c<0xdc00 || s==limit || (c2=*(limit-1))<0xd800 || c2>=0xdc00) {
|
|
// surrogate code point
|
|
if(containsSlow(c, list4kStarts[0xd], list4kStarts[0xe])) {
|
|
break;
|
|
}
|
|
} else {
|
|
// surrogate pair
|
|
if(containsSlow(U16_GET_SUPPLEMENTARY(c2, c), list4kStarts[0x10], list4kStarts[0x11])) {
|
|
break;
|
|
}
|
|
--limit;
|
|
}
|
|
if(s==limit) {
|
|
return s;
|
|
}
|
|
}
|
|
}
|
|
return limit+1;
|
|
}
|
|
|
|
/*
|
|
* Precheck for sufficient trail bytes at end of string only once per span.
|
|
* Check validity.
|
|
*/
|
|
const uint8_t *
|
|
BMPSet::spanUTF8(const uint8_t *s, int32_t length, USetSpanCondition spanCondition) const {
|
|
const uint8_t *limit=s+length;
|
|
uint8_t b=*s;
|
|
if(U8_IS_SINGLE(b)) {
|
|
// Initial all-ASCII span.
|
|
if(spanCondition) {
|
|
do {
|
|
if(!latin1Contains[b] || ++s==limit) {
|
|
return s;
|
|
}
|
|
b=*s;
|
|
} while(U8_IS_SINGLE(b));
|
|
} else {
|
|
do {
|
|
if(latin1Contains[b] || ++s==limit) {
|
|
return s;
|
|
}
|
|
b=*s;
|
|
} while(U8_IS_SINGLE(b));
|
|
}
|
|
length=(int32_t)(limit-s);
|
|
}
|
|
|
|
if(spanCondition!=USET_SPAN_NOT_CONTAINED) {
|
|
spanCondition=USET_SPAN_CONTAINED; // Pin to 0/1 values.
|
|
}
|
|
|
|
const uint8_t *limit0=limit;
|
|
|
|
/*
|
|
* Make sure that the last 1/2/3/4-byte sequence before limit is complete
|
|
* or runs into a lead byte.
|
|
* In the span loop compare s with limit only once
|
|
* per multi-byte character.
|
|
*
|
|
* Give a trailing illegal sequence the same value as the result of contains(FFFD),
|
|
* including it if that is part of the span, otherwise set limit0 to before
|
|
* the truncated sequence.
|
|
*/
|
|
b=*(limit-1);
|
|
if((int8_t)b<0) {
|
|
// b>=0x80: lead or trail byte
|
|
if(b<0xc0) {
|
|
// single trail byte, check for preceding 3- or 4-byte lead byte
|
|
if(length>=2 && (b=*(limit-2))>=0xe0) {
|
|
limit-=2;
|
|
if(containsFFFD!=spanCondition) {
|
|
limit0=limit;
|
|
}
|
|
} else if(b<0xc0 && b>=0x80 && length>=3 && (b=*(limit-3))>=0xf0) {
|
|
// 4-byte lead byte with only two trail bytes
|
|
limit-=3;
|
|
if(containsFFFD!=spanCondition) {
|
|
limit0=limit;
|
|
}
|
|
}
|
|
} else {
|
|
// lead byte with no trail bytes
|
|
--limit;
|
|
if(containsFFFD!=spanCondition) {
|
|
limit0=limit;
|
|
}
|
|
}
|
|
}
|
|
|
|
uint8_t t1, t2, t3;
|
|
|
|
while(s<limit) {
|
|
b=*s;
|
|
if(U8_IS_SINGLE(b)) {
|
|
// ASCII
|
|
if(spanCondition) {
|
|
do {
|
|
if(!latin1Contains[b]) {
|
|
return s;
|
|
} else if(++s==limit) {
|
|
return limit0;
|
|
}
|
|
b=*s;
|
|
} while(U8_IS_SINGLE(b));
|
|
} else {
|
|
do {
|
|
if(latin1Contains[b]) {
|
|
return s;
|
|
} else if(++s==limit) {
|
|
return limit0;
|
|
}
|
|
b=*s;
|
|
} while(U8_IS_SINGLE(b));
|
|
}
|
|
}
|
|
++s; // Advance past the lead byte.
|
|
if(b>=0xe0) {
|
|
if(b<0xf0) {
|
|
if( /* handle U+0000..U+FFFF inline */
|
|
(t1=(uint8_t)(s[0]-0x80)) <= 0x3f &&
|
|
(t2=(uint8_t)(s[1]-0x80)) <= 0x3f
|
|
) {
|
|
b&=0xf;
|
|
uint32_t twoBits=(bmpBlockBits[t1]>>b)&0x10001;
|
|
if(twoBits<=1) {
|
|
// All 64 code points with this lead byte and middle trail byte
|
|
// are either in the set or not.
|
|
if(twoBits!=(uint32_t)spanCondition) {
|
|
return s-1;
|
|
}
|
|
} else {
|
|
// Look up the code point in its 4k block of code points.
