ICU-59117.0.1.tar.gz

[apple/icu.git] / icuSources / common / unicode / utf8.h

diff --git a/icuSources/common/unicode/utf8.h b/icuSources/common/unicode/utf8.h
index 20c168e1c55b9a323498c5c531f324a086d0a666..18e7e9455c97f9fe74b64891ea1efc7c4a75a00d 100644 (file)
--- a/icuSources/common/unicode/utf8.h
+++ b/icuSources/common/unicode/utf8.h
@@ -1,12 +1,14 @@
+// © 2016 and later: Unicode, Inc. and others.
+// License & terms of use: http://www.unicode.org/copyright.html

 /*
 *******************************************************************************
 *
 /*
 *******************************************************************************
 *

-*   Copyright (C) 1999-2001, International Business Machines
+*   Copyright (C) 1999-2015, International Business Machines

 *   Corporation and others.  All Rights Reserved.
 *
 *******************************************************************************
 *   file name:  utf8.h
 *   Corporation and others.  All Rights Reserved.
 *
 *******************************************************************************
 *   file name:  utf8.h

-*   encoding:   US-ASCII
+*   encoding:   UTF-8

 *   tab size:   8 (not used)
 *   indentation:4
 *
 *   tab size:   8 (not used)
 *   indentation:4
 *


@@ -19,11 +21,9 @@
  * \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.
  * \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
  *
  * For more information see utf.h and the ICU User Guide Strings chapter

- * (http://oss.software.ibm.com/icu/userguide/).
+ * (http://userguide.icu-project.org/strings).

  *
  * <em>Usage:</em>
  * ICU coding guidelines for if() statements should be followed when using these macros.
  *
  * <em>Usage:</em>
  * ICU coding guidelines for if() statements should be followed when using these macros.


@@ -31,68 +31,110 @@
  * bodies and all macro statements should be terminated with semicolon.
  */
 
  * bodies and all macro statements should be terminated with semicolon.
  */
 

-/* utf.h must be included first. */
+#ifndef __UTF8_H__
+#define __UTF8_H__
+
+#include "unicode/umachine.h"

 #ifndef __UTF_H__
 #   include "unicode/utf.h"
 #endif
 
 #ifndef __UTF_H__
 #   include "unicode/utf.h"
 #endif
 

-#ifndef __UTF8_H__
-#define __UTF8_H__
-

 /* internal definitions ----------------------------------------------------- */
 
 /* 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
  */
  * @internal
  */

-#ifdef U_UTF8_IMPL
-U_CAPI const uint8_t 
-utf8_countTrailBytes[256];
-#else
-U_CFUNC U_IMPORT const uint8_t /* U_IMPORT2? */ /*U_IMPORT*/ 
-utf8_countTrailBytes[256];
-#endif
+#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
  */
  * @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.
 
 /**
  * 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.
  * @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
  */
  * @internal
  */

-U_CAPI 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.
 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
  */
  * @internal
  */

-U_CAPI 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.
 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
  */
  * @internal
  */

-U_CAPI 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.
 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
  */
  * @internal
  */

-U_CAPI 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 -------------------------------------------- */
 utf8_back1SafeBody(const uint8_t *s, int32_t start, int32_t i);
 
 /* single-code point definitions -------------------------------------------- */


@@ -101,7 +143,7 @@ utf8_back1SafeBody(const uint8_t *s, int32_t start, int32_t i);
  * Does this code unit (byte) encode a code point by itself (US-ASCII 0..0x7f)?
  * @param c 8-bit code unit (byte)
  * @return TRUE or FALSE
  * Does this code unit (byte) encode a code point by itself (US-ASCII 0..0x7f)?
  * @param c 8-bit code unit (byte)
  * @return TRUE or FALSE

- * @draft ICU 2.4
+ * @stable ICU 2.4

  */
 #define U8_IS_SINGLE(c) (((c)&0x80)==0)
 
  */
 #define U8_IS_SINGLE(c) (((c)&0x80)==0)
 


