michael@0: /*
michael@0:  * Copyright 2001-2004 Unicode, Inc.
michael@0:  *
michael@0:  * Disclaimer
michael@0:  *
michael@0:  * This source code is provided as is by Unicode, Inc. No claims are
michael@0:  * made as to fitness for any particular purpose. No warranties of any
michael@0:  * kind are expressed or implied. The recipient agrees to determine
michael@0:  * applicability of information provided. If this file has been
michael@0:  * purchased on magnetic or optical media from Unicode, Inc., the
michael@0:  * sole remedy for any claim will be exchange of defective media
michael@0:  * within 90 days of receipt.
michael@0:  *
michael@0:  * Limitations on Rights to Redistribute This Code
michael@0:  *
michael@0:  * Unicode, Inc. hereby grants the right to freely use the information
michael@0:  * supplied in this file in the creation of products supporting the
michael@0:  * Unicode Standard, and to make copies of this file in any form
michael@0:  * for internal or external distribution as long as this notice
michael@0:  * remains attached.
michael@0:  */
michael@0: 
michael@0: /* ---------------------------------------------------------------------
michael@0: 
michael@0: Conversions between UTF32, UTF-16, and UTF-8. Source code file.
michael@0: Author: Mark E. Davis, 1994.
michael@0: Rev History: Rick McGowan, fixes & updates May 2001.
michael@0: Sept 2001: fixed const & error conditions per
michael@0: mods suggested by S. Parent & A. Lillich.
michael@0: June 2002: Tim Dodd added detection and handling of incomplete
michael@0: source sequences, enhanced error detection, added casts
michael@0: to eliminate compiler warnings.
michael@0: July 2003: slight mods to back out aggressive FFFE detection.
michael@0: Jan 2004: updated switches in from-UTF8 conversions.
michael@0: Oct 2004: updated to use UNI_MAX_LEGAL_UTF32 in UTF-32 conversions.
michael@0: 
michael@0: See the header file "ConvertUTF.h" for complete documentation.
michael@0: 
michael@0: ------------------------------------------------------------------------ */
michael@0: 
michael@0: 
michael@0: #include "convert_UTF.h"
michael@0: #ifdef CVTUTF_DEBUG
michael@0: #include <stdio.h>
michael@0: #endif
michael@0: 
michael@0: static const int halfShift  = 10; /* used for shifting by 10 bits */
michael@0: 
michael@0: static const UTF32 halfBase = 0x0010000UL;
michael@0: static const UTF32 halfMask = 0x3FFUL;
michael@0: 
michael@0: #define UNI_SUR_HIGH_START  (UTF32)0xD800
michael@0: #define UNI_SUR_HIGH_END    (UTF32)0xDBFF
michael@0: #define UNI_SUR_LOW_START   (UTF32)0xDC00
michael@0: #define UNI_SUR_LOW_END     (UTF32)0xDFFF
michael@0: #define false	   0
michael@0: #define true	    1
michael@0: 
michael@0: /* --------------------------------------------------------------------- */
michael@0: 
michael@0: ConversionResult ConvertUTF32toUTF16 (const UTF32** sourceStart, const UTF32* sourceEnd,
michael@0:                                       UTF16** targetStart, UTF16* targetEnd, ConversionFlags flags) {
michael@0:   ConversionResult result = conversionOK;
michael@0:   const UTF32* source = *sourceStart;
michael@0:   UTF16* target = *targetStart;
michael@0:   while (source < sourceEnd) {
michael@0:     UTF32 ch;
michael@0:     if (target >= targetEnd) {
michael@0: 	    result = targetExhausted; break;
michael@0:     }
michael@0:     ch = *source++;
michael@0:     if (ch <= UNI_MAX_BMP) { /* Target is a character <= 0xFFFF */
michael@0: 	    /* UTF-16 surrogate values are illegal in UTF-32; 0xffff or 0xfffe are both reserved values */
michael@0: 	    if (ch >= UNI_SUR_HIGH_START && ch <= UNI_SUR_LOW_END) {
michael@0:         if (flags == strictConversion) {
michael@0:           --source; /* return to the illegal value itself */
michael@0:           result = sourceIllegal;
michael@0:           break;
michael@0:         } else {
michael@0:           *target++ = UNI_REPLACEMENT_CHAR;
michael@0:         }
michael@0: 	    } else {