|
|
UChar32 c=(b<<12)|(t1<<6)|t2;
|
|
if(containsSlow(c, list4kStarts[b], list4kStarts[b+1]) != spanCondition) {
|
|
return s-1;
|
|
}
|
|
}
|
|
s+=2;
|
|
continue;
|
|
}
|
|
} else if( /* handle U+10000..U+10FFFF inline */
|
|
(t1=(uint8_t)(s[0]-0x80)) <= 0x3f &&
|
|
(t2=(uint8_t)(s[1]-0x80)) <= 0x3f &&
|
|
(t3=(uint8_t)(s[2]-0x80)) <= 0x3f
|
|
) {
|
|
// Give an illegal sequence the same value as the result of contains(FFFD).
|
|
UChar32 c=((UChar32)(b-0xf0)<<18)|((UChar32)t1<<12)|(t2<<6)|t3;
|
|
if( ( (0x10000<=c && c<=0x10ffff) ?
|
|
containsSlow(c, list4kStarts[0x10], list4kStarts[0x11]) :
|
|
containsFFFD
|
|
) != spanCondition
|
|
) {
|
|
return s-1;
|
|
}
|
|
s+=3;
|
|
continue;
|
|
}
|
|
} else {
|
|
if( /* handle U+0000..U+07FF inline */
|
|
b>=0xc0 &&
|
|
(t1=(uint8_t)(*s-0x80)) <= 0x3f
|
|
) {
|
|
if((USetSpanCondition)((table7FF[t1]&((uint32_t)1<<(b&0x1f)))!=0) != spanCondition) {
|
|
return s-1;
|
|
}
|
|
++s;
|
|
continue;
|
|
}
|
|
}
|
|
|
|
// Give an illegal sequence the same value as the result of contains(FFFD).
|
|
// Handle each byte of an illegal sequence separately to simplify the code;
|
|
// no need to optimize error handling.
|
|
if(containsFFFD!=spanCondition) {
|
|
return s-1;
|
|
}
|
|
}
|
|
|
|
return limit0;
|
|
}
|
|
|
|
/*
|
|
* While going backwards through UTF-8 optimize only for ASCII.
|
|
* Unlike UTF-16, UTF-8 is not forward-backward symmetrical, that is, it is not
|
|
* possible to tell from the last byte in a multi-byte sequence how many
|
|
* preceding bytes there should be. Therefore, going backwards through UTF-8
|
|
* is much harder than going forward.
|
|
*/
|
|
int32_t
|
|
BMPSet::spanBackUTF8(const uint8_t *s, int32_t length, USetSpanCondition spanCondition) const {
|
|
if(spanCondition!=USET_SPAN_NOT_CONTAINED) {
|
|
spanCondition=USET_SPAN_CONTAINED; // Pin to 0/1 values.
|
|
}
|
|
|
|
uint8_t b;
|
|
|
|
do {
|
|
b=s[--length];
|
|
if(U8_IS_SINGLE(b)) {
|
|
// ASCII sub-span
|
|
if(spanCondition) {
|
|
do {
|
|
if(!latin1Contains[b]) {
|
|
return length+1;
|
|
} else if(length==0) {
|
|
return 0;
|
|
}
|
|
b=s[--length];
|
|
} while(U8_IS_SINGLE(b));
|
|
} else {
|
|
do {
|
|
if(latin1Contains[b]) {
|
|
return length+1;
|
|
} else if(length==0) {
|
|
return 0;
|
|
}
|
|
b=s[--length];
|
|
} while(U8_IS_SINGLE(b));
|
|
}
|
|
}
|
|
|
|
int32_t prev=length;
|
|
UChar32 c;
|
|
// trail byte: collect a multi-byte character
|
|
// (or lead byte in last-trail position)
|
|
c=utf8_prevCharSafeBody(s, 0, &length, b, -3);
|
|
// c is a valid code point, not ASCII, not a surrogate
|
|
if(c<=0x7ff) {
|
|
if((USetSpanCondition)((table7FF[c&0x3f]&((uint32_t)1<<(c>>6)))!=0) != spanCondition) {
|
|
return prev+1;
|
|
}
|
|
} else if(c<=0xffff) {
|
|
int lead=c>>12;
|
|
uint32_t twoBits=(bmpBlockBits[(c>>6)&0x3f]>>lead)&0x10001;
|
|
if(twoBits<=1) {
|
|
// All 64 code points with the same bits 15..6
|
|
// are either in the set or not.
|
|
if(twoBits!=(uint32_t)spanCondition) {
|
|
return prev+1;
|
|
}
|
|
} else {
|
|
// Look up the code point in its 4k block of code points.
|
|
if(containsSlow(c, list4kStarts[lead], list4kStarts[lead+1]) != spanCondition) {
|
|
return prev+1;
|
|
}
|
|
}
|
|
} else {
|
|
if(containsSlow(c, list4kStarts[0x10], list4kStarts[0x11]) != spanCondition) {
|
|
return prev+1;
|
|
}
|
|
}
|
|
} while(length>0);
|
|
return 0;
|
|
}
|
|
|
|
U_NAMESPACE_END
|