michael@0: /*
michael@0: *******************************************************************************
michael@0: *
michael@0: *   Copyright (C) 1999-2013, International Business Machines
michael@0: *   Corporation and others.  All Rights Reserved.
michael@0: *
michael@0: *******************************************************************************
michael@0: *   file name:  utf8.h
michael@0: *   encoding:   US-ASCII
michael@0: *   tab size:   8 (not used)
michael@0: *   indentation:4
michael@0: *
michael@0: *   created on: 1999sep13
michael@0: *   created by: Markus W. Scherer
michael@0: */
michael@0: 
michael@0: /**
michael@0:  * \file
michael@0:  * \brief C API: 8-bit Unicode handling macros
michael@0:  * 
michael@0:  * This file defines macros to deal with 8-bit Unicode (UTF-8) code units (bytes) and strings.
michael@0:  *
michael@0:  * For more information see utf.h and the ICU User Guide Strings chapter
michael@0:  * (http://userguide.icu-project.org/strings).
michael@0:  *
michael@0:  * <em>Usage:</em>
michael@0:  * ICU coding guidelines for if() statements should be followed when using these macros.
michael@0:  * Compound statements (curly braces {}) must be used  for if-else-while... 
michael@0:  * bodies and all macro statements should be terminated with semicolon.
michael@0:  */
michael@0: 
michael@0: #ifndef __UTF8_H__
michael@0: #define __UTF8_H__
michael@0: 
michael@0: #include "unicode/umachine.h"
michael@0: #ifndef __UTF_H__
michael@0: #   include "unicode/utf.h"
michael@0: #endif
michael@0: 
michael@0: /* internal definitions ----------------------------------------------------- */
michael@0: 
michael@0: /**
michael@0:  * \var utf8_countTrailBytes
michael@0:  * Internal array with numbers of trail bytes for any given byte used in
michael@0:  * lead byte position.
michael@0:  *
michael@0:  * This is internal since it is not meant to be called directly by external clients;
michael@0:  * however it is called by public macros in this file and thus must remain stable,
michael@0:  * and should not be hidden when other internal functions are hidden (otherwise
michael@0:  * public macros would fail to compile).
michael@0:  * @internal
michael@0:  */
michael@0: #ifdef U_UTF8_IMPL
michael@0: U_EXPORT const uint8_t 
michael@0: #elif defined(U_STATIC_IMPLEMENTATION) || defined(U_COMMON_IMPLEMENTATION)
michael@0: U_CFUNC const uint8_t
michael@0: #else
michael@0: U_CFUNC U_IMPORT const uint8_t /* U_IMPORT2? */ /*U_IMPORT*/ 
michael@0: #endif
michael@0: utf8_countTrailBytes[256];
michael@0: 
michael@0: /**
michael@0:  * Counts the trail bytes for a UTF-8 lead byte.
michael@0:  * Returns 0 for 0..0xbf as well as for 0xfe and 0xff.
michael@0:  *
michael@0:  * This is internal since it is not meant to be called directly by external clients;
michael@0:  * however it is called by public macros in this file and thus must remain stable.
michael@0:  *
michael@0:  * Note: Beginning with ICU 50, the implementation uses a multi-condition expression
michael@0:  * which was shown in 2012 (on x86-64) to compile to fast, branch-free code.
michael@0:  * leadByte is evaluated multiple times.
michael@0:  *
michael@0:  * The pre-ICU 50 implementation used the exported array utf8_countTrailBytes:
michael@0:  * #define U8_COUNT_TRAIL_BYTES(leadByte) (utf8_countTrailBytes[leadByte])
michael@0:  * leadByte was evaluated exactly once.
michael@0:  *
michael@0:  * @param leadByte The first byte of a UTF-8 sequence. Must be 0..0xff.
michael@0:  * @internal
michael@0:  */
michael@0: #define U8_COUNT_TRAIL_BYTES(leadByte) \
michael@0:     ((leadByte)<0xf0 ? \
michael@0:         ((leadByte)>=0xc0)+((leadByte)>=0xe0) : \
michael@0:         (leadByte)<0xfe ? 3+((leadByte)>=0xf8)+((leadByte)>=0xfc) : 0)
michael@0: 
michael@0: /**
michael@0:  * Counts the trail bytes for a UTF-8 lead byte of a valid UTF-8 sequence.
michael@0:  * The maximum supported lead byte is 0xf4 corresponding to U+10FFFF.
michael@0:  * leadByte might be evaluated multiple times.
michael@0:  *
michael@0:  * This is internal since it is not meant to be called directly by external clients;
michael@0:  * however it is called by public macros in this file and thus must remain stable.
michael@0:  *
michael@0:  * @param leadByte The first byte of a UTF-8 sequence. Must be 0..0xff.
michael@0:  * @internal
michael@0:  */
michael@0: #define U8_COUNT_TRAIL_BYTES_UNSAFE(leadByte) \
michael@0:     (((leadByte)>=0xc0)+((leadByte)>=0xe0)+((leadByte)>=0xf0))
michael@0: 
michael@0: /**
michael@0:  * Mask a UTF-8 lead byte, leave only the lower bits that form part of the code point value.
michael@0:  *
michael@0:  * This is internal since it is not meant to be called directly by external clients;
michael@0:  * however it is called by public macros in this file and thus must remain stable.
