michael@0: 
michael@0: /*
michael@0:  * Copyright 2006 The Android Open Source Project
michael@0:  *
michael@0:  * Use of this source code is governed by a BSD-style license that can be
michael@0:  * found in the LICENSE file.
michael@0:  */
michael@0: 
michael@0: 
michael@0: #include "SkUtils.h"
michael@0: 
michael@0: #if 0
michael@0: #define assign_16_longs(dst, value)             \
michael@0:     do {                                        \
michael@0:         (dst)[0] = value;   (dst)[1] = value;   \
michael@0:         (dst)[2] = value;   (dst)[3] = value;   \
michael@0:         (dst)[4] = value;   (dst)[5] = value;   \
michael@0:         (dst)[6] = value;   (dst)[7] = value;   \
michael@0:         (dst)[8] = value;   (dst)[9] = value;   \
michael@0:         (dst)[10] = value;  (dst)[11] = value;  \
michael@0:         (dst)[12] = value;  (dst)[13] = value;  \
michael@0:         (dst)[14] = value;  (dst)[15] = value;  \
michael@0:     } while (0)
michael@0: #else
michael@0: #define assign_16_longs(dst, value)             \
michael@0:     do {                                        \
michael@0:         *(dst)++ = value;   *(dst)++ = value;   \
michael@0:         *(dst)++ = value;   *(dst)++ = value;   \
michael@0:         *(dst)++ = value;   *(dst)++ = value;   \
michael@0:         *(dst)++ = value;   *(dst)++ = value;   \
michael@0:         *(dst)++ = value;   *(dst)++ = value;   \
michael@0:         *(dst)++ = value;   *(dst)++ = value;   \
michael@0:         *(dst)++ = value;   *(dst)++ = value;   \
michael@0:         *(dst)++ = value;   *(dst)++ = value;   \
michael@0:     } while (0)
michael@0: #endif
michael@0: 
michael@0: ///////////////////////////////////////////////////////////////////////////////
michael@0: 
michael@0: void sk_memset16_portable(uint16_t dst[], uint16_t value, int count) {
michael@0:     SkASSERT(dst != NULL && count >= 0);
michael@0: 
michael@0:     if (count <= 0) {
michael@0:         return;
michael@0:     }
michael@0: 
michael@0:     // not sure if this helps to short-circuit on small values of count
michael@0:     if (count < 8) {
michael@0:         do {
michael@0:             *dst++ = (uint16_t)value;
michael@0:         } while (--count != 0);
michael@0:         return;
michael@0:     }
michael@0: 
michael@0:     // ensure we're on a long boundary
michael@0:     if ((size_t)dst & 2) {
michael@0:         *dst++ = (uint16_t)value;
michael@0:         count -= 1;
michael@0:     }
michael@0: 
michael@0:     uint32_t value32 = ((uint32_t)value << 16) | value;
michael@0: 
michael@0:     // handle the bulk with our unrolled macro
michael@0:     {
michael@0:         int sixteenlongs = count >> 5;
michael@0:         if (sixteenlongs) {
michael@0:             uint32_t* dst32 = (uint32_t*)dst;
michael@0:             do {
michael@0:                 assign_16_longs(dst32, value32);
michael@0:             } while (--sixteenlongs != 0);
michael@0:             dst = (uint16_t*)dst32;
michael@0:             count &= 31;
michael@0:         }
michael@0:     }
michael@0: 
michael@0:     // handle (most) of the rest
michael@0:     {
michael@0:         int longs = count >> 1;
michael@0:         if (longs) {
michael@0:             do {
michael@0:                 *(uint32_t*)dst = value32;
michael@0:                 dst += 2;
michael@0:             } while (--longs != 0);
michael@0:         }
michael@0:     }
michael@0: 
michael@0:     // cleanup a possible trailing short
michael@0:     if (count & 1) {
michael@0:         *dst = (uint16_t)value;
michael@0:     }
michael@0: }
michael@0: 
michael@0: void sk_memset32_portable(uint32_t dst[], uint32_t value, int count) {
