1 files changed, 44 insertions, 0 deletions
diff --git a/libmisc/utf8.c b/libmisc/utf8.c
new file mode 100644
index 0000000..28357f0
--- /dev/null
+++ b/libmisc/utf8.c
@@ -0,0 +1,44 @@
+/* libmisc/utf8.c - UTF-8 routines
+ *
+ * Copyright (C) 2024-2025  Luke T. Shumaker <lukeshu@lukeshu.com>
+ * SPDX-License-Identifier: AGPL-3.0-or-later
+ */
+
+#include <libmisc/utf8.h>
+
+void utf8_decode_codepoint(const uint8_t *str, size_t len, uint32_t *ret_ch, uint8_t *ret_chlen) {
+	uint32_t ch;
+	uint8_t  chlen;
+	uint32_t chmin;
+	if      ((str[0] & 0b10000000) == 0b00000000) { ch = str[0] & 0b01111111; chlen = 1; chmin = 0;               } /* bits=7+(0*6)= 7 */
+	else if ((str[0] & 0b11100000) == 0b11000000) { ch = str[0] & 0b00011111; chlen = 2; chmin = UINT32_C(1)<< 7; } /* bits=5+(1*6)=11 */
+	else if ((str[0] & 0b11110000) == 0b11100000) { ch = str[0] & 0b00001111; chlen = 3; chmin = UINT32_C(1)<<11; } /* bits=4+(2*6)=16 */
+	else if ((str[0] & 0b11111000) == 0b11110000) { ch = str[0] & 0b00000111; chlen = 4; chmin = UINT32_C(1)<<16; } /* bits=3+(3*6)=21 */
+	else goto invalid;
+	if (chlen > len)
+		goto invalid;
+	for (uint8_t i = 1; i < chlen; i++) {
+		if ((str[i] & 0b11000000) != 0b10000000)
+			goto invalid;
+		ch = (ch << 6) | (str[i] & 0b00111111);
+	}
+	if (ch > 0x10FFFF || ch < chmin)
+		goto invalid;
+	*ret_ch    = ch;
+	*ret_chlen = chlen;
+	return;
+ invalid:
+	*ret_chlen = 0;
+}
+
+bool _utf8_is_valid(const uint8_t *str, size_t len, bool forbid_nul) {
+	for (size_t pos = 0; pos < len;) {
+		uint32_t ch;
+		uint8_t  chlen;
+		utf8_decode_codepoint(&str[pos], len-pos, &ch, &chlen);
+		if (chlen == 0 || (forbid_nul && ch == 0))
+