michael@0:         *target++ = (UTF16)ch; /* normal case */
michael@0: 	    }
michael@0:     } else if (ch > UNI_MAX_LEGAL_UTF32) {
michael@0: 	    if (flags == strictConversion) {
michael@0:         result = sourceIllegal;
michael@0: 	    } else {
michael@0:         *target++ = UNI_REPLACEMENT_CHAR;
michael@0: 	    }
michael@0:     } else {
michael@0: 	    /* target is a character in range 0xFFFF - 0x10FFFF. */
michael@0: 	    if (target + 1 >= targetEnd) {
michael@0:         --source; /* Back up source pointer! */
michael@0:         result = targetExhausted; break;
michael@0: 	    }
michael@0: 	    ch -= halfBase;
michael@0: 	    *target++ = (UTF16)((ch >> halfShift) + UNI_SUR_HIGH_START);
michael@0: 	    *target++ = (UTF16)((ch & halfMask) + UNI_SUR_LOW_START);
michael@0:     }
michael@0:   }
michael@0: *sourceStart = source;
michael@0: *targetStart = target;
michael@0: return result;
michael@0: }
michael@0: 
michael@0: /* --------------------------------------------------------------------- */
michael@0: 
michael@0: ConversionResult ConvertUTF16toUTF32 (const UTF16** sourceStart, const UTF16* sourceEnd,
michael@0:                                       UTF32** targetStart, UTF32* targetEnd, ConversionFlags flags) {
michael@0:   ConversionResult result = conversionOK;
michael@0:   const UTF16* source = *sourceStart;
michael@0:   UTF32* target = *targetStart;
michael@0:   UTF32 ch, ch2;
michael@0:   while (source < sourceEnd) {
michael@0:     const UTF16* oldSource = source; /*  In case we have to back up because of target overflow. */
michael@0:     ch = *source++;
michael@0:     /* If we have a surrogate pair, convert to UTF32 first. */
michael@0:     if (ch >= UNI_SUR_HIGH_START && ch <= UNI_SUR_HIGH_END) {
michael@0: 	    /* If the 16 bits following the high surrogate are in the source buffer... */
michael@0: 	    if (source < sourceEnd) {
michael@0:         ch2 = *source;
michael@0:         /* If it's a low surrogate, convert to UTF32. */
michael@0:         if (ch2 >= UNI_SUR_LOW_START && ch2 <= UNI_SUR_LOW_END) {
michael@0:           ch = ((ch - UNI_SUR_HIGH_START) << halfShift)
michael@0:           + (ch2 - UNI_SUR_LOW_START) + halfBase;
michael@0:           ++source;
michael@0:         } else if (flags == strictConversion) { /* it's an unpaired high surrogate */
michael@0:           --source; /* return to the illegal value itself */
michael@0:           result = sourceIllegal;
michael@0:           break;
michael@0:         }
michael@0: 	    } else { /* We don't have the 16 bits following the high surrogate. */
michael@0:         --source; /* return to the high surrogate */
michael@0:         result = sourceExhausted;
michael@0:         break;
michael@0: 	    }
michael@0:     } else if (flags == strictConversion) {
michael@0: 	    /* UTF-16 surrogate values are illegal in UTF-32 */
michael@0: 	    if (ch >= UNI_SUR_LOW_START && ch <= UNI_SUR_LOW_END) {
michael@0:         --source; /* return to the illegal value itself */
michael@0:         result = sourceIllegal;
michael@0:         break;
michael@0: 	    }
michael@0:     }
michael@0:     if (target >= targetEnd) {
michael@0: 	    source = oldSource; /* Back up source pointer! */
michael@0: 	    result = targetExhausted; break;
michael@0:     }
michael@0:     *target++ = ch;
michael@0:   }
michael@0:   *sourceStart = source;
michael@0:   *targetStart = target;
michael@0: #ifdef CVTUTF_DEBUG
michael@0:   if (result == sourceIllegal) {
michael@0:     fprintf(stderr, "ConvertUTF16toUTF32 illegal seq 0x%04x,%04x\n", ch, ch2);
michael@0:     fflush(stderr);
michael@0:   }
michael@0: #endif
michael@0:   return result;
michael@0: }
michael@0: 
michael@0: /* --------------------------------------------------------------------- */
michael@0: 
michael@0: /*
michael@0:  * Index into the table below with the first byte of a UTF-8 sequence to
michael@0:  * get the number of trailing bytes that are supposed to follow it.
