From 7ed5c95e0f4bee427c48150dab92ae3229bc8c68 Mon Sep 17 00:00:00 2001 From: DTB Date: Wed, 29 May 2024 16:32:31 -0600 Subject: [PATCH] libutf: utf32be_to_utf8 --- src/libenc/libutf.c | 84 +++++++++++++++++++++++++++++++++++---------- src/libenc/libutf.h | 2 ++ 2 files changed, 68 insertions(+), 18 deletions(-) diff --git a/src/libenc/libutf.c b/src/libenc/libutf.c index ef5cf36..8812577 100644 --- a/src/libenc/libutf.c +++ b/src/libenc/libutf.c @@ -1,14 +1,37 @@ #include /* size_t */ #include "libutf.h" -/* UTF-32BE is the big-endian literal encoding of a Unicode codepoint, - * including 11 bits of padding. The following functions convert from and to - * UTF-32. */ +/* This is functionally equivalent to the UTF-32-specific conversion functions + * but very slightly slower than each. */ +/* This operation is symmetrical; swab32(swab32(c)) will always return c. */ +/* big-endian | ZZZZ YYYY | XXXX WWWW | VVVV UUUU | TTTT SSSS */ +/* little-endian | TTTT SSSS | VVVV UUUU | */ +rune_t +swab32(rune_t c){ return + ((c & 0xFF000000) >> 24) + | ((c & 0x00FF0000) >> 8) + | ((c & 0x0000FF00) << 8) + | ((c & 0x000000FF) << 24); +} /* From , UTF-8 is encoded as follows: * The codepoint 0bZYYYY_XXXXWWWW_VVVVUUUU is nestled within the literal bits 0 * and 1; letters denote nybbles here, little-endian, and certain bits are * labeled, also little-endian. */ + +/* utf-8 bits |32 |24 |16 |8 + * U+001000 to U+10FFFF | 1111 0ZYY | 10YY XXXX | 10WW WWVV | 10VV UUUU * + * U+000800 to U+00FFFF | 0000 0000 | 1110 XXXX | 10WW WWVV | 10VV UUUU * + * U+000080 to U+0007FF | 0000 0000 | 0000 0000 | 110W WWVV | 10VV UUUU * + * U+000000 to U+00007F | 0000 0000 | 0000 0000 | 0000 0000 | 0VVV UUUU */ +size_t +utf8_size(rune_t c){ return + 1 + ((c & 0x80000000) >> 31) /* 4B? */ + + ((c & 0x00800000) >> 23) /* 3B? */ + + ((c & 0x00008000) >> 15) /* 2B? */ + + ((c & 0x00000080) >> 7); /* 1B? */ +} + /* utf-8 bits |32 |27 |22 |14 |7 * U+001000 to U+10FFFF | 1111 0ZYY | 10YY XXXX | 10WW WWVV | 10VV UUUU * * U+000800 to U+00FFFF | 0000 0000 | 1110 XXXX | 10WW WWVV | 10VV UUUU * @@ -16,7 +39,7 @@ * U+000000 to U+00007F | 0000 0000 | 0000 0000 | 0000 0000 | 0VVV UUUU */ /* utf-32be bits |32 |21 |18 |12 |6 * U+000000 to U+10FFFF | 0000 0000 | 000Z YYYY | XXXX WWWW | VVVV UUUU */ -codepoint_t +rune_t utf8_to_utf32be(rune_t c){ return ((c & 0x07000000) >> 6) /* UTF-8 b27-25 -> UTF-32 b21-19 */ | (((c & 0x003F0000) /* UTF-8 b22-17 -> UTF-32 b18-13 */ @@ -24,6 +47,41 @@ utf8_to_utf32be(rune_t c){ return | ((c & 0x00003F00) >> 2) /* UTF-8 b14- 9 -> UTF-32 b12- 7 */ | (c & 0x0000007F); /* UTF-32 b 7- 1 */ } +/* utf-32be bits |32 |21 |18 |12 |6 + * U+000000 to U+10FFFF | 0000 0000 | 000Z YYYY | XXXX WWWW | VVVV UUUU */ +/* utf-8 bits |32 |27 |22 |14 |7 + * U+001000 