michael@0:     SkASSERT(dst != NULL && count >= 0);
michael@0: 
michael@0:     int sixteenlongs = count >> 4;
michael@0:     if (sixteenlongs) {
michael@0:         do {
michael@0:             assign_16_longs(dst, value);
michael@0:         } while (--sixteenlongs != 0);
michael@0:         count &= 15;
michael@0:     }
michael@0: 
michael@0:     if (count) {
michael@0:         do {
michael@0:             *dst++ = value;
michael@0:         } while (--count != 0);
michael@0:     }
michael@0: }
michael@0: 
michael@0: static void sk_memset16_stub(uint16_t dst[], uint16_t value, int count) {
michael@0:     SkMemset16Proc proc = SkMemset16GetPlatformProc();
michael@0:     sk_memset16 = proc ? proc : sk_memset16_portable;
michael@0:     sk_memset16(dst, value, count);
michael@0: }
michael@0: 
michael@0: SkMemset16Proc sk_memset16 = sk_memset16_stub;
michael@0: 
michael@0: static void sk_memset32_stub(uint32_t dst[], uint32_t value, int count) {
michael@0:     SkMemset32Proc proc = SkMemset32GetPlatformProc();
michael@0:     sk_memset32 = proc ? proc : sk_memset32_portable;
michael@0:     sk_memset32(dst, value, count);
michael@0: }
michael@0: 
michael@0: SkMemset32Proc sk_memset32 = sk_memset32_stub;
michael@0: 
michael@0: ///////////////////////////////////////////////////////////////////////////////
michael@0: 
michael@0: /*  0xxxxxxx    1 total
michael@0:     10xxxxxx    // never a leading byte
michael@0:     110xxxxx    2 total
michael@0:     1110xxxx    3 total
michael@0:     11110xxx    4 total
michael@0: 
michael@0:     11 10 01 01 xx xx xx xx 0...
michael@0:     0xE5XX0000
michael@0:     0xE5 << 24
michael@0: */
michael@0: 
michael@0: #ifdef SK_DEBUG
michael@0:     static void assert_utf8_leadingbyte(unsigned c) {
michael@0:         SkASSERT(c <= 0xF7);    // otherwise leading byte is too big (more than 4 bytes)
michael@0:         SkASSERT((c & 0xC0) != 0x80);   // can't begin with a middle char
michael@0:     }
michael@0: 
michael@0:     int SkUTF8_LeadByteToCount(unsigned c) {
michael@0:         assert_utf8_leadingbyte(c);
michael@0:         return (((0xE5 << 24) >> (c >> 4 << 1)) & 3) + 1;
michael@0:     }
michael@0: #else
michael@0:     #define assert_utf8_leadingbyte(c)
michael@0: #endif
michael@0: 
michael@0: int SkUTF8_CountUnichars(const char utf8[]) {
michael@0:     SkASSERT(utf8);
michael@0: 
michael@0:     int count = 0;
michael@0: 
michael@0:     for (;;) {
michael@0:         int c = *(const uint8_t*)utf8;
michael@0:         if (c == 0) {
michael@0:             break;
michael@0:         }
michael@0:         utf8 += SkUTF8_LeadByteToCount(c);
michael@0:         count += 1;
michael@0:     }
michael@0:     return count;
michael@0: }
michael@0: 
michael@0: int SkUTF8_CountUnichars(const char utf8[], size_t byteLength) {
michael@0:     SkASSERT(NULL != utf8 || 0 == byteLength);
michael@0: 
michael@0:     int         count = 0;
michael@0:     const char* stop = utf8 + byteLength;
michael@0: 
michael@0:     while (utf8 < stop) {
michael@0:         utf8 += SkUTF8_LeadByteToCount(*(const uint8_t*)utf8);
michael@0:         count += 1;
michael@0:     }
michael@0:     return count;
michael@0: }
michael@0: 
michael@0: SkUnichar SkUTF8_ToUnichar(const char utf8[]) {
michael@0:     SkASSERT(NULL != utf8);
michael@0: 
michael@0:     const uint8_t*  p = (const uint8_t*)utf8;
michael@0:     int             c = *p;
michael@0:     int             hic = c << 24;
michael@0: 
michael@0:     assert_utf8_leadingbyte(c);
michael@0: 
michael@0:     if (hic < 0) {
michael@0:         uint32_t mask = (uint32_t)~0x3F;