michael@0:  * @internal
michael@0:  */
michael@0: #define U8_MASK_LEAD_BYTE(leadByte, countTrailBytes) ((leadByte)&=(1<<(6-(countTrailBytes)))-1)
michael@0: 
michael@0: /**
michael@0:  * Function for handling "next code point" with error-checking.
michael@0:  *
michael@0:  * This is internal since it is not meant to be called directly by external clients;
michael@0:  * however it is U_STABLE (not U_INTERNAL) since it is called by public macros in this
michael@0:  * file and thus must remain stable, and should not be hidden when other internal
michael@0:  * functions are hidden (otherwise public macros would fail to compile).
michael@0:  * @internal
michael@0:  */
michael@0: U_STABLE UChar32 U_EXPORT2
michael@0: utf8_nextCharSafeBody(const uint8_t *s, int32_t *pi, int32_t length, UChar32 c, UBool strict);
michael@0: 
michael@0: /**
michael@0:  * Function for handling "append code point" with error-checking.
michael@0:  *
michael@0:  * This is internal since it is not meant to be called directly by external clients;
michael@0:  * however it is U_STABLE (not U_INTERNAL) since it is called by public macros in this
michael@0:  * file and thus must remain stable, and should not be hidden when other internal
michael@0:  * functions are hidden (otherwise public macros would fail to compile).
michael@0:  * @internal
michael@0:  */
michael@0: U_STABLE int32_t U_EXPORT2
michael@0: utf8_appendCharSafeBody(uint8_t *s, int32_t i, int32_t length, UChar32 c, UBool *pIsError);
michael@0: 
michael@0: /**
michael@0:  * Function for handling "previous code point" with error-checking.
michael@0:  *
michael@0:  * This is internal since it is not meant to be called directly by external clients;
michael@0:  * however it is U_STABLE (not U_INTERNAL) since it is called by public macros in this
michael@0:  * file and thus must remain stable, and should not be hidden when other internal
michael@0:  * functions are hidden (otherwise public macros would fail to compile).
michael@0:  * @internal
michael@0:  */
michael@0: U_STABLE UChar32 U_EXPORT2
michael@0: utf8_prevCharSafeBody(const uint8_t *s, int32_t start, int32_t *pi, UChar32 c, UBool strict);
michael@0: 
michael@0: /**
michael@0:  * Function for handling "skip backward one code point" with error-checking.
michael@0:  *
michael@0:  * This is internal since it is not meant to be called directly by external clients;
michael@0:  * however it is U_STABLE (not U_INTERNAL) since it is called by public macros in this
michael@0:  * file and thus must remain stable, and should not be hidden when other internal
michael@0:  * functions are hidden (otherwise public macros would fail to compile).
michael@0:  * @internal
michael@0:  */
michael@0: U_STABLE int32_t U_EXPORT2
michael@0: utf8_back1SafeBody(const uint8_t *s, int32_t start, int32_t i);
michael@0: 
michael@0: /* single-code point definitions -------------------------------------------- */
michael@0: 
michael@0: /**
michael@0:  * Does this code unit (byte) encode a code point by itself (US-ASCII 0..0x7f)?
michael@0:  * @param c 8-bit code unit (byte)
michael@0:  * @return TRUE or FALSE
michael@0:  * @stable ICU 2.4
michael@0:  */
michael@0: #define U8_IS_SINGLE(c) (((c)&0x80)==0)
michael@0: 
michael@0: /**
michael@0:  * Is this code unit (byte) a UTF-8 lead byte?
michael@0:  * @param c 8-bit code unit (byte)
michael@0:  * @return TRUE or FALSE
michael@0:  * @stable ICU 2.4
michael@0:  */
michael@0: #define U8_IS_LEAD(c) ((uint8_t)((c)-0xc0)<0x3e)
michael@0: 
michael@0: /**
michael@0:  * Is this code unit (byte) a UTF-8 trail byte?
michael@0:  * @param c 8-bit code unit (byte)
michael@0:  * @return TRUE or FALSE
michael@0:  * @stable ICU 2.4
michael@0:  */
michael@0: #define U8_IS_TRAIL(c) (((c)&0xc0)==0x80)
michael@0: 
michael@0: /**
michael@0:  * How many code units (bytes) are used for the UTF-8 encoding
michael@0:  * of this Unicode code point?
michael@0:  * @param c 32-bit code point
michael@0:  * @return 1..4, or 0 if c is a surrogate or not a Unicode code point
michael@0:  * @stable ICU 2.4
michael@0:  */
michael@0: #define U8_LENGTH(c) \
michael@0:     ((uint32_t)(c)<=0x7f ? 1 : \
michael@0:         ((uint32_t)(c)<=0x7ff ? 2 : \
michael@0:             ((uint32_t)(c)<=0xd7ff ? 3 : \
michael@0:                 ((uint32_t)(c)<=0xdfff || (uint32_t)(c)>0x10ffff ? 0 : \
michael@0:                     ((uint32_t)(c)<=0xffff ? 3 : 4)\
michael@0:                 ) \
michael@0:             ) \
michael@0:         ) \
michael@0:     )
michael@0: 
michael@0: /**
michael@0:  * The maximum number of UTF-8 code units (bytes) per Unicode code point (U+0000..U+10ffff).
michael@0:  * @return 4
michael@0:  * @stable ICU 2.4
michael@0:  */
michael@0: #define U8_MAX_LENGTH 4
michael@0: 
michael@0: /**
michael@0:  * Get a code point from a string at a random-access offset,
michael@0:  * without changing the offset.