@@ -109,7 +151,7 @@ utf8_back1SafeBody(const uint8_t *s, int32_t start, int32_t i);
  * Is this code unit (byte) a UTF-8 lead byte?
  * @param c 8-bit code unit (byte)
  * @return TRUE or FALSE
  * Is this code unit (byte) a UTF-8 lead byte?
  * @param c 8-bit code unit (byte)
  * @return TRUE or FALSE

- * @draft ICU 2.4
+ * @stable ICU 2.4

  */
 #define U8_IS_LEAD(c) ((uint8_t)((c)-0xc0)<0x3e)
 
  */
 #define U8_IS_LEAD(c) ((uint8_t)((c)-0xc0)<0x3e)
 


@@ -117,7 +159,7 @@ utf8_back1SafeBody(const uint8_t *s, int32_t start, int32_t i);
  * Is this code unit (byte) a UTF-8 trail byte?
  * @param c 8-bit code unit (byte)
  * @return TRUE or FALSE
  * Is this code unit (byte) a UTF-8 trail byte?
  * @param c 8-bit code unit (byte)
  * @return TRUE or FALSE

- * @draft ICU 2.4
+ * @stable ICU 2.4

  */
 #define U8_IS_TRAIL(c) (((c)&0xc0)==0x80)
 
  */
 #define U8_IS_TRAIL(c) (((c)&0xc0)==0x80)
 


@@ -126,7 +168,7 @@ utf8_back1SafeBody(const uint8_t *s, int32_t start, int32_t i);
  * of this Unicode code point?
  * @param c 32-bit code point
  * @return 1..4, or 0 if c is a surrogate or not a Unicode code point
  * of this Unicode code point?
  * @param c 32-bit code point
  * @return 1..4, or 0 if c is a surrogate or not a Unicode code point

- * @draft ICU 2.4
+ * @stable ICU 2.4

  */
 #define U8_LENGTH(c) \
     ((uint32_t)(c)<=0x7f ? 1 : \
  */
 #define U8_LENGTH(c) \
     ((uint32_t)(c)<=0x7f ? 1 : \


@@ -142,7 +184,7 @@ utf8_back1SafeBody(const uint8_t *s, int32_t start, int32_t i);
 /**
  * The maximum number of UTF-8 code units (bytes) per Unicode code point (U+0000..U+10ffff).
  * @return 4
 /**
  * The maximum number of UTF-8 code units (bytes) per Unicode code point (U+0000..U+10ffff).
  * @return 4

- * @draft ICU 2.4
+ * @stable ICU 2.4

  */
 #define U8_MAX_LENGTH 4
 
  */
 #define U8_MAX_LENGTH 4
 


@@ -156,16 +198,16 @@ utf8_back1SafeBody(const uint8_t *s, int32_t start, int32_t i);
  * byte sequence.
  * Iteration through a string is more efficient with U8_NEXT_UNSAFE or U8_NEXT.
  *
  * byte sequence.
  * Iteration through a string is more efficient with U8_NEXT_UNSAFE or U8_NEXT.
  *

- * @param s const UChar * string
+ * @param s const uint8_t * string

  * @param i string offset
  * @param c output UChar32 variable
  * @see U8_GET
  * @param i string offset
  * @param c output UChar32 variable
  * @see U8_GET

- * @draft ICU 2.4
+ * @stable ICU 2.4

  */
 #define U8_GET_UNSAFE(s, i, c) { \
  */
 #define U8_GET_UNSAFE(s, i, c) { \

-    int32_t __I=(int32_t)(i); \
-    U8_SET_CP_START_UNSAFE(s, __I); \
-    U8_NEXT_UNSAFE(s, __I, c); \
+    int32_t _u8_get_unsafe_index=(int32_t)(i); \
+    U8_SET_CP_START_UNSAFE(s, _u8_get_unsafe_index); \
+    U8_NEXT_UNSAFE(s, _u8_get_unsafe_index, c); \

 }
 
 /**
 }
 
 /**


@@ -174,22 +216,56 @@ utf8_back1SafeBody(const uint8_t *s, int32_t start, int32_t i);
  * 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 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.
  *
  * 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 UChar * string
- * @param start starting string offset
- * @param i string offset, start<=i<length
- * @param length string length
+ * @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 <0 in case of an error
  * @see U8_GET_UNSAFE
  * @param c output UChar32 variable, set to <0 in case of an error
  * @see U8_GET_UNSAFE