to U+10FFFF | 1111 0ZYY | 10YY XXXX | 10WW WWVV | 10VV UUUU * + * U+000800 to U+00FFFF | 0000 0000 | 1110 XXXX | 10WW WWVV | 10VV UUUU * + * U+000080 to U+0007FF | 0000 0000 | 0000 0000 | 110W WWVV | 10VV UUUU * + * U+000000 to U+00007F | 0000 0000 | 0000 0000 | 0000 0000 | 0VVV UUUU */ +/* m is the minimum amount of bytes into which to encode the codepoint c. If m + * is greater than 0, this function may return overlong-encoded UTF-8. */ +rune_t +utf32be_to_utf8(rune_t c, size_t m){ + rune_t r; + { size_t n; + if((n = utf8_size(c)) > m) + m = n; } /* This avoids calculating the size twice. */ + switch(m){ /* "Trin's device" if this is a novel use of a switch. */ + case 4: r = 0xF0000000 /* UTF-8 b32-29 */ + | ((c & 0x1C0000) << 6) /* UTF-32 b21-19 -> UTF-8 b27-25 */ + | 0x00800000 /* UTF-8 b24-23 */ + | ((c & 0x03F000) << 4); /* UTF-32 b18-13 -> UTF-8 b24-17 */ + if(m == 3) + case 3: r = (0xE << 4) /* UTF-8 b24-21 */ + | ((c & 0x00F000) << 4); /* UTF-32 b16-13 -> UTF-8 b20-17 */ + r |= 0x00008000 /* UTF-8 b16-15 */ + | ((c & 0x000FC0) << 2); /* UTF-32 b12- 7 -> UTF-8 b14- 9 */ + if(m == 2) + case 2: r = (0xE << 3) /* UTF-8 b16-14 */ + | ((c & 0x007C00) << 2); /* UTF-32 b11- 7 -> UTF-8 b13- 9 */ + r |= 0x00000080 /* UTF-8 b 8- 7 */ + | (c & 0x00003F); /* UTF-8 b 6- 1 */ + if(m == 1) + case 1: r = c & 0x00007F; /* UTF-8 b 7- 1 */ + } + return r; +} /* is a good * explanation of this. */ @@ -31,25 +89,15 @@ utf8_to_utf32be(rune_t c){ return * U+000000 to U+10FFFF | 0000 0000 | 000Z YYYY | XXXX WWWW | VVVV UUUU */ /* utf-32le bits * U+000000 to U+10FFFF | VVVV UUUU | XXXX WWWW | 000Z YYYY | 0000 0000 */ -codepoint_t -utf32be_to_utf32le(codepoint_t c){ return +rune_t +utf32be_to_utf32le(rune_t c){ return ((c & 0x000000FF) << 24) | ((c & 0x0000FF00) << 8) | ((c & 0x001F0000) >> 8); } -codepoint_t -utf32le_to_utf32be(codepoint_t c){ return +rune_t +utf32le_to_utf32be(rune_t c){ return ((c & 0xFF000000) >> 24) | ((c & 0x00FF0000) >> 8) | ((c & 0x00001F00) << 8); } -/* This operation is symmetrical; swab32(swab32(c)) will always return c. It's - * (very slightly) slower than the specific UTF-32 conversion functions but may - * be useful. */ -codepoint_t -swab32(codepoint_t c){ return - ((c & 0xFF000000) >> 24) - | ((c & 0x00FF0000) >> 8) - | ((c & 0x0000FF00) << 8) - | ((c & 0x000000FF) << 24); -} diff --git a/src/libenc/libutf.h b/src/libenc/libutf.h index 6b004f3..feafde6 100644 --- a/src/libenc/libutf.h +++ b/src/libenc/libutf.h @@ -7,8 +7,10 @@ typedef uint32_t rune_t; * */ typedef unsigned long int rune_t; #endif +#include /* size_t */ rune_t swab32(rune_t c); rune_t utf8_to_utf32be(rune_t c); +rune_t utf32be_to_utf8(rune_t c, size_t m); rune_t utf32be_to_utf32le(rune_t c); rune_t utf32le_to_utf32be(rune_t c);