michael@0:  * The offset may point to either the lead byte or one of the trail bytes
michael@0:  * for a code point, in which case the macro will read all of the bytes
michael@0:  * for the code point.
michael@0:  * The result is undefined if the offset points to an illegal UTF-8
michael@0:  * byte sequence.
michael@0:  * Iteration through a string is more efficient with U8_NEXT_UNSAFE or U8_NEXT.
michael@0:  *
michael@0:  * @param s const uint8_t * string
michael@0:  * @param i string offset
michael@0:  * @param c output UChar32 variable
michael@0:  * @see U8_GET
michael@0:  * @stable ICU 2.4
michael@0:  */
michael@0: #define U8_GET_UNSAFE(s, i, c) { \
michael@0:     int32_t _u8_get_unsafe_index=(int32_t)(i); \
michael@0:     U8_SET_CP_START_UNSAFE(s, _u8_get_unsafe_index); \
michael@0:     U8_NEXT_UNSAFE(s, _u8_get_unsafe_index, c); \
michael@0: }
michael@0: 
michael@0: /**
michael@0:  * Get a code point from a string at a random-access offset,
michael@0:  * without changing the offset.
michael@0:  * The offset may point to either the lead byte or one of the trail bytes
michael@0:  * for a code point, in which case the macro will read all of the bytes
michael@0:  * for the code point.
michael@0:  *
michael@0:  * The length can be negative for a NUL-terminated string.
michael@0:  *
michael@0:  * If the offset points to an illegal UTF-8 byte sequence, then
michael@0:  * c is set to a negative value.
michael@0:  * Iteration through a string is more efficient with U8_NEXT_UNSAFE or U8_NEXT.
michael@0:  *
michael@0:  * @param s const uint8_t * string
michael@0:  * @param start int32_t starting string offset
michael@0:  * @param i int32_t string offset, must be start<=i<length
michael@0:  * @param length int32_t string length
michael@0:  * @param c output UChar32 variable, set to <0 in case of an error
michael@0:  * @see U8_GET_UNSAFE
michael@0:  * @stable ICU 2.4
michael@0:  */
michael@0: #define U8_GET(s, start, i, length, c) { \
michael@0:     int32_t _u8_get_index=(i); \
michael@0:     U8_SET_CP_START(s, start, _u8_get_index); \
michael@0:     U8_NEXT(s, _u8_get_index, length, c); \
michael@0: }
michael@0: 
michael@0: #ifndef U_HIDE_DRAFT_API
michael@0: /**
michael@0:  * Get a code point from a string at a random-access offset,
michael@0:  * without changing the offset.
michael@0:  * The offset may point to either the lead byte or one of the trail bytes
michael@0:  * for a code point, in which case the macro will read all of the bytes
michael@0:  * for the code point.
michael@0:  *
michael@0:  * The length can be negative for a NUL-terminated string.
michael@0:  *
michael@0:  * If the offset points to an illegal UTF-8 byte sequence, then
michael@0:  * c is set to U+FFFD.
michael@0:  * Iteration through a string is more efficient with U8_NEXT_UNSAFE or U8_NEXT_OR_FFFD.
michael@0:  *
michael@0:  * This macro does not distinguish between a real U+FFFD in the text
michael@0:  * and U+FFFD returned for an ill-formed sequence.
michael@0:  * Use U8_GET() if that distinction is important.
michael@0:  *
michael@0:  * @param s const uint8_t * string
michael@0:  * @param start int32_t starting string offset
michael@0:  * @param i int32_t string offset, must be start<=i<length
michael@0:  * @param length int32_t string length
michael@0:  * @param c output UChar32 variable, set to U+FFFD in case of an error
michael@0:  * @see U8_GET
michael@0:  * @draft ICU 51
michael@0:  */
michael@0: #define U8_GET_OR_FFFD(s, start, i, length, c) { \
michael@0:     int32_t _u8_get_index=(i); \
michael@0:     U8_SET_CP_START(s, start, _u8_get_index); \
michael@0:     U8_NEXT_OR_FFFD(s, _u8_get_index, length, c); \
michael@0: }
michael@0: #endif /* U_HIDE_DRAFT_API */
michael@0: 
michael@0: /* definitions with forward iteration --------------------------------------- */
michael@0: 
michael@0: /**
michael@0:  * Get a code point from a string at a code point boundary offset,
michael@0:  * and advance the offset to the next code point boundary.
michael@0:  * (Post-incrementing forward iteration.)
michael@0:  * "Unsafe" macro, assumes well-formed UTF-8.
michael@0:  *
michael@0:  * The offset may point to the lead byte of a multi-byte sequence,
michael@0:  * in which case the macro will read the whole sequence.
michael@0:  * The result is undefined if the offset points to a trail byte
michael@0:  * or an illegal UTF-8 sequence.