michael@0:  * Note that *legal* UTF-8 values can't have 4 or 5-bytes. The table is
michael@0:  * left as-is for anyone who may want to do such conversion, which was
michael@0:  * allowed in earlier algorithms.
michael@0:  */
michael@0: static const char trailingBytesForUTF8[256] = {
michael@0:   0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
michael@0:   0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
michael@0:   0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
michael@0:   0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
michael@0:   0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
michael@0:   0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
michael@0:   1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,
michael@0:   2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2, 3,3,3,3,3,3,3,3,4,4,4,4,5,5,5,5
michael@0: };
michael@0: 
michael@0: /*
michael@0:  * Magic values subtracted from a buffer value during UTF8 conversion.
michael@0:  * This table contains as many values as there might be trailing bytes
michael@0:  * in a UTF-8 sequence.
michael@0:  */
michael@0: static const UTF32 offsetsFromUTF8[6] = { 0x00000000UL, 0x00003080UL, 0x000E2080UL,
michael@0:   0x03C82080UL, 0xFA082080UL, 0x82082080UL };
michael@0: 
michael@0: /*
michael@0:  * Once the bits are split out into bytes of UTF-8, this is a mask OR-ed
michael@0:  * into the first byte, depending on how many bytes follow.  There are
michael@0:  * as many entries in this table as there are UTF-8 sequence types.
michael@0:  * (I.e., one byte sequence, two byte... etc.). Remember that sequencs
michael@0:  * for *legal* UTF-8 will be 4 or fewer bytes total.
michael@0:  */
michael@0: static const UTF8 firstByteMark[7] = { 0x00, 0x00, 0xC0, 0xE0, 0xF0, 0xF8, 0xFC };
michael@0: 
michael@0: /* --------------------------------------------------------------------- */
michael@0: 
michael@0: /* The interface converts a whole buffer to avoid function-call overhead.
michael@0: * Constants have been gathered. Loops & conditionals have been removed as
michael@0: * much as possible for efficiency, in favor of drop-through switches.
michael@0: * (See "Note A" at the bottom of the file for equivalent code.)
michael@0: * If your compiler supports it, the "isLegalUTF8" call can be turned
michael@0: * into an inline function.
michael@0: */
michael@0: 
michael@0: /* --------------------------------------------------------------------- */
michael@0: 
michael@0: ConversionResult ConvertUTF16toUTF8 (const UTF16** sourceStart, const UTF16* sourceEnd,
michael@0:                                      UTF8** targetStart, UTF8* targetEnd, ConversionFlags flags) {
michael@0:   ConversionResult result = conversionOK;
michael@0:   const UTF16* source = *sourceStart;
michael@0:   UTF8* target = *targetStart;
michael@0:   while (source < sourceEnd) {
michael@0:     UTF32 ch;
michael@0:     unsigned short bytesToWrite = 0;
michael@0:     const UTF32 byteMask = 0xBF;
michael@0:     const UTF32 byteMark = 0x80;
michael@0:     const UTF16* oldSource = source; /* In case we have to back up because of target overflow. */
michael@0:     ch = *source++;
michael@0:     /* If we have a surrogate pair, convert to UTF32 first. */
michael@0:     if (ch >= UNI_SUR_HIGH_START && ch <= UNI_SUR_HIGH_END) {