michael@0:         hic <<= 1;
michael@0:         do {
michael@0:             c = (c << 6) | (*++p & 0x3F);
michael@0:             mask <<= 5;
michael@0:         } while ((hic <<= 1) < 0);
michael@0:         c &= ~mask;
michael@0:     }
michael@0:     return c;
michael@0: }
michael@0: 
michael@0: SkUnichar SkUTF8_NextUnichar(const char** ptr) {
michael@0:     SkASSERT(NULL != ptr && NULL != *ptr);
michael@0: 
michael@0:     const uint8_t*  p = (const uint8_t*)*ptr;
michael@0:     int             c = *p;
michael@0:     int             hic = c << 24;
michael@0: 
michael@0:     assert_utf8_leadingbyte(c);
michael@0: 
michael@0:     if (hic < 0) {
michael@0:         uint32_t mask = (uint32_t)~0x3F;
michael@0:         hic <<= 1;
michael@0:         do {
michael@0:             c = (c << 6) | (*++p & 0x3F);
michael@0:             mask <<= 5;
michael@0:         } while ((hic <<= 1) < 0);
michael@0:         c &= ~mask;
michael@0:     }
michael@0:     *ptr = (char*)p + 1;
michael@0:     return c;
michael@0: }
michael@0: 
michael@0: SkUnichar SkUTF8_PrevUnichar(const char** ptr) {
michael@0:     SkASSERT(NULL != ptr && NULL != *ptr);
michael@0: 
michael@0:     const char* p = *ptr;
michael@0: 
michael@0:     if (*--p & 0x80) {
michael@0:         while (*--p & 0x40) {
michael@0:             ;
michael@0:         }
michael@0:     }
michael@0: 
michael@0:     *ptr = (char*)p;
michael@0:     return SkUTF8_NextUnichar(&p);
michael@0: }
michael@0: 
michael@0: size_t SkUTF8_FromUnichar(SkUnichar uni, char utf8[]) {
michael@0:     if ((uint32_t)uni > 0x10FFFF) {
michael@0:         SkDEBUGFAIL("bad unichar");
michael@0:         return 0;
michael@0:     }
michael@0: 
michael@0:     if (uni <= 127) {
michael@0:         if (utf8) {
michael@0:             *utf8 = (char)uni;
michael@0:         }
michael@0:         return 1;
michael@0:     }
michael@0: 
michael@0:     char    tmp[4];
michael@0:     char*   p = tmp;
michael@0:     size_t  count = 1;
michael@0: 
michael@0:     SkDEBUGCODE(SkUnichar orig = uni;)
michael@0: 
michael@0:     while (uni > 0x7F >> count) {
michael@0:         *p++ = (char)(0x80 | (uni & 0x3F));
michael@0:         uni >>= 6;
michael@0:         count += 1;
michael@0:     }
michael@0: 
michael@0:     if (utf8) {
michael@0:         p = tmp;
michael@0:         utf8 += count;
michael@0:         while (p < tmp + count - 1) {
michael@0:             *--utf8 = *p++;
michael@0:         }
michael@0:         *--utf8 = (char)(~(0xFF >> count) | uni);
michael@0:     }
michael@0: 
michael@0:     SkASSERT(utf8 == NULL || orig == SkUTF8_ToUnichar(utf8));
michael@0:     return count;
michael@0: }
michael@0: 
michael@0: ///////////////////////////////////////////////////////////////////////////////
michael@0: 
michael@0: int SkUTF16_CountUnichars(const uint16_t src[]) {
michael@0:     SkASSERT(src);
michael@0: 
michael@0:     int count = 0;
michael@0:     unsigned c;
michael@0:     while ((c = *src++) != 0) {
michael@0:         SkASSERT(!SkUTF16_IsLowSurrogate(c));
michael@0:         if (SkUTF16_IsHighSurrogate(c)) {
michael@0:             c = *src++;
michael@0:             SkASSERT(SkUTF16_IsLowSurrogate(c));
michael@0:         }
michael@0:         count += 1;
michael@0:     }
michael@0:     return count;
michael@0: }
michael@0: 
michael@0: int SkUTF16_CountUnichars(const uint16_t src[], int numberOf16BitValues) {
michael@0:     SkASSERT(src);
michael@0: 
michael@0:     const uint16_t* stop = src + numberOf16BitValues;
michael@0:     int count = 0;
michael@0:     while (src < stop) {
michael@0:         unsigned c = *src++;
michael@0:         SkASSERT(!SkUTF16_IsLowSurrogate(c));