michael@0:  *
michael@0:  * @param s const uint8_t * string
michael@0:  * @param i string offset
michael@0:  * @param c output UChar32 variable
michael@0:  * @see U8_NEXT
michael@0:  * @stable ICU 2.4
michael@0:  */
michael@0: #define U8_NEXT_UNSAFE(s, i, c) { \
michael@0:     (c)=(uint8_t)(s)[(i)++]; \
michael@0:     if((c)>=0x80) { \
michael@0:         if((c)<0xe0) { \
michael@0:             (c)=(((c)&0x1f)<<6)|((s)[(i)++]&0x3f); \
michael@0:         } else if((c)<0xf0) { \
michael@0:             /* no need for (c&0xf) because the upper bits are truncated after <<12 in the cast to (UChar) */ \
michael@0:             (c)=(UChar)(((c)<<12)|(((s)[i]&0x3f)<<6)|((s)[(i)+1]&0x3f)); \
michael@0:             (i)+=2; \
michael@0:         } else { \
michael@0:             (c)=(((c)&7)<<18)|(((s)[i]&0x3f)<<12)|(((s)[(i)+1]&0x3f)<<6)|((s)[(i)+2]&0x3f); \
michael@0:             (i)+=3; \
michael@0:         } \
michael@0:     } \
michael@0: }
michael@0: 
michael@0: /**
michael@0:  * Get a code point from a string at a code point boundary offset,
michael@0:  * and advance the offset to the next code point boundary.
michael@0:  * (Post-incrementing forward iteration.)
michael@0:  * "Safe" macro, checks for illegal sequences and for string boundaries.
michael@0:  *
michael@0:  * The length can be negative for a NUL-terminated string.
michael@0:  *
michael@0:  * The offset may point to the lead byte of a multi-byte sequence,
michael@0:  * in which case the macro will read the whole sequence.
michael@0:  * If the offset points to a trail byte or an illegal UTF-8 sequence, then
michael@0:  * c is set to a negative value.
michael@0:  *
michael@0:  * @param s const uint8_t * string
michael@0:  * @param i int32_t string offset, must be i<length
michael@0:  * @param length int32_t string length
michael@0:  * @param c output UChar32 variable, set to <0 in case of an error
michael@0:  * @see U8_NEXT_UNSAFE
michael@0:  * @stable ICU 2.4
michael@0:  */
michael@0: #define U8_NEXT(s, i, length, c) { \
michael@0:     (c)=(uint8_t)(s)[(i)++]; \
michael@0:     if((c)>=0x80) { \
michael@0:         uint8_t __t1, __t2; \
michael@0:         if( /* handle U+1000..U+CFFF inline */ \
michael@0:             (0xe0<(c) && (c)<=0xec) && \
michael@0:             (((i)+1)<(length) || (length)<0) && \
michael@0:             (__t1=(uint8_t)((s)[i]-0x80))<=0x3f && \
michael@0:             (__t2=(uint8_t)((s)[(i)+1]-0x80))<= 0x3f \
michael@0:         ) { \
michael@0:             /* no need for (c&0xf) because the upper bits are truncated after <<12 in the cast to (UChar) */ \
michael@0:             (c)=(UChar)(((c)<<12)|(__t1<<6)|__t2); \
michael@0:             (i)+=2; \
michael@0:         } else if( /* handle U+0080..U+07FF inline */ \
michael@0:             ((c)<0xe0 && (c)>=0xc2) && \
michael@0:             ((i)!=(length)) && \
michael@0:             (__t1=(uint8_t)((s)[i]-0x80))<=0x3f \
michael@0:         ) { \
michael@0:             (c)=(((c)&0x1f)<<6)|__t1; \
michael@0:             ++(i); \
michael@0:         } else { \
michael@0:             /* function call for "complicated" and error cases */ \
michael@0:             (c)=utf8_nextCharSafeBody((const uint8_t *)s, &(i), (length), c, -1); \
michael@0:         } \
michael@0:     } \
michael@0: }
michael@0: 
michael@0: #ifndef U_HIDE_DRAFT_API
michael@0: /**
michael@0:  * Get a code point from a string at a code point boundary offset,
michael@0:  * and advance the offset to the next code point boundary.
michael@0:  * (Post-incrementing forward iteration.)
michael@0:  * "Safe" macro, checks for illegal sequences and for string boundaries.
michael@0:  *
michael@0:  * The length can be negative for a NUL-terminated string.
michael@0:  *
michael@0:  * The offset may point to the lead byte of a multi-byte sequence,
michael@0:  * in which case the macro will read the whole sequence.
michael@0:  * If the offset points to a trail byte or an illegal UTF-8 sequence, then
michael@0:  * c is set to U+FFFD.
michael@0:  *
michael@0:  * This macro does not distinguish between a real U+FFFD in the text
michael@0:  * and U+FFFD returned for an ill-formed sequence.
michael@0:  * Use U8_NEXT() if that distinction is important.