michael@0: 	    /* If the 16 bits following the high surrogate are in the source buffer... */
michael@0: 	    if (source < sourceEnd) {
michael@0:         UTF32 ch2 = *source;
michael@0:         /* If it's a low surrogate, convert to UTF32. */
michael@0:         if (ch2 >= UNI_SUR_LOW_START && ch2 <= UNI_SUR_LOW_END) {
michael@0:           ch = ((ch - UNI_SUR_HIGH_START) << halfShift)
michael@0:           + (ch2 - UNI_SUR_LOW_START) + halfBase;
michael@0:           ++source;
michael@0:         } else if (flags == strictConversion) { /* it's an unpaired high surrogate */
michael@0:           --source; /* return to the illegal value itself */
michael@0:           result = sourceIllegal;
michael@0:           break;
michael@0:         }
michael@0: 	    } else { /* We don't have the 16 bits following the high surrogate. */
michael@0:         --source; /* return to the high surrogate */
michael@0:         result = sourceExhausted;
michael@0:         break;
michael@0: 	    }
michael@0:     } else if (flags == strictConversion) {
michael@0: 	    /* UTF-16 surrogate values are illegal in UTF-32 */
michael@0: 	    if (ch >= UNI_SUR_LOW_START && ch <= UNI_SUR_LOW_END) {
michael@0:         --source; /* return to the illegal value itself */
michael@0:         result = sourceIllegal;
michael@0:         break;
michael@0: 	    }
michael@0:     }
michael@0:     /* Figure out how many bytes the result will require */
michael@0:     if (ch < (UTF32)0x80) {	     bytesToWrite = 1;
michael@0:     } else if (ch < (UTF32)0x800) {     bytesToWrite = 2;
michael@0:     } else if (ch < (UTF32)0x10000) {   bytesToWrite = 3;
michael@0:     } else if (ch < (UTF32)0x110000) {  bytesToWrite = 4;
michael@0:     } else {			    bytesToWrite = 3;
michael@0:       ch = UNI_REPLACEMENT_CHAR;
michael@0:     }
michael@0: 
michael@0:     target += bytesToWrite;
michael@0:     if (target > targetEnd) {
michael@0: 	    source = oldSource; /* Back up source pointer! */
michael@0: 	    target -= bytesToWrite; result = targetExhausted; break;
michael@0:     }
michael@0:     switch (bytesToWrite) { /* note: everything falls through. */
michael@0: 	    case 4: *--target = (UTF8)((ch | byteMark) & byteMask); ch >>= 6;
michael@0: 	    case 3: *--target = (UTF8)((ch | byteMark) & byteMask); ch >>= 6;
michael@0: 	    case 2: *--target = (UTF8)((ch | byteMark) & byteMask); ch >>= 6;
michael@0: 	    case 1: *--target =  (UTF8)(ch | firstByteMark[bytesToWrite]);
michael@0:     }
michael@0:     target += bytesToWrite;
michael@0:   }
michael@0: *sourceStart = source;
michael@0: *targetStart = target;
michael@0: return result;
michael@0: }
michael@0: 
michael@0: /* --------------------------------------------------------------------- */
michael@0: 
michael@0: /*
michael@0:  * Utility routine to tell whether a sequence of bytes is legal UTF-8.
michael@0:  * This must be called with the length pre-determined by the first byte.
michael@0:  * If not calling this from ConvertUTF8to*, then the length can be set by:
michael@0:  *  length = trailingBytesForUTF8[*source]+1;
michael@0:  * and the sequence is illegal right away if there aren't that many bytes
michael@0:  * available.
michael@0:  * If presented with a length > 4, this returns false.  The Unicode
michael@0:  * definition of UTF-8 goes up to 4-byte sequences.