- * @draft ICU 2.4
+ * @stable ICU 2.4

  */
 #define U8_GET(s, start, i, length, c) { \
  */
 #define U8_GET(s, start, i, length, c) { \

-    int32_t __I=(int32_t)(i); \
-    U8_SET_CP_START(s, start, __I); \
-    U8_NEXT(s, __I, length, c); \
+    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 --------------------------------------- */
 }
 
 /* definitions with forward iteration --------------------------------------- */


@@ -205,27 +281,24 @@ utf8_back1SafeBody(const uint8_t *s, int32_t start, int32_t i);
  * The result is undefined if the offset points to a trail byte
  * or an illegal UTF-8 sequence.
  *
  * The result is undefined if the offset points to a trail byte
  * or an illegal UTF-8 sequence.
  *

- * @param s const UChar * string
+ * @param s const uint8_t * string

  * @param i string offset
  * @param c output UChar32 variable
  * @see U8_NEXT
  * @param i string offset
  * @param c output UChar32 variable
  * @see U8_NEXT

- * @draft ICU 2.4
+ * @stable ICU 2.4

  */
 #define U8_NEXT_UNSAFE(s, i, c) { \
  */
 #define U8_NEXT_UNSAFE(s, i, c) { \

-    (c)=(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; \
+    (c)=(uint8_t)(s)[(i)++]; \
+    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; \

         } \
     } \
 }
         } \
     } \
 }


@@ -236,25 +309,94 @@ utf8_back1SafeBody(const uint8_t *s, int32_t start, int32_t i);
  * (Post-incrementing forward iteration.)
  * "Safe" macro, checks for illegal sequences and for string boundaries.
  *
  * (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.
  *
  * 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 UChar * string
- * @param i string offset, i<length
- * @param length string length
+ * @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 <0 in case of an error
  * @see U8_NEXT_UNSAFE
  * @param c output UChar32 variable, set to <0 in case of an error
  * @see U8_NEXT_UNSAFE

- * @draft ICU 2.4
+ * @stable ICU 2.4

  */
 #define U8_NEXT(s, i, length, c) { \
  */
 #define U8_NEXT(s, i, length, c) { \

-    (c)=(s)[(i)++]; \
+    (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 { \
+            /* function call for "complicated" and error cases */ \
+            (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) { \
     if((c)>=0x80) { \

-        if(U8_IS_LEAD(c)) { \
-            (c)=utf8_nextCharSafeBody(s, &(i), (int32_t)(length), c, -1); \
+        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 { \
         } else { \

-            (c)=U_SENTINEL; \
+            /* function call for "complicated" and error cases */ \
+            (c)=utf8_nextCharSafeBody((const uint8_t *)s, &(i), (length), c, -3); \

         } \
     } \
 }
         } \
     } \
 }


@@ -266,11 +408,11 @@ utf8_back1SafeBody(const uint8_t *s, int32_t start, int32_t i);
  * "Unsafe" macro, assumes a valid code point and sufficient space in the string.
  * Otherwise, the result is undefined.
  *
  * "Unsafe" macro, assumes a valid code point and sufficient space in the string.
  * Otherwise, the result is undefined.
  *

- * @param s const UChar * string buffer
+ * @param s const uint8_t * string buffer

  * @param i string offset
  * @param c code point to append
  * @see U8_APPEND
  * @param i string offset
  * @param c code point to append
  * @see U8_APPEND

- * @draft ICU 2.4
+ * @stable ICU 2.4

  */
 #define U8_APPEND_UNSAFE(s, i, c) { \
     if((uint32_t)(c)<=0x7f) { \
  */
 #define U8_APPEND_UNSAFE(s, i, c) { \
     if((uint32_t)(c)<=0x7f) { \


@@ -292,7 +434,7 @@ utf8_back1SafeBody(const uint8_t *s, int32_t start, int32_t i);
 }
 
 /**
 }
 
 /**

- * Append a code point to a string, overwriting 1 or 2 code units.
+ * Append a code point to a string, overwriting 1 to 4 bytes.