michael@0:  *
michael@0:  * @param s const uint8_t * string
michael@0:  * @param i int32_t string offset, must be i<length
michael@0:  * @param length int32_t string length
michael@0:  * @param c output UChar32 variable, set to U+FFFD in case of an error
michael@0:  * @see U8_NEXT
michael@0:  * @draft ICU 51
michael@0:  */
michael@0: #define U8_NEXT_OR_FFFD(s, i, length, c) { \
michael@0:     (c)=(uint8_t)(s)[(i)++]; \
michael@0:     if((c)>=0x80) { \
michael@0:         uint8_t __t1, __t2; \
michael@0:         if( /* handle U+1000..U+CFFF inline */ \
michael@0:             (0xe0<(c) && (c)<=0xec) && \
michael@0:             (((i)+1)<(length) || (length)<0) && \
michael@0:             (__t1=(uint8_t)((s)[i]-0x80))<=0x3f && \
michael@0:             (__t2=(uint8_t)((s)[(i)+1]-0x80))<= 0x3f \
michael@0:         ) { \
michael@0:             /* no need for (c&0xf) because the upper bits are truncated after <<12 in the cast to (UChar) */ \
michael@0:             (c)=(UChar)(((c)<<12)|(__t1<<6)|__t2); \
michael@0:             (i)+=2; \
michael@0:         } else if( /* handle U+0080..U+07FF inline */ \
michael@0:             ((c)<0xe0 && (c)>=0xc2) && \
michael@0:             ((i)!=(length)) && \
michael@0:             (__t1=(uint8_t)((s)[i]-0x80))<=0x3f \
michael@0:         ) { \
michael@0:             (c)=(((c)&0x1f)<<6)|__t1; \
michael@0:             ++(i); \
michael@0:         } else { \
michael@0:             /* function call for "complicated" and error cases */ \
michael@0:             (c)=utf8_nextCharSafeBody((const uint8_t *)s, &(i), (length), c, -3); \
michael@0:         } \
michael@0:     } \
michael@0: }
michael@0: #endif /* U_HIDE_DRAFT_API */
michael@0: 
michael@0: /**
michael@0:  * Append a code point to a string, overwriting 1 to 4 bytes.
michael@0:  * The offset points to the current end of the string contents
michael@0:  * and is advanced (post-increment).
michael@0:  * "Unsafe" macro, assumes a valid code point and sufficient space in the string.
michael@0:  * Otherwise, the result is undefined.
michael@0:  *
michael@0:  * @param s const uint8_t * string buffer
michael@0:  * @param i string offset
michael@0:  * @param c code point to append
michael@0:  * @see U8_APPEND
michael@0:  * @stable ICU 2.4
michael@0:  */
michael@0: #define U8_APPEND_UNSAFE(s, i, c) { \
michael@0:     if((uint32_t)(c)<=0x7f) { \
michael@0:         (s)[(i)++]=(uint8_t)(c); \
michael@0:     } else { \
michael@0:         if((uint32_t)(c)<=0x7ff) { \
michael@0:             (s)[(i)++]=(uint8_t)(((c)>>6)|0xc0); \
michael@0:         } else { \
michael@0:             if((uint32_t)(c)<=0xffff) { \
michael@0:                 (s)[(i)++]=(uint8_t)(((c)>>12)|0xe0); \
michael@0:             } else { \
michael@0:                 (s)[(i)++]=(uint8_t)(((c)>>18)|0xf0); \
michael@0:                 (s)[(i)++]=(uint8_t)((((c)>>12)&0x3f)|0x80); \
michael@0:             } \
michael@0:             (s)[(i)++]=(uint8_t)((((c)>>6)&0x3f)|0x80); \
michael@0:         } \
michael@0:         (s)[(i)++]=(uint8_t)(((c)&0x3f)|0x80); \
michael@0:     } \
michael@0: }
michael@0: 
michael@0: /**
michael@0:  * Append a code point to a string, overwriting 1 to 4 bytes.
michael@0:  * The offset points to the current end of the string contents
michael@0:  * and is advanced (post-increment).
michael@0:  * "Safe" macro, checks for a valid code point.
michael@0:  * If a non-ASCII code point is written, checks for sufficient space in the string.
michael@0:  * If the code point is not valid or trail bytes do not fit,
michael@0:  * then isError is set to TRUE.
michael@0:  *
michael@0:  * @param s const uint8_t * string buffer
michael@0:  * @param i int32_t string offset, must be i<capacity
michael@0:  * @param capacity int32_t size of the string buffer
michael@0:  * @param c UChar32 code point to append
michael@0:  * @param isError output UBool set to TRUE if an error occurs, otherwise not modified
michael@0:  * @see U8_APPEND_UNSAFE
michael@0:  * @stable ICU 2.4
michael@0:  */
michael@0: #define U8_APPEND(s, i, capacity, c, isError) { \
michael@0:     if((uint32_t)(c)<=0x7f) { \
michael@0:         (s)[(i)++]=(uint8_t)(c); \
michael@0:     } else if((uint32_t)(c)<=0x7ff && (i)+1<(capacity)) { \
michael@0:         (s)[(i)++]=(uint8_t)(((c)>>6)|0xc0); \
michael@0:         (s)[(i)++]=(uint8_t)(((c)&0x3f)|0x80); \
michael@0:     } else if((uint32_t)(c)<=0xd7ff && (i)+2<(capacity)) { \
michael@0:         (s)[(i)++]=(uint8_t)(((c)>>12)|0xe0); \
michael@0:         (s)[(i)++]=(uint8_t)((((c)>>6)&0x3f)|0x80); \
michael@0:         (s)[(i)++]=(uint8_t)(((c)&0x3f)|0x80); \
michael@0:     } else { \
michael@0:         (i)=utf8_appendCharSafeBody(s, (i), (capacity), c, &(isError)); \
michael@0:     } \
michael@0: }
michael@0: 
michael@0: /**
michael@0:  * Advance the string offset from one code point boundary to the next.
michael@0:  * (Post-incrementing iteration.)
michael@0:  * "Unsafe" macro, assumes well-formed UTF-8.
