+// © 2016 and later: Unicode, Inc. and others.
+// License & terms of use: http://www.unicode.org/copyright.html
/*
*******************************************************************************
*
-* Copyright (C) 1999-2007, International Business Machines
+* Copyright (C) 1999-2015, International Business Machines
* Corporation and others. All Rights Reserved.
*
*******************************************************************************
* file name: utf8.h
-* encoding: US-ASCII
+* encoding: UTF-8
* tab size: 8 (not used)
* indentation:4
*
* \brief C API: 8-bit Unicode handling macros
*
* This file defines macros to deal with 8-bit Unicode (UTF-8) code units (bytes) and strings.
- * utf8.h is included by utf.h after unicode/umachine.h
- * and some common definitions.
*
* For more information see utf.h and the ICU User Guide Strings chapter
- * (http://icu-project.org/userguide/strings.html).
+ * (http://userguide.icu-project.org/strings).
*
* <em>Usage:</em>
* ICU coding guidelines for if() statements should be followed when using these macros.
#ifndef __UTF8_H__
#define __UTF8_H__
-/* utf.h must be included first. */
+#include "unicode/umachine.h"
#ifndef __UTF_H__
# include "unicode/utf.h"
#endif
/* internal definitions ----------------------------------------------------- */
+
+
/**
- * \var utf8_countTrailBytes
- * Internal array with numbers of trail bytes for any given byte used in
- * lead byte position.
+ * Counts the trail bytes for a UTF-8 lead byte.
+ * Returns 0 for 0..0xbf as well as for 0xfe and 0xff.
+ *
+ * This is internal since it is not meant to be called directly by external clients;
+ * however it is called by public macros in this file and thus must remain stable.
+ *
+ * Note: Beginning with ICU 50, the implementation uses a multi-condition expression
+ * which was shown in 2012 (on x86-64) to compile to fast, branch-free code.
+ * leadByte is evaluated multiple times.
+ *
+ * The pre-ICU 50 implementation used the exported array utf8_countTrailBytes:
+ * #define U8_COUNT_TRAIL_BYTES(leadByte) (utf8_countTrailBytes[leadByte])
+ * leadByte was evaluated exactly once.
+ *
+ * @param leadByte The first byte of a UTF-8 sequence. Must be 0..0xff.
* @internal
*/
-#ifdef U_UTF8_IMPL
-U_EXPORT const uint8_t
-#elif defined(U_STATIC_IMPLEMENTATION) || defined(U_COMMON_IMPLEMENTATION)
-U_CFUNC const uint8_t
-#else
-U_CFUNC U_IMPORT const uint8_t /* U_IMPORT2? */ /*U_IMPORT*/
-#endif
-utf8_countTrailBytes[256];
+#define U8_COUNT_TRAIL_BYTES(leadByte) \
+ ((uint8_t)(leadByte)<0xf0 ? \
+ ((uint8_t)(leadByte)>=0xc0)+((uint8_t)(leadByte)>=0xe0) : \
+ (uint8_t)(leadByte)<0xfe ? 3+((uint8_t)(leadByte)>=0xf8)+((uint8_t)(leadByte)>=0xfc) : 0)
/**
- * Count the trail bytes for a UTF-8 lead byte.
+ * Counts the trail bytes for a UTF-8 lead byte of a valid UTF-8 sequence.
+ * The maximum supported lead byte is 0xf4 corresponding to U+10FFFF.
+ * leadByte might be evaluated multiple times.
+ *
+ * This is internal since it is not meant to be called directly by external clients;
+ * however it is called by public macros in this file and thus must remain stable.
+ *
+ * @param leadByte The first byte of a UTF-8 sequence. Must be 0..0xff.
* @internal
*/
-#define U8_COUNT_TRAIL_BYTES(leadByte) (utf8_countTrailBytes[(uint8_t)leadByte])
+#define U8_COUNT_TRAIL_BYTES_UNSAFE(leadByte) \
+ (((leadByte)>=0xc0)+((leadByte)>=0xe0)+((leadByte)>=0xf0))
/**
* Mask a UTF-8 lead byte, leave only the lower bits that form part of the code point value.
+ *
+ * This is internal since it is not meant to be called directly by external clients;
+ * however it is called by public macros in this file and thus must remain stable.
* @internal
*/
#define U8_MASK_LEAD_BYTE(leadByte, countTrailBytes) ((leadByte)&=(1<<(6-(countTrailBytes)))-1)
/**
* Function for handling "next code point" with error-checking.