michael@0:  */
michael@0: 
michael@0: static Boolean isLegalUTF8(const UTF8 *source, int length) {
michael@0:   UTF8 a;
michael@0:   const UTF8 *srcptr = source+length;
michael@0:   switch (length) {
michael@0:     default: return false;
michael@0:       /* Everything else falls through when "true"... */
michael@0:     case 4: if ((a = (*--srcptr)) < 0x80 || a > 0xBF) return false;
michael@0:     case 3: if ((a = (*--srcptr)) < 0x80 || a > 0xBF) return false;
michael@0:     case 2: if ((a = (*--srcptr)) > 0xBF) return false;
michael@0: 
michael@0:       switch (*source) {
michael@0:         /* no fall-through in this inner switch */
michael@0:         case 0xE0: if (a < 0xA0) return false; break;
michael@0:         case 0xED: if (a > 0x9F) return false; break;
michael@0:         case 0xF0: if (a < 0x90) return false; break;
michael@0:         case 0xF4: if (a > 0x8F) return false; break;
michael@0:         default:   if (a < 0x80) return false;
michael@0:       }
michael@0: 
michael@0:       case 1: if (*source >= 0x80 && *source < 0xC2) return false;
michael@0:   }
michael@0:   if (*source > 0xF4) return false;
michael@0:   return true;
michael@0: }
michael@0: 
michael@0: /* --------------------------------------------------------------------- */
michael@0: 
michael@0: /*
michael@0:  * Exported function to return whether a UTF-8 sequence is legal or not.
michael@0:  * This is not used here; it's just exported.
michael@0:  */
michael@0: Boolean isLegalUTF8Sequence(const UTF8 *source, const UTF8 *sourceEnd) {
michael@0:   int length = trailingBytesForUTF8[*source]+1;
michael@0:   if (source+length > sourceEnd) {
michael@0:     return false;
michael@0:   }
michael@0:   return isLegalUTF8(source, length);
michael@0: }
michael@0: 
michael@0: /* --------------------------------------------------------------------- */
michael@0: 
michael@0: ConversionResult ConvertUTF8toUTF16 (const UTF8** sourceStart, const UTF8* sourceEnd,
michael@0:                                      UTF16** targetStart, UTF16* targetEnd, ConversionFlags flags) {
michael@0:   ConversionResult result = conversionOK;
michael@0:   const UTF8* source = *sourceStart;
michael@0:   UTF16* target = *targetStart;
michael@0:   while (source < sourceEnd) {
michael@0:     UTF32 ch = 0;
michael@0:     unsigned short extraBytesToRead = trailingBytesForUTF8[*source];
michael@0:     if (source + extraBytesToRead >= sourceEnd) {
michael@0: 	    result = sourceExhausted; break;
michael@0:     }
michael@0:     /* Do this check whether lenient or strict */
michael@0:     if (! isLegalUTF8(source, extraBytesToRead+1)) {
michael@0: 	    result = sourceIllegal;
michael@0: 	    break;
michael@0:     }
michael@0:     /*
michael@0:      * The cases all fall through. See "Note A" below.
michael@0:      */
michael@0:     switch (extraBytesToRead) {
michael@0: 	    case 5: ch += *source++; ch <<= 6; /* remember, illegal UTF-8 */
michael@0: 	    case 4: ch += *source++; ch <<= 6; /* remember, illegal UTF-8 */
michael@0: 	    case 3: ch += *source++; ch <<= 6;
michael@0: 	    case 2: ch += *source++; ch <<= 6;
michael@0: 	    case 1: ch += *source++; ch <<= 6;
michael@0: 	    case 0: ch += *source++;
michael@0:     }
michael@0:     ch -= offsetsFromUTF8[extraBytesToRead];
michael@0: 
michael@0:     if (target >= targetEnd) {
michael@0: 	    source -= (extraBytesToRead+1); /* Back up source pointer! */
michael@0: 	    result = targetExhausted; break;
michael@0:     }
michael@0:     if (ch <= UNI_MAX_BMP) { /* Target is a character <= 0xFFFF */
michael@0: 	    /* UTF-16 surrogate values are illegal in UTF-32 */
michael@0: 	    if (ch >= UNI_SUR_HIGH_START && ch <= UNI_SUR_LOW_END) {