michael@0:  *
michael@0:  * @param s const uint8_t * string
michael@0:  * @param i string offset
michael@0:  * @see U8_FWD_1
michael@0:  * @stable ICU 2.4
michael@0:  */
michael@0: #define U8_FWD_1_UNSAFE(s, i) { \
michael@0:     (i)+=1+U8_COUNT_TRAIL_BYTES_UNSAFE((uint8_t)(s)[i]); \
michael@0: }
michael@0: 
michael@0: /**
michael@0:  * Advance the string offset from one code point boundary to the next.
michael@0:  * (Post-incrementing iteration.)
michael@0:  * "Safe" macro, checks for illegal sequences and for string boundaries.
michael@0:  *
michael@0:  * The length can be negative for a NUL-terminated string.
michael@0:  *
michael@0:  * @param s const uint8_t * string
michael@0:  * @param i int32_t string offset, must be i<length
michael@0:  * @param length int32_t string length
michael@0:  * @see U8_FWD_1_UNSAFE
michael@0:  * @stable ICU 2.4
michael@0:  */
michael@0: #define U8_FWD_1(s, i, length) { \
michael@0:     uint8_t __b=(uint8_t)(s)[(i)++]; \
michael@0:     if(U8_IS_LEAD(__b)) { \
michael@0:         uint8_t __count=U8_COUNT_TRAIL_BYTES(__b); \
michael@0:         if((i)+__count>(length) && (length)>=0) { \
michael@0:             __count=(uint8_t)((length)-(i)); \
michael@0:         } \
michael@0:         while(__count>0 && U8_IS_TRAIL((s)[i])) { \
michael@0:             ++(i); \
michael@0:             --__count; \
michael@0:         } \
michael@0:     } \
michael@0: }
michael@0: 
michael@0: /**
michael@0:  * Advance the string offset from one code point boundary to the n-th next one,
michael@0:  * i.e., move forward by n code points.
michael@0:  * (Post-incrementing iteration.)
michael@0:  * "Unsafe" macro, assumes well-formed UTF-8.
michael@0:  *
michael@0:  * @param s const uint8_t * string
michael@0:  * @param i string offset
michael@0:  * @param n number of code points to skip
michael@0:  * @see U8_FWD_N
michael@0:  * @stable ICU 2.4
michael@0:  */
michael@0: #define U8_FWD_N_UNSAFE(s, i, n) { \
michael@0:     int32_t __N=(n); \
michael@0:     while(__N>0) { \
michael@0:         U8_FWD_1_UNSAFE(s, i); \
michael@0:         --__N; \
michael@0:     } \
michael@0: }
michael@0: 
michael@0: /**
michael@0:  * Advance the string offset from one code point boundary to the n-th next one,
michael@0:  * i.e., move forward by n code points.
michael@0:  * (Post-incrementing iteration.)
michael@0:  * "Safe" macro, checks for illegal sequences and for string boundaries.
michael@0:  *
michael@0:  * The length can be negative for a NUL-terminated string.
michael@0:  *
michael@0:  * @param s const uint8_t * string
michael@0:  * @param i int32_t string offset, must be i<length
michael@0:  * @param length int32_t string length
michael@0:  * @param n number of code points to skip
michael@0:  * @see U8_FWD_N_UNSAFE
michael@0:  * @stable ICU 2.4
michael@0:  */
michael@0: #define U8_FWD_N(s, i, length, n) { \
michael@0:     int32_t __N=(n); \
michael@0:     while(__N>0 && ((i)<(length) || ((length)<0 && (s)[i]!=0))) { \
michael@0:         U8_FWD_1(s, i, length); \
michael@0:         --__N; \
michael@0:     } \
michael@0: }
michael@0: 
michael@0: /**
michael@0:  * Adjust a random-access offset to a code point boundary
michael@0:  * at the start of a code point.
michael@0:  * If the offset points to a UTF-8 trail byte,
michael@0:  * then the offset is moved backward to the corresponding lead byte.
michael@0:  * Otherwise, it is not modified.
michael@0:  * "Unsafe" macro, assumes well-formed UTF-8.
michael@0:  *
michael@0:  * @param s const uint8_t * string
michael@0:  * @param i string offset
michael@0:  * @see U8_SET_CP_START
michael@0:  * @stable ICU 2.4
michael@0:  */
michael@0: #define U8_SET_CP_START_UNSAFE(s, i) { \
michael@0:     while(U8_IS_TRAIL((s)[i])) { --(i); } \
michael@0: }
michael@0: 
michael@0: /**
michael@0:  * Adjust a random-access offset to a code point boundary
michael@0:  * at the start of a code point.
michael@0:  * If the offset points to a UTF-8 trail byte,
michael@0:  * then the offset is moved backward to the corresponding lead byte.
michael@0:  * Otherwise, it is not modified.
michael@0:  * "Safe" macro, checks for illegal sequences and for string boundaries.
michael@0:  *
michael@0:  * @param s const uint8_t * string
michael@0:  * @param start int32_t starting string offset (usually 0)
michael@0:  * @param i int32_t string offset, must be start<=i
michael@0:  * @see U8_SET_CP_START_UNSAFE
michael@0:  * @stable ICU 2.4
michael@0:  */
michael@0: #define U8_SET_CP_START(s, start, i) { \
michael@0:     if(U8_IS_TRAIL((s)[(i)])) { \
michael@0:         (i)=utf8_back1SafeBody(s, start, (i)); \
michael@0:     } \
michael@0: }
michael@0: 
michael@0: /* definitions with backward iteration -------------------------------------- */
michael@0: 
michael@0: /**
michael@0:  * Move the string offset from one code point boundary to the previous one
michael@0:  * and get the code point between them.