+ *
+ * This is internal since it is not meant to be called directly by external clients;
+ * however it is U_STABLE (not U_INTERNAL) since it is called by public macros in this
+ * file and thus must remain stable, and should not be hidden when other internal
+ * functions are hidden (otherwise public macros would fail to compile).
* @internal
*/
-U_INTERNAL UChar32 U_EXPORT2
+U_STABLE UChar32 U_EXPORT2
utf8_nextCharSafeBody(const uint8_t *s, int32_t *pi, int32_t length, UChar32 c, UBool strict);
/**
* Function for handling "append code point" with error-checking.
+ *
+ * This is internal since it is not meant to be called directly by external clients;
+ * however it is U_STABLE (not U_INTERNAL) since it is called by public macros in this
+ * file and thus must remain stable, and should not be hidden when other internal
+ * functions are hidden (otherwise public macros would fail to compile).
* @internal
*/
-U_INTERNAL int32_t U_EXPORT2
+U_STABLE int32_t U_EXPORT2
utf8_appendCharSafeBody(uint8_t *s, int32_t i, int32_t length, UChar32 c, UBool *pIsError);
/**
* Function for handling "previous code point" with error-checking.
+ *
+ * This is internal since it is not meant to be called directly by external clients;
+ * however it is U_STABLE (not U_INTERNAL) since it is called by public macros in this
+ * file and thus must remain stable, and should not be hidden when other internal
+ * functions are hidden (otherwise public macros would fail to compile).
* @internal
*/
-U_INTERNAL UChar32 U_EXPORT2
+U_STABLE UChar32 U_EXPORT2
utf8_prevCharSafeBody(const uint8_t *s, int32_t start, int32_t *pi, UChar32 c, UBool strict);
/**
* Function for handling "skip backward one code point" with error-checking.
+ *
+ * This is internal since it is not meant to be called directly by external clients;
+ * however it is U_STABLE (not U_INTERNAL) since it is called by public macros in this
+ * file and thus must remain stable, and should not be hidden when other internal
+ * functions are hidden (otherwise public macros would fail to compile).
* @internal
*/
-U_INTERNAL int32_t U_EXPORT2
+U_STABLE int32_t U_EXPORT2
utf8_back1SafeBody(const uint8_t *s, int32_t start, int32_t i);
/* single-code point definitions -------------------------------------------- */
* The offset may point to either the lead byte or one of the trail bytes
* for a code point, in which case the macro will read all of the bytes
* for the code point.
+ *
+ * The length can be negative for a NUL-terminated string.
+ *
* If the offset points to an illegal UTF-8 byte sequence, then
* c is set to a negative value.
* Iteration through a string is more efficient with U8_NEXT_UNSAFE or U8_NEXT.
*
* @param s const uint8_t * string
- * @param start starting string offset
- * @param i string offset, must be start<=i<length
- * @param length string length
+ * @param start int32_t starting string offset
+ * @param i int32_t string offset, must be start<=i<length
+ * @param length int32_t string length
* @param c output UChar32 variable, set to <0 in case of an error
* @see U8_GET_UNSAFE
* @stable ICU 2.4
*/
#define U8_GET(s, start, i, length, c) { \
- int32_t _u8_get_index=(int32_t)(i); \
+ int32_t _u8_get_index=(i); \
U8_SET_CP_START(s, start, _u8_get_index); \
U8_NEXT(s, _u8_get_index, length, c); \
}
+/**
+ * Get a code point from a string at a random-access offset,
+ * without changing the offset.
+ * The offset may point to either the lead byte or one of the trail bytes
+ * for a code point, in which case the macro will read all of the bytes
+ * for the code point.
+ *
+ * The length can be negative for a NUL-terminated string.
+ *
+ * If the offset points to an illegal UTF-8 byte sequence, then
+ * c is set to U+FFFD.
+ * Iteration through a string is more efficient with U8_NEXT_UNSAFE or U8_NEXT_OR_FFFD.
+ *
+ * This macro does not distinguish between a real U+FFFD in the text
+ * and U+FFFD returned for an ill-formed sequence.
+ * Use U8_GET() if that distinction is important.