michael@0:         if (flags == strictConversion) {
michael@0:           source -= (extraBytesToRead+1); /* return to the illegal value itself */
michael@0:           result = sourceIllegal;
michael@0:           break;
michael@0:         } else {
michael@0:           *target++ = UNI_REPLACEMENT_CHAR;
michael@0:         }
michael@0: 	    } else {
michael@0:         *target++ = (UTF16)ch; /* normal case */
michael@0: 	    }
michael@0:     } else if (ch > UNI_MAX_UTF16) {
michael@0: 	    if (flags == strictConversion) {
michael@0:         result = sourceIllegal;
michael@0:         source -= (extraBytesToRead+1); /* return to the start */
michael@0:         break; /* Bail out; shouldn't continue */
michael@0: 	    } else {
michael@0:         *target++ = UNI_REPLACEMENT_CHAR;
michael@0: 	    }
michael@0:     } else {
michael@0: 	    /* target is a character in range 0xFFFF - 0x10FFFF. */
michael@0: 	    if (target + 1 >= targetEnd) {
michael@0:         source -= (extraBytesToRead+1); /* Back up source pointer! */
michael@0:         result = targetExhausted; break;
michael@0: 	    }
michael@0: 	    ch -= halfBase;
michael@0: 	    *target++ = (UTF16)((ch >> halfShift) + UNI_SUR_HIGH_START);
michael@0: 	    *target++ = (UTF16)((ch & halfMask) + UNI_SUR_LOW_START);
michael@0:     }
michael@0:   }
michael@0: *sourceStart = source;
michael@0: *targetStart = target;
michael@0: return result;
michael@0: }
michael@0: 
michael@0: /* --------------------------------------------------------------------- */
michael@0: 
michael@0: ConversionResult ConvertUTF32toUTF8 (const UTF32** sourceStart, const UTF32* sourceEnd,
michael@0:                                      UTF8** targetStart, UTF8* targetEnd, ConversionFlags flags) {
michael@0:   ConversionResult result = conversionOK;
michael@0:   const UTF32* source = *sourceStart;
michael@0:   UTF8* target = *targetStart;
michael@0:   while (source < sourceEnd) {
michael@0:     UTF32 ch;
michael@0:     unsigned short bytesToWrite = 0;
michael@0:     const UTF32 byteMask = 0xBF;
michael@0:     const UTF32 byteMark = 0x80;
michael@0:     ch = *source++;
michael@0:     if (flags == strictConversion ) {
michael@0: 	    /* UTF-16 surrogate values are illegal in UTF-32 */
michael@0: 	    if (ch >= UNI_SUR_HIGH_START && ch <= UNI_SUR_LOW_END) {
michael@0:         --source; /* return to the illegal value itself */
michael@0:         result = sourceIllegal;
michael@0:         break;
michael@0: 	    }
michael@0:     }
michael@0:     /*
michael@0:      * Figure out how many bytes the result will require. Turn any
michael@0:      * illegally large UTF32 things (> Plane 17) into replacement chars.
michael@0:      */
michael@0:     if (ch < (UTF32)0x80) {	     bytesToWrite = 1;
michael@0:     } else if (ch < (UTF32)0x800) {     bytesToWrite = 2;
michael@0:     } else if (ch < (UTF32)0x10000) {   bytesToWrite = 3;
michael@0:     } else if (ch <= UNI_MAX_LEGAL_UTF32) {  bytesToWrite = 4;
michael@0:     } else {			    bytesToWrite = 3;
michael@0:       ch = UNI_REPLACEMENT_CHAR;
michael@0:       result = sourceIllegal;
michael@0:     }
michael@0: 
michael@0:     target += bytesToWrite;
michael@0:     if (target > targetEnd) {
michael@0: 	    --source; /* Back up source pointer! */
michael@0: 	    target -= bytesToWrite; result = targetExhausted; break;
michael@0:     }
michael@0:     switch (bytesToWrite) { /* note: everything falls through. */
michael@0: 	    case 4: *--target = (UTF8)((ch | byteMark) & byteMask); ch >>= 6;
michael@0: 	    case 3: *--target = (UTF8)((ch | byteMark) & byteMask); ch >>= 6;
michael@0: 	    case 2: *--target = (UTF8)((ch | byteMark) & byteMask); ch >>= 6;