michael@0:  * (Pre-decrementing backward iteration.)
michael@0:  * "Unsafe" macro, assumes well-formed UTF-8.
michael@0:  *
michael@0:  * The input offset may be the same as the string length.
michael@0:  * If the offset is behind a multi-byte sequence, then the macro will read
michael@0:  * the whole sequence.
michael@0:  * If the offset is behind a lead byte, then that itself
michael@0:  * will be returned as the code point.
michael@0:  * The result is undefined if the offset is behind an illegal UTF-8 sequence.
michael@0:  *
michael@0:  * @param s const uint8_t * string
michael@0:  * @param i string offset
michael@0:  * @param c output UChar32 variable
michael@0:  * @see U8_PREV
michael@0:  * @stable ICU 2.4
michael@0:  */
michael@0: #define U8_PREV_UNSAFE(s, i, c) { \
michael@0:     (c)=(uint8_t)(s)[--(i)]; \
michael@0:     if(U8_IS_TRAIL(c)) { \
michael@0:         uint8_t __b, __count=1, __shift=6; \
michael@0: \
michael@0:         /* c is a trail byte */ \
michael@0:         (c)&=0x3f; \
michael@0:         for(;;) { \
michael@0:             __b=(uint8_t)(s)[--(i)]; \
michael@0:             if(__b>=0xc0) { \
michael@0:                 U8_MASK_LEAD_BYTE(__b, __count); \
michael@0:                 (c)|=(UChar32)__b<<__shift; \
michael@0:                 break; \
michael@0:             } else { \
michael@0:                 (c)|=(UChar32)(__b&0x3f)<<__shift; \
michael@0:                 ++__count; \
michael@0:                 __shift+=6; \
michael@0:             } \
michael@0:         } \
michael@0:     } \
michael@0: }
michael@0: 
michael@0: /**
michael@0:  * Move the string offset from one code point boundary to the previous one
michael@0:  * and get the code point between them.
michael@0:  * (Pre-decrementing backward iteration.)
michael@0:  * "Safe" macro, checks for illegal sequences and for string boundaries.
michael@0:  *
michael@0:  * The input offset may be the same as the string length.
michael@0:  * If the offset is behind a multi-byte sequence, then the macro will read
michael@0:  * the whole sequence.
michael@0:  * If the offset is behind a lead byte, then that itself
michael@0:  * will be returned as the code point.
michael@0:  * If the offset is behind an illegal UTF-8 sequence, then c is set to a negative value.
michael@0:  *
michael@0:  * @param s const uint8_t * string
michael@0:  * @param start int32_t starting string offset (usually 0)
michael@0:  * @param i int32_t string offset, must be start<i
michael@0:  * @param c output UChar32 variable, set to <0 in case of an error
michael@0:  * @see U8_PREV_UNSAFE
michael@0:  * @stable ICU 2.4
michael@0:  */
michael@0: #define U8_PREV(s, start, i, c) { \
michael@0:     (c)=(uint8_t)(s)[--(i)]; \
michael@0:     if((c)>=0x80) { \
michael@0:         (c)=utf8_prevCharSafeBody((const uint8_t *)s, start, &(i), c, -1); \
michael@0:     } \
michael@0: }
michael@0: 
michael@0: #ifndef U_HIDE_DRAFT_API
michael@0: /**
michael@0:  * Move the string offset from one code point boundary to the previous one
michael@0:  * and get the code point between them.
michael@0:  * (Pre-decrementing backward iteration.)
michael@0:  * "Safe" macro, checks for illegal sequences and for string boundaries.
michael@0:  *
michael@0:  * The input offset may be the same as the string length.
michael@0:  * If the offset is behind a multi-byte sequence, then the macro will read
michael@0:  * the whole sequence.
michael@0:  * If the offset is behind a lead byte, then that itself
michael@0:  * will be returned as the code point.
michael@0:  * If the offset is behind an illegal UTF-8 sequence, then c is set to U+FFFD.
michael@0:  *
michael@0:  * This macro does not distinguish between a real U+FFFD in the text
michael@0:  * and U+FFFD returned for an ill-formed sequence.
michael@0:  * Use U8_PREV() if that distinction is important.
michael@0:  *
michael@0:  * @param s const uint8_t * string
michael@0:  * @param start int32_t starting string offset (usually 0)
michael@0:  * @param i int32_t string offset, must be start<i
michael@0:  * @param c output UChar32 variable, set to U+FFFD in case of an error
michael@0:  * @see U8_PREV
michael@0:  * @draft ICU 51
michael@0:  */
michael@0: #define U8_PREV_OR_FFFD(s, start, i, c) { \
michael@0:     (c)=(uint8_t)(s)[--(i)]; \
michael@0:     if((c)>=0x80) { \
michael@0:         (c)=utf8_prevCharSafeBody((const uint8_t *)s, start, &(i), c, -3); \
michael@0:     } \
michael@0: }
michael@0: #endif /* U_HIDE_DRAFT_API */
michael@0: 
michael@0: /**
michael@0:  * Move the string offset from one code point boundary to the previous one.
michael@0:  * (Pre-decrementing backward iteration.)
michael@0:  * The input offset may be the same as the string length.
michael@0:  * "Unsafe" macro, assumes well-formed UTF-8.