+ *
+ * @param s const uint8_t * string
+ * @param start int32_t starting string offset
+ * @param i int32_t string offset, must be start<=i<length
+ * @param length int32_t string length
+ * @param c output UChar32 variable, set to U+FFFD in case of an error
+ * @see U8_GET
+ * @stable ICU 51
+ */
+#define U8_GET_OR_FFFD(s, start, i, length, c) { \
+ int32_t _u8_get_index=(i); \
+ U8_SET_CP_START(s, start, _u8_get_index); \
+ U8_NEXT_OR_FFFD(s, _u8_get_index, length, c); \
+}
+
/* definitions with forward iteration --------------------------------------- */
/**
*/
#define U8_NEXT_UNSAFE(s, i, c) { \
(c)=(uint8_t)(s)[(i)++]; \
- if((uint8_t)((c)-0xc0)<0x35) { \
- uint8_t __count=U8_COUNT_TRAIL_BYTES(c); \
- U8_MASK_LEAD_BYTE(c, __count); \
- switch(__count) { \
- /* each following branch falls through to the next one */ \
- case 3: \
- (c)=((c)<<6)|((s)[(i)++]&0x3f); \
- case 2: \
- (c)=((c)<<6)|((s)[(i)++]&0x3f); \
- case 1: \
- (c)=((c)<<6)|((s)[(i)++]&0x3f); \
- /* no other branches to optimize switch() */ \
- break; \
+ if((c)>=0x80) { \
+ if((c)<0xe0) { \
+ (c)=(((c)&0x1f)<<6)|((s)[(i)++]&0x3f); \
+ } else if((c)<0xf0) { \
+ /* no need for (c&0xf) because the upper bits are truncated after <<12 in the cast to (UChar) */ \
+ (c)=(UChar)(((c)<<12)|(((s)[i]&0x3f)<<6)|((s)[(i)+1]&0x3f)); \
+ (i)+=2; \
+ } else { \
+ (c)=(((c)&7)<<18)|(((s)[i]&0x3f)<<12)|(((s)[(i)+1]&0x3f)<<6)|((s)[(i)+2]&0x3f); \
+ (i)+=3; \
} \
} \
}
* (Post-incrementing forward iteration.)
* "Safe" macro, checks for illegal sequences and for string boundaries.
*
+ * The length can be negative for a NUL-terminated string.
+ *
* The offset may point to the lead byte of a multi-byte sequence,
* in which case the macro will read the whole sequence.
* If the offset points to a trail byte or an illegal UTF-8 sequence, then
* c is set to a negative value.
*
* @param s const uint8_t * string
- * @param i string offset, must be i<length
- * @param length string length
+ * @param i int32_t string offset, must be i<length
+ * @param length int32_t string length
* @param c output UChar32 variable, set to <0 in case of an error
* @see U8_NEXT_UNSAFE
* @stable ICU 2.4
uint8_t __t1, __t2; \
if( /* handle U+1000..U+CFFF inline */ \
(0xe0<(c) && (c)<=0xec) && \
- (((i)+1)<(length)) && \
+ (((i)+1)<(length) || (length)<0) && \
(__t1=(uint8_t)((s)[i]-0x80))<=0x3f && \
(__t2=(uint8_t)((s)[(i)+1]-0x80))<= 0x3f \
) { \
(i)+=2; \
} else if( /* handle U+0080..U+07FF inline */ \
((c)<0xe0 && (c)>=0xc2) && \
- ((i)<(length)) && \
+ ((i)!=(length)) && \
(__t1=(uint8_t)((s)[i]-0x80))<=0x3f \
) { \
- (c)=(UChar)((((c)&0x1f)<<6)|__t1); \
+ (c)=(((c)&0x1f)<<6)|__t1; \
++(i); \
- } else if(U8_IS_LEAD(c)) { \
+ } else { \
/* function call for "complicated" and error cases */ \
- (c)=utf8_nextCharSafeBody((const uint8_t *)s, &(i), (int32_t)(length), c, -1); \
+ (c)=utf8_nextCharSafeBody((const uint8_t *)s, &(i), (length), c, -1); \
+ } \
+ } \
+}
+
+/**
+ * Get a code point from a string at a code point boundary offset,
+ * and advance the offset to the next code point boundary.
+ * (Post-incrementing forward iteration.)
+ * "Safe" macro, checks for illegal sequences and for string boundaries.
+ *
+ * The length can be negative for a NUL-terminated string.
+ *
+ * The offset may point to the lead byte of a multi-byte sequence,
+ * in which case the macro will read the whole sequence.
+ * If the offset points to a trail byte or an illegal UTF-8 sequence, then
+ * c is set to U+FFFD.
+ *
+ * This macro does not distinguish between a real U+FFFD in the text
+ * and U+FFFD returned for an ill-formed sequence.
+ * Use U8_NEXT() if that distinction is important.