  * The offset points to the current end of the string contents
  * and is advanced (post-increment).
  * "Safe" macro, checks for a valid code point.
  * The offset points to the current end of the string contents
  * and is advanced (post-increment).
  * "Safe" macro, checks for a valid code point.


@@ -300,19 +442,26 @@ utf8_back1SafeBody(const uint8_t *s, int32_t start, int32_t i);
  * If the code point is not valid or trail bytes do not fit,
  * then isError is set to TRUE.
  *
  * If the code point is not valid or trail bytes do not fit,
  * then isError is set to TRUE.
  *

- * @param s const UChar * string buffer
- * @param i string offset, i<length
- * @param length size of the string buffer
- * @param c code point to append
+ * @param s const uint8_t * string buffer
+ * @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
  * @param isError output UBool set to TRUE if an error occurs, otherwise not modified
  * @see U8_APPEND_UNSAFE

- * @draft ICU 2.4
+ * @stable ICU 2.4

  */
  */

-#define U8_APPEND(s, i, length, c, isError) { \
+#define U8_APPEND(s, i, capacity, c, isError) { \

     if((uint32_t)(c)<=0x7f) { \
         (s)[(i)++]=(uint8_t)(c); \
     if((uint32_t)(c)<=0x7f) { \
         (s)[(i)++]=(uint8_t)(c); \

+    } else if((uint32_t)(c)<=0x7ff && (i)+1<(capacity)) { \
+        (s)[(i)++]=(uint8_t)(((c)>>6)|0xc0); \
+        (s)[(i)++]=(uint8_t)(((c)&0x3f)|0x80); \
+    } else if((uint32_t)(c)<=0xd7ff && (i)+2<(capacity)) { \
+        (s)[(i)++]=(uint8_t)(((c)>>12)|0xe0); \
+        (s)[(i)++]=(uint8_t)((((c)>>6)&0x3f)|0x80); \
+        (s)[(i)++]=(uint8_t)(((c)&0x3f)|0x80); \

     } else { \
     } else { \

-        (i)=utf8_appendCharSafeBody(s, (int32_t)(i), (int32_t)(length), c, &(isError)); \
+        (i)=utf8_appendCharSafeBody(s, (i), (capacity), c, &(isError)); \

     } \
 }
 
     } \
 }
 


@@ -321,13 +470,13 @@ utf8_back1SafeBody(const uint8_t *s, int32_t start, int32_t i);
  * (Post-incrementing iteration.)
  * "Unsafe" macro, assumes well-formed UTF-8.
  *
  * (Post-incrementing iteration.)
  * "Unsafe" macro, assumes well-formed UTF-8.
  *

- * @param s const UChar * string
+ * @param s const uint8_t * string

  * @param i string offset
  * @see U8_FWD_1
  * @param i string offset
  * @see U8_FWD_1

- * @draft ICU 2.4
+ * @stable ICU 2.4

  */
 #define U8_FWD_1_UNSAFE(s, i) { \
  */
 #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]); \

 }
 
 /**
 }
 
 /**


@@ -335,17 +484,19 @@ utf8_back1SafeBody(const uint8_t *s, int32_t start, int32_t i);
  * (Post-incrementing iteration.)
  * "Safe" macro, checks for illegal sequences and for string boundaries.
  *
  * (Post-incrementing iteration.)
  * "Safe" macro, checks for illegal sequences and for string boundaries.
  *

- * @param s const UChar * string
- * @param i string offset, i<length
- * @param length string length
+ * The length can be negative for a NUL-terminated string.
+ *
+ * @param s const uint8_t * string
+ * @param i int32_t string offset, must be i<length
+ * @param length int32_t string length