michael@0:  *
michael@0:  * @param s const uint8_t * string
michael@0:  * @param i string offset
michael@0:  * @see U8_BACK_1
michael@0:  * @stable ICU 2.4
michael@0:  */
michael@0: #define U8_BACK_1_UNSAFE(s, i) { \
michael@0:     while(U8_IS_TRAIL((s)[--(i)])) {} \
michael@0: }
michael@0: 
michael@0: /**
michael@0:  * Move the string offset from one code point boundary to the previous one.
michael@0:  * (Pre-decrementing backward iteration.)
michael@0:  * The input offset may be the same as the string length.
michael@0:  * "Safe" macro, checks for illegal sequences and for string boundaries.
michael@0:  *
michael@0:  * @param s const uint8_t * string
michael@0:  * @param start int32_t starting string offset (usually 0)
michael@0:  * @param i int32_t string offset, must be start<i
michael@0:  * @see U8_BACK_1_UNSAFE
michael@0:  * @stable ICU 2.4
michael@0:  */
michael@0: #define U8_BACK_1(s, start, i) { \
michael@0:     if(U8_IS_TRAIL((s)[--(i)])) { \
michael@0:         (i)=utf8_back1SafeBody(s, start, (i)); \
michael@0:     } \
michael@0: }
michael@0: 
michael@0: /**
michael@0:  * Move the string offset from one code point boundary to the n-th one before it,
michael@0:  * i.e., move backward by n code points.
michael@0:  * (Pre-decrementing backward iteration.)
michael@0:  * The input offset may be the same as the string length.
michael@0:  * "Unsafe" macro, assumes well-formed UTF-8.
michael@0:  *
michael@0:  * @param s const uint8_t * string
michael@0:  * @param i string offset
michael@0:  * @param n number of code points to skip
michael@0:  * @see U8_BACK_N
michael@0:  * @stable ICU 2.4
michael@0:  */
michael@0: #define U8_BACK_N_UNSAFE(s, i, n) { \
michael@0:     int32_t __N=(n); \
michael@0:     while(__N>0) { \
michael@0:         U8_BACK_1_UNSAFE(s, i); \
michael@0:         --__N; \
michael@0:     } \
michael@0: }
michael@0: 
michael@0: /**
michael@0:  * Move the string offset from one code point boundary to the n-th one before it,
michael@0:  * i.e., move backward by n code points.
michael@0:  * (Pre-decrementing backward iteration.)
michael@0:  * The input offset may be the same as the string length.
michael@0:  * "Safe" macro, checks for illegal sequences and for string boundaries.
michael@0:  *
michael@0:  * @param s const uint8_t * string
michael@0:  * @param start int32_t index of the start of the string
michael@0:  * @param i int32_t string offset, must be start<i
michael@0:  * @param n number of code points to skip
michael@0:  * @see U8_BACK_N_UNSAFE
michael@0:  * @stable ICU 2.4
michael@0:  */
michael@0: #define U8_BACK_N(s, start, i, n) { \
michael@0:     int32_t __N=(n); \
michael@0:     while(__N>0 && (i)>(start)) { \
michael@0:         U8_BACK_1(s, start, i); \
michael@0:         --__N; \
michael@0:     } \
michael@0: }
michael@0: 
michael@0: /**
michael@0:  * Adjust a random-access offset to a code point boundary after a code point.
michael@0:  * If the offset is behind a partial multi-byte sequence,
michael@0:  * then the offset is incremented to behind the whole sequence.
michael@0:  * Otherwise, it is not modified.
michael@0:  * The input offset may be the same as the string length.
michael@0:  * "Unsafe" macro, assumes well-formed UTF-8.
michael@0:  *
michael@0:  * @param s const uint8_t * string
michael@0:  * @param i string offset
michael@0:  * @see U8_SET_CP_LIMIT
michael@0:  * @stable ICU 2.4
michael@0:  */
michael@0: #define U8_SET_CP_LIMIT_UNSAFE(s, i) { \
michael@0:     U8_BACK_1_UNSAFE(s, i); \
michael@0:     U8_FWD_1_UNSAFE(s, i); \
michael@0: }
michael@0: 
michael@0: /**
michael@0:  * Adjust a random-access offset to a code point boundary after a code point.
michael@0:  * If the offset is behind a partial multi-byte sequence,
michael@0:  * then the offset is incremented to behind the whole sequence.
michael@0:  * Otherwise, it is not modified.
michael@0:  * The input offset may be the same as the string length.
michael@0:  * "Safe" macro, checks for illegal sequences and for string boundaries.
michael@0:  *
michael@0:  * The length can be negative for a NUL-terminated string.
michael@0:  *
michael@0:  * @param s const uint8_t * string
michael@0:  * @param start int32_t starting string offset (usually 0)
michael@0:  * @param i int32_t string offset, must be start<=i<=length
michael@0:  * @param length int32_t string length
michael@0:  * @see U8_SET_CP_LIMIT_UNSAFE
michael@0:  * @stable ICU 2.4
michael@0:  */
michael@0: #define U8_SET_CP_LIMIT(s, start, i, length) { \
michael@0:     if((start)<(i) && ((i)<(length) || ((length)<0 && (s)[i]!=0))) { \
michael@0:         U8_BACK_1(s, start, i); \
michael@0:         U8_FWD_1(s, i, length); \
michael@0:     } \
michael@0: }
michael@0: 
michael@0: #endif