+ *
+ * @param s const uint8_t * string
+ * @param i int32_t string offset, must be i<length
+ * @param length int32_t string length
+ * @param c output UChar32 variable, set to U+FFFD in case of an error
+ * @see U8_NEXT
+ * @stable ICU 51
+ */
+#define U8_NEXT_OR_FFFD(s, i, length, c) { \
+ (c)=(uint8_t)(s)[(i)++]; \
+ if((c)>=0x80) { \
+ uint8_t __t1, __t2; \
+ if( /* handle U+1000..U+CFFF inline */ \
+ (0xe0<(c) && (c)<=0xec) && \
+ (((i)+1)<(length) || (length)<0) && \
+ (__t1=(uint8_t)((s)[i]-0x80))<=0x3f && \
+ (__t2=(uint8_t)((s)[(i)+1]-0x80))<= 0x3f \
+ ) { \
+ /* no need for (c&0xf) because the upper bits are truncated after <<12 in the cast to (UChar) */ \
+ (c)=(UChar)(((c)<<12)|(__t1<<6)|__t2); \
+ (i)+=2; \
+ } else if( /* handle U+0080..U+07FF inline */ \
+ ((c)<0xe0 && (c)>=0xc2) && \
+ ((i)!=(length)) && \
+ (__t1=(uint8_t)((s)[i]-0x80))<=0x3f \
+ ) { \
+ (c)=(((c)&0x1f)<<6)|__t1; \
+ ++(i); \
} else { \
- (c)=U_SENTINEL; \
+ /* function call for "complicated" and error cases */ \
+ (c)=utf8_nextCharSafeBody((const uint8_t *)s, &(i), (length), c, -3); \
} \
} \
}
* then isError is set to TRUE.
*
* @param s const uint8_t * string buffer
- * @param i string offset, must be i<capacity
- * @param capacity size of the string buffer
- * @param c code point to append
+ * @param i int32_t string offset, must be i<capacity
+ * @param capacity int32_t size of the string buffer
+ * @param c UChar32 code point to append
* @param isError output UBool set to TRUE if an error occurs, otherwise not modified
* @see U8_APPEND_UNSAFE
* @stable ICU 2.4
(s)[(i)++]=(uint8_t)((((c)>>6)&0x3f)|0x80); \
(s)[(i)++]=(uint8_t)(((c)&0x3f)|0x80); \
} else { \
- (i)=utf8_appendCharSafeBody(s, (int32_t)(i), (int32_t)(capacity), c, &(isError)); \
+ (i)=utf8_appendCharSafeBody(s, (i), (capacity), c, &(isError)); \
} \
}
* @stable ICU 2.4
*/
#define U8_FWD_1_UNSAFE(s, i) { \
- (i)+=1+U8_COUNT_TRAIL_BYTES((s)[i]); \
+ (i)+=1+U8_COUNT_TRAIL_BYTES_UNSAFE((uint8_t)(s)[i]); \
}
/**
* (Post-incrementing iteration.)
* "Safe" macro, checks for illegal sequences and for string boundaries.
*
+ * The length can be negative for a NUL-terminated string.
+ *
* @param s const uint8_t * string
- * @param i string offset, must be i<length
- * @param length string length
+ * @param i int32_t string offset, must be i<length
+ * @param length int32_t string length
* @see U8_FWD_1_UNSAFE
* @stable ICU 2.4
*/
uint8_t __b=(uint8_t)(s)[(i)++]; \
if(U8_IS_LEAD(__b)) { \
uint8_t __count=U8_COUNT_TRAIL_BYTES(__b); \
- if((i)+__count>(length)) { \
+ if((i)+__count>(length) && (length)>=0) { \
__count=(uint8_t)((length)-(i)); \
} \
while(__count>0 && U8_IS_TRAIL((s)[i])) { \
* (Post-incrementing iteration.)
* "Safe" macro, checks for illegal sequences and for string boundaries.
*
+ * The length can be negative for a NUL-terminated string.
+ *
* @param s const uint8_t * string
- * @param i string offset, must be i<length
- * @param length string length
+ * @param i int32_t string offset, must be i<length
+ * @param length int32_t string length
* @param n number of code points to skip
* @see U8_FWD_N_UNSAFE
* @stable ICU 2.4
*/
#define U8_FWD_N(s, i, length, n) { \
int32_t __N=(n); \
- while(__N>0 && (i)<(length)) { \
+ while(__N>0 && ((i)<(length) || ((length)<0 && (s)[i]!=0))) { \
U8_FWD_1(s, i, length); \
--__N; \
} \
* "Safe" macro, checks for illegal sequences and for string boundaries.