  * @see U8_FWD_1_UNSAFE
  * @see U8_FWD_1_UNSAFE

- * @draft ICU 2.4
+ * @stable ICU 2.4

  */
 #define U8_FWD_1(s, i, length) { \
  */
 #define U8_FWD_1(s, i, length) { \

-    uint8_t __b=(s)[(i)++]; \
+    uint8_t __b=(uint8_t)(s)[(i)++]; \

     if(U8_IS_LEAD(__b)) { \
         uint8_t __count=U8_COUNT_TRAIL_BYTES(__b); \
     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])) { \
             __count=(uint8_t)((length)-(i)); \
         } \
         while(__count>0 && U8_IS_TRAIL((s)[i])) { \


@@ -361,11 +512,11 @@ utf8_back1SafeBody(const uint8_t *s, int32_t start, int32_t i);
  * (Post-incrementing iteration.)
  * "Unsafe" macro, assumes well-formed UTF-8.
  *
  * (Post-incrementing iteration.)
  * "Unsafe" macro, assumes well-formed UTF-8.
  *

- * @param s const UChar * string
+ * @param s const uint8_t * string

  * @param i string offset
  * @param n number of code points to skip
  * @see U8_FWD_N
  * @param i string offset
  * @param n number of code points to skip
  * @see U8_FWD_N

- * @draft ICU 2.4
+ * @stable ICU 2.4

  */
 #define U8_FWD_N_UNSAFE(s, i, n) { \
     int32_t __N=(n); \
  */
 #define U8_FWD_N_UNSAFE(s, i, n) { \
     int32_t __N=(n); \


@@ -381,16 +532,18 @@ utf8_back1SafeBody(const uint8_t *s, int32_t start, int32_t i);
  * (Post-incrementing iteration.)
  * "Safe" macro, checks for illegal sequences and for string boundaries.
  *
  * (Post-incrementing iteration.)
  * "Safe" macro, checks for illegal sequences and for string boundaries.
  *

- * @param s const UChar * string
- * @param i string offset, i<length
- * @param length string length
+ * The length can be negative for a NUL-terminated string.
+ *
+ * @param s const uint8_t * string
+ * @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
  * @param n number of code points to skip
  * @see U8_FWD_N_UNSAFE

- * @draft ICU 2.4
+ * @stable ICU 2.4

  */
 #define U8_FWD_N(s, i, length, n) { \
     int32_t __N=(n); \
  */
 #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; \
     } \
         U8_FWD_1(s, i, length); \
         --__N; \
     } \


@@ -404,10 +557,10 @@ utf8_back1SafeBody(const uint8_t *s, int32_t start, int32_t i);
  * Otherwise, it is not modified.
  * "Unsafe" macro, assumes well-formed UTF-8.
  *
  * Otherwise, it is not modified.
  * "Unsafe" macro, assumes well-formed UTF-8.
  *

- * @param s const UChar * string
+ * @param s const uint8_t * string

  * @param i string offset
  * @see U8_SET_CP_START
  * @param i string offset
  * @see U8_SET_CP_START

- * @draft ICU 2.4
+ * @stable ICU 2.4

  */
 #define U8_SET_CP_START_UNSAFE(s, i) { \
     while(U8_IS_TRAIL((s)[i])) { --(i); } \
  */
 #define U8_SET_CP_START_UNSAFE(s, i) { \
     while(U8_IS_TRAIL((s)[i])) { --(i); } \


@@ -421,15 +574,15 @@ utf8_back1SafeBody(const uint8_t *s, int32_t start, int32_t i);
  * Otherwise, it is not modified.
  * "Safe" macro, checks for illegal sequences and for string boundaries.
  *
  * Otherwise, it is not modified.
  * "Safe" macro, checks for illegal sequences and for string boundaries.
  *

- * @param s const UChar * string
- * @param start starting string offset (usually 0)
- * @param i string offset, start<=i
+ * @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

  * @see U8_SET_CP_START_UNSAFE
  * @see U8_SET_CP_START_UNSAFE

- * @draft ICU 2.4
+ * @stable ICU 2.4

  */
 #define U8_SET_CP_START(s, start, i) { \
     if(U8_IS_TRAIL((s)[(i)])) { \
  */
 #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)); \

     } \
 }
 
     } \
 }
 


@@ -448,21 +601,21 @@ utf8_back1SafeBody(const uint8_t *s, int32_t start, int32_t i);
  * will be returned as the code point.
  * The result is undefined if the offset is behind an illegal UTF-8 sequence.
  *
  * will be returned as the code point.
  * The result is undefined if the offset is behind an illegal UTF-8 sequence.
  *