michael@0: 	    case 1: *--target = (UTF8) (ch | firstByteMark[bytesToWrite]);
michael@0:     }
michael@0:     target += bytesToWrite;
michael@0:   }
michael@0: *sourceStart = source;
michael@0: *targetStart = target;
michael@0: return result;
michael@0: }
michael@0: 
michael@0: /* --------------------------------------------------------------------- */
michael@0: 
michael@0: ConversionResult ConvertUTF8toUTF32 (const UTF8** sourceStart, const UTF8* sourceEnd,
michael@0:                                      UTF32** targetStart, UTF32* targetEnd, ConversionFlags flags) {
michael@0:   ConversionResult result = conversionOK;
michael@0:   const UTF8* source = *sourceStart;
michael@0:   UTF32* target = *targetStart;
michael@0:   while (source < sourceEnd) {
michael@0:     UTF32 ch = 0;
michael@0:     unsigned short extraBytesToRead = trailingBytesForUTF8[*source];
michael@0:     if (source + extraBytesToRead >= sourceEnd) {
michael@0: 	    result = sourceExhausted; break;
michael@0:     }
michael@0:     /* Do this check whether lenient or strict */
michael@0:     if (! isLegalUTF8(source, extraBytesToRead+1)) {
michael@0: 	    result = sourceIllegal;
michael@0: 	    break;
michael@0:     }
michael@0:     /*
michael@0:      * The cases all fall through. See "Note A" below.
michael@0:      */
michael@0:     switch (extraBytesToRead) {
michael@0: 	    case 5: ch += *source++; ch <<= 6;
michael@0: 	    case 4: ch += *source++; ch <<= 6;
michael@0: 	    case 3: ch += *source++; ch <<= 6;
michael@0: 	    case 2: ch += *source++; ch <<= 6;
michael@0: 	    case 1: ch += *source++; ch <<= 6;
michael@0: 	    case 0: ch += *source++;
michael@0:     }
michael@0:     ch -= offsetsFromUTF8[extraBytesToRead];
michael@0: 
michael@0:     if (target >= targetEnd) {
michael@0: 	    source -= (extraBytesToRead+1); /* Back up the source pointer! */
michael@0: 	    result = targetExhausted; break;
michael@0:     }
michael@0:     if (ch <= UNI_MAX_LEGAL_UTF32) {
michael@0: 	    /*
michael@0: 	     * UTF-16 surrogate values are illegal in UTF-32, and anything
michael@0: 	     * over Plane 17 (> 0x10FFFF) is illegal.
michael@0: 	     */
michael@0: 	    if (ch >= UNI_SUR_HIGH_START && ch <= UNI_SUR_LOW_END) {
michael@0:         if (flags == strictConversion) {
michael@0:           source -= (extraBytesToRead+1); /* return to the illegal value itself */
michael@0:           result = sourceIllegal;
michael@0:           break;
michael@0:         } else {
michael@0:           *target++ = UNI_REPLACEMENT_CHAR;
michael@0:         }
michael@0: 	    } else {
michael@0:         *target++ = ch;
michael@0: 	    }
michael@0:     } else { /* i.e., ch > UNI_MAX_LEGAL_UTF32 */
michael@0: 	    result = sourceIllegal;
michael@0: 	    *target++ = UNI_REPLACEMENT_CHAR;
michael@0:     }
michael@0:   }
michael@0:   *sourceStart = source;
michael@0:   *targetStart = target;
michael@0:   return result;
michael@0: }
michael@0: 
michael@0: /* ---------------------------------------------------------------------
michael@0: 
michael@0: Note A.
michael@0: The fall-through switches in UTF-8 reading code save a
michael@0: temp variable, some decrements & conditionals.  The switches
michael@0: are equivalent to the following loop:
michael@0: {
michael@0:   int tmpBytesToRead = extraBytesToRead+1;
michael@0:   do {
michael@0: 		ch += *source++;
michael@0: 		--tmpBytesToRead;
michael@0: 		if (tmpBytesToRead) ch <<= 6;
michael@0:   } while (tmpBytesToRead > 0);
michael@0: }
michael@0: In UTF-8 writing code, the switches on "bytesToWrite" are
michael@0: similarly unrolled loops.
michael@0: 
michael@0: --------------------------------------------------------------------- */