*
* @param s const uint8_t * string
- * @param start starting string offset (usually 0)
- * @param i string offset, must be start<=i
+ * @param start int32_t starting string offset (usually 0)
+ * @param i int32_t string offset, must be start<=i
* @see U8_SET_CP_START_UNSAFE
* @stable ICU 2.4
*/
#define U8_SET_CP_START(s, start, i) { \
if(U8_IS_TRAIL((s)[(i)])) { \
- (i)=utf8_back1SafeBody(s, start, (int32_t)(i)); \
+ (i)=utf8_back1SafeBody(s, start, (i)); \
} \
}
* If the offset is behind an illegal UTF-8 sequence, then c is set to a negative value.
*
* @param s const uint8_t * string
- * @param start starting string offset (usually 0)
- * @param i string offset, must be start<i
+ * @param start int32_t starting string offset (usually 0)
+ * @param i int32_t string offset, must be start<i
* @param c output UChar32 variable, set to <0 in case of an error
* @see U8_PREV_UNSAFE
* @stable ICU 2.4
#define U8_PREV(s, start, i, c) { \
(c)=(uint8_t)(s)[--(i)]; \
if((c)>=0x80) { \
- if((c)<=0xbf) { \
- (c)=utf8_prevCharSafeBody((const uint8_t *)s, start, &(i), c, -1); \
- } else { \
- (c)=U_SENTINEL; \
- } \
+ (c)=utf8_prevCharSafeBody((const uint8_t *)s, start, &(i), c, -1); \
+ } \
+}
+
+/**
+ * Move the string offset from one code point boundary to the previous one
+ * and get the code point between them.
+ * (Pre-decrementing backward iteration.)
+ * "Safe" macro, checks for illegal sequences and for string boundaries.
+ *
+ * The input offset may be the same as the string length.
+ * If the offset is behind a multi-byte sequence, then the macro will read
+ * the whole sequence.
+ * If the offset is behind a lead byte, then that itself
+ * will be returned as the code point.
+ * If the offset is behind an illegal UTF-8 sequence, then c is set to U+FFFD.
+ *
+ * This macro does not distinguish between a real U+FFFD in the text
+ * and U+FFFD returned for an ill-formed sequence.
+ * Use U8_PREV() if that distinction is important.
+ *
+ * @param s const uint8_t * string
+ * @param start int32_t starting string offset (usually 0)
+ * @param i int32_t string offset, must be start<i
+ * @param c output UChar32 variable, set to U+FFFD in case of an error
+ * @see U8_PREV
+ * @stable ICU 51
+ */
+#define U8_PREV_OR_FFFD(s, start, i, c) { \
+ (c)=(uint8_t)(s)[--(i)]; \
+ if((c)>=0x80) { \
+ (c)=utf8_prevCharSafeBody((const uint8_t *)s, start, &(i), c, -3); \
} \
}
* "Safe" macro, checks for illegal sequences and for string boundaries.
*
* @param s const uint8_t * string
- * @param start starting string offset (usually 0)
- * @param i string offset, must be start<i
+ * @param start int32_t starting string offset (usually 0)
+ * @param i int32_t string offset, must be start<i
* @see U8_BACK_1_UNSAFE
* @stable ICU 2.4
*/
#define U8_BACK_1(s, start, i) { \
if(U8_IS_TRAIL((s)[--(i)])) { \
- (i)=utf8_back1SafeBody(s, start, (int32_t)(i)); \
+ (i)=utf8_back1SafeBody(s, start, (i)); \
} \
}
* "Safe" macro, checks for illegal sequences and for string boundaries.
*
* @param s const uint8_t * string
- * @param start index of the start of the string
- * @param i string offset, must be start<i
+ * @param start int32_t index of the start of the string
+ * @param i int32_t string offset, must be start<i
* @param n number of code points to skip
* @see U8_BACK_N_UNSAFE
* @stable ICU 2.4
* The input offset may be the same as the string length.
* "Safe" macro, checks for illegal sequences and for string boundaries.
*
+ * The length can be negative for a NUL-terminated string.
+ *
* @param s const uint8_t * string
- * @param start starting string offset (usually 0)
- * @param i string offset, must be start<=i<=length
- * @param length string length
+ * @param start int32_t starting string offset (usually 0)
+ * @param i int32_t string offset, must be start<=i<=length
+ * @param length int32_t string length
* @see U8_SET_CP_LIMIT_UNSAFE
* @stable ICU 2.4
*/
#define U8_SET_CP_LIMIT(s, start, i, length) { \
- if((start)<(i) && (i)<(length)) { \
+ if((start)<(i) && ((i)<(length) || (length)<0)) { \
U8_BACK_1(s, start, i); \
U8_FWD_1(s, i, length); \
} \
projects / apple / icu.git / blobdiff