- * @param s const UChar * string
+ * @param s const uint8_t * string

  * @param i string offset
  * @param c output UChar32 variable
  * @see U8_PREV
  * @param i string offset
  * @param c output UChar32 variable
  * @see U8_PREV

- * @draft ICU 2.4
+ * @stable ICU 2.4

  */
 #define U8_PREV_UNSAFE(s, i, c) { \
  */
 #define U8_PREV_UNSAFE(s, i, c) { \

-    (c)=(s)[--(i)]; \
+    (c)=(uint8_t)(s)[--(i)]; \

     if(U8_IS_TRAIL(c)) { \
         uint8_t __b, __count=1, __shift=6; \
 \
         /* c is a trail byte */ \
         (c)&=0x3f; \
         for(;;) { \
     if(U8_IS_TRAIL(c)) { \
         uint8_t __b, __count=1, __shift=6; \
 \
         /* c is a trail byte */ \
         (c)&=0x3f; \
         for(;;) { \

-            __b=(s)[--(i)]; \
+            __b=(uint8_t)(s)[--(i)]; \

             if(__b>=0xc0) { \
                 U8_MASK_LEAD_BYTE(__b, __count); \
                 (c)|=(UChar32)__b<<__shift; \
             if(__b>=0xc0) { \
                 U8_MASK_LEAD_BYTE(__b, __count); \
                 (c)|=(UChar32)__b<<__shift; \


@@ -489,21 +642,48 @@ utf8_back1SafeBody(const uint8_t *s, int32_t start, int32_t i);
  * will be returned as the code point.
  * If the offset is behind an illegal UTF-8 sequence, then c is set to a negative value.
  *
  * will be returned as the code point.
  * If the offset is behind an illegal UTF-8 sequence, then c is set to a negative value.
  *

- * @param s const UChar * string
- * @param start starting string offset (usually 0)
- * @param i string offset, start<=i
+ * @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 <0 in case of an error
  * @see U8_PREV_UNSAFE
  * @param c output UChar32 variable, set to <0 in case of an error
  * @see U8_PREV_UNSAFE

- * @draft ICU 2.4
+ * @stable ICU 2.4

  */
 #define U8_PREV(s, start, i, c) { \
  */
 #define U8_PREV(s, start, i, c) { \

-    (c)=(s)[--(i)]; \
+    (c)=(uint8_t)(s)[--(i)]; \

     if((c)>=0x80) { \
     if((c)>=0x80) { \

-        if((c)<=0xbf) { \
-            (c)=utf8_prevCharSafeBody(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); \

     } \
 }
 
     } \
 }
 


@@ -513,10 +693,10 @@ utf8_back1SafeBody(const uint8_t *s, int32_t start, int32_t i);
  * The input offset may be the same as the string length.
  * "Unsafe" macro, assumes well-formed UTF-8.
  *
  * The input offset may be the same as the string length.
  * "Unsafe" macro, assumes well-formed UTF-8.
  *

- * @param s const UChar * string
+ * @param s const uint8_t * string

  * @param i string offset
  * @see U8_BACK_1
  * @param i string offset
  * @see U8_BACK_1

- * @draft ICU 2.4
+ * @stable ICU 2.4

  */
 #define U8_BACK_1_UNSAFE(s, i) { \
     while(U8_IS_TRAIL((s)[--(i)])) {} \
  */
 #define U8_BACK_1_UNSAFE(s, i) { \
     while(U8_IS_TRAIL((s)[--(i)])) {} \


@@ -528,15 +708,15 @@ utf8_back1SafeBody(const uint8_t *s, int32_t start, int32_t i);
  * The input offset may be the same as the string length.
  * "Safe" macro, checks for illegal sequences and for string boundaries.
  *
  * The input offset may be the same as the string length.
  * "Safe" macro, checks for illegal sequences and for string boundaries.
  *

- * @param s const UChar * string
- * @param start starting string offset (usually 0)
- * @param i string offset, start<=i
+ * @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

