From 2479ab63d016e93b7ebaa5f05d875250b5210687 Mon Sep 17 00:00:00 2001
From: DTB <trinity@trinity.moe>
Date: Mon, 27 May 2024 17:38:35 -0600
Subject: [PATCH] libutf: UTF conversion functionality

---
 src/libenc/libutf.c | 55 +++++++++++++++++++++++++++++++++++++++++++++
 src/libenc/libutf.h | 14 ++++++++++++
 2 files changed, 69 insertions(+)
 create mode 100644 src/libenc/libutf.c
 create mode 100644 src/libenc/libutf.h

diff --git a/src/libenc/libutf.c b/src/libenc/libutf.c
new file mode 100644
index 0000000..ef5cf36
--- /dev/null
+++ b/src/libenc/libutf.c
@@ -0,0 +1,55 @@
+#include <stdlib.h> /* 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. */
+
+/* From <https://en.wikipedia.org/wiki/UTF-8>, 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   |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 */
+/* utf-32be bits          |32            |21 |18       |12      |6
+ * U+000000 to U+10FFFF | 0000 0000 | 000Z YYYY | XXXX WWWW | VVVV UUUU */
+codepoint_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 */
+		^ ((c & (1 << 22)) >> 1)) >> 4) /* (if UTF-8 b23, zero b22) */
+	|  ((c & 0x00003F00) >> 2)     /* UTF-8 b14- 9 -> UTF-32 b12- 7 */
+	|   (c & 0x0000007F);                          /* UTF-32 b 7- 1 */
+}
+
+/* <https://www.herongyang.com/Unicode/UTF-32-UTF-32-Encoding.html> is a good
+ * explanation of this. */
+/* utf-32be bits
+ * 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
+	  ((c & 0x000000FF) << 24)
+	| ((c & 0x0000FF00) << 8)
+	| ((c & 0x001F0000) >> 8);
+}
+codepoint_t
+utf32le_to_utf32be(codepoint_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
new file mode 100644
index 0000000..6b004f3
--- /dev/null
+++ b/src/libenc/libutf.h
@@ -0,0 +1,14 @@
+#if __STDC_VERSION__ >= 199901L
+/* C99 type definitions */
+#	include <stdint.h>
+typedef uint32_t rune_t;
+#else
+/* Must hold at least 32b; see the C89 draft 2.2.4.2
+ * <http://jfxpt.com/library/c89-draft.html#2.2.4.2> */
+typedef unsigned long int rune_t;
+#endif
+
+rune_t swab32(rune_t c);
+rune_t utf8_to_utf32be(rune_t c);
+rune_t utf32be_to_utf32le(rune_t c);
+rune_t utf32le_to_utf32be(rune_t c);