michael@0:         if (SkUTF16_IsHighSurrogate(c)) {
michael@0:             SkASSERT(src < stop);
michael@0:             c = *src++;
michael@0:             SkASSERT(SkUTF16_IsLowSurrogate(c));
michael@0:         }
michael@0:         count += 1;
michael@0:     }
michael@0:     return count;
michael@0: }
michael@0: 
michael@0: SkUnichar SkUTF16_NextUnichar(const uint16_t** srcPtr) {
michael@0:     SkASSERT(srcPtr && *srcPtr);
michael@0: 
michael@0:     const uint16_t* src = *srcPtr;
michael@0:     SkUnichar       c = *src++;
michael@0: 
michael@0:     SkASSERT(!SkUTF16_IsLowSurrogate(c));
michael@0:     if (SkUTF16_IsHighSurrogate(c)) {
michael@0:         unsigned c2 = *src++;
michael@0:         SkASSERT(SkUTF16_IsLowSurrogate(c2));
michael@0: 
michael@0:         // c = ((c & 0x3FF) << 10) + (c2 & 0x3FF) + 0x10000
michael@0:         // c = (((c & 0x3FF) + 64) << 10) + (c2 & 0x3FF)
michael@0:         c = (c << 10) + c2 + (0x10000 - (0xD800 << 10) - 0xDC00);
michael@0:     }
michael@0:     *srcPtr = src;
michael@0:     return c;
michael@0: }
michael@0: 
michael@0: SkUnichar SkUTF16_PrevUnichar(const uint16_t** srcPtr) {
michael@0:     SkASSERT(srcPtr && *srcPtr);
michael@0: 
michael@0:     const uint16_t* src = *srcPtr;
michael@0:     SkUnichar       c = *--src;
michael@0: 
michael@0:     SkASSERT(!SkUTF16_IsHighSurrogate(c));
michael@0:     if (SkUTF16_IsLowSurrogate(c)) {
michael@0:         unsigned c2 = *--src;
michael@0:         SkASSERT(SkUTF16_IsHighSurrogate(c2));
michael@0:         c = (c2 << 10) + c + (0x10000 - (0xD800 << 10) - 0xDC00);
michael@0:     }
michael@0:     *srcPtr = src;
michael@0:     return c;
michael@0: }
michael@0: 
michael@0: size_t SkUTF16_FromUnichar(SkUnichar uni, uint16_t dst[]) {
michael@0:     SkASSERT((unsigned)uni <= 0x10FFFF);
michael@0: 
michael@0:     int extra = (uni > 0xFFFF);
michael@0: 
michael@0:     if (dst) {
michael@0:         if (extra) {
michael@0:             // dst[0] = SkToU16(0xD800 | ((uni - 0x10000) >> 10));
michael@0:             // dst[0] = SkToU16(0xD800 | ((uni >> 10) - 64));
michael@0:             dst[0] = SkToU16((0xD800 - 64) + (uni >> 10));
michael@0:             dst[1] = SkToU16(0xDC00 | (uni & 0x3FF));
michael@0: 
michael@0:             SkASSERT(SkUTF16_IsHighSurrogate(dst[0]));
michael@0:             SkASSERT(SkUTF16_IsLowSurrogate(dst[1]));
michael@0:         } else {
michael@0:             dst[0] = SkToU16(uni);
michael@0:             SkASSERT(!SkUTF16_IsHighSurrogate(dst[0]));
michael@0:             SkASSERT(!SkUTF16_IsLowSurrogate(dst[0]));
michael@0:         }
michael@0:     }
michael@0:     return 1 + extra;
michael@0: }
michael@0: 
michael@0: size_t SkUTF16_ToUTF8(const uint16_t utf16[], int numberOf16BitValues,
michael@0:                       char utf8[]) {
michael@0:     SkASSERT(numberOf16BitValues >= 0);
michael@0:     if (numberOf16BitValues <= 0) {
michael@0:         return 0;
michael@0:     }
michael@0: 
michael@0:     SkASSERT(utf16 != NULL);
michael@0: 
michael@0:     const uint16_t* stop = utf16 + numberOf16BitValues;
michael@0:     size_t          size = 0;
michael@0: 
michael@0:     if (utf8 == NULL) {    // just count
michael@0:         while (utf16 < stop) {
michael@0:             size += SkUTF8_FromUnichar(SkUTF16_NextUnichar(&utf16), NULL);
michael@0:         }
michael@0:     } else {
michael@0:         char* start = utf8;
michael@0:         while (utf16 < stop) {
michael@0:             utf8 += SkUTF8_FromUnichar(SkUTF16_NextUnichar(&utf16), utf8);
michael@0:         }
michael@0:         size = utf8 - start;
michael@0:     }
michael@0:     return size;
michael@0: }