  * @see U8_BACK_1_UNSAFE
  * @see U8_BACK_1_UNSAFE

- * @draft ICU 2.4
+ * @stable ICU 2.4

  */
 #define U8_BACK_1(s, start, i) { \
     if(U8_IS_TRAIL((s)[--(i)])) { \
  */
 #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)); \

     } \
 }
 
     } \
 }
 


@@ -547,11 +727,11 @@ utf8_back1SafeBody(const uint8_t *s, int32_t start, int32_t i);
  * The input offset may be the same as the string length.
  * "Unsafe" macro, assumes well-formed UTF-8.
  *
  * The input offset may be the same as the string length.
  * "Unsafe" macro, assumes well-formed UTF-8.
  *

- * @param s const UChar * string
+ * @param s const uint8_t * string

  * @param i string offset
  * @param n number of code points to skip
  * @see U8_BACK_N
  * @param i string offset
  * @param n number of code points to skip
  * @see U8_BACK_N

- * @draft ICU 2.4
+ * @stable ICU 2.4

  */
 #define U8_BACK_N_UNSAFE(s, i, n) { \
     int32_t __N=(n); \
  */
 #define U8_BACK_N_UNSAFE(s, i, n) { \
     int32_t __N=(n); \


@@ -568,12 +748,12 @@ utf8_back1SafeBody(const uint8_t *s, int32_t start, int32_t i);
  * The input offset may be the same as the string length.
  * "Safe" macro, checks for illegal sequences and for string boundaries.
  *
  * The input offset may be the same as the string length.
  * "Safe" macro, checks for illegal sequences and for string boundaries.
  *

- * @param s const UChar * string
- * @param start index of the start of the string
- * @param i string offset, i<length
+ * @param s const uint8_t * string
+ * @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
  * @param n number of code points to skip
  * @see U8_BACK_N_UNSAFE

- * @draft ICU 2.4
+ * @stable ICU 2.4

  */
 #define U8_BACK_N(s, start, i, n) { \
     int32_t __N=(n); \
  */
 #define U8_BACK_N(s, start, i, n) { \
     int32_t __N=(n); \


@@ -591,10 +771,10 @@ utf8_back1SafeBody(const uint8_t *s, int32_t start, int32_t i);
  * The input offset may be the same as the string length.
  * "Unsafe" macro, assumes well-formed UTF-8.
  *
  * The input offset may be the same as the string length.
  * "Unsafe" macro, assumes well-formed UTF-8.
  *

- * @param s const UChar * string
+ * @param s const uint8_t * string

  * @param i string offset
  * @see U8_SET_CP_LIMIT
  * @param i string offset
  * @see U8_SET_CP_LIMIT

- * @draft ICU 2.4
+ * @stable ICU 2.4

  */
 #define U8_SET_CP_LIMIT_UNSAFE(s, i) { \
     U8_BACK_1_UNSAFE(s, i); \
  */
 #define U8_SET_CP_LIMIT_UNSAFE(s, i) { \
     U8_BACK_1_UNSAFE(s, i); \


@@ -609,15 +789,17 @@ utf8_back1SafeBody(const uint8_t *s, int32_t start, int32_t i);
  * The input offset may be the same as the string length.
  * "Safe" macro, checks for illegal sequences and for string boundaries.
  *
  * The input offset may be the same as the string length.
  * "Safe" macro, checks for illegal sequences and for string boundaries.
  *

- * @param s const UChar * string
- * @param start starting string offset (usually 0)
- * @param i string offset, start<=i<=length
- * @param length string length
+ * The length can be negative for a NUL-terminated string.
+ *
+ * @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<=length
+ * @param length int32_t string length

  * @see U8_SET_CP_LIMIT_UNSAFE
  * @see U8_SET_CP_LIMIT_UNSAFE

- * @draft ICU 2.4
+ * @stable ICU 2.4

  */
 #define U8_SET_CP_LIMIT(s, start, i, length) { \
  */
 #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); \
     } \
         U8_BACK_1(s, start, i); \
         U8_FWD_1(s, i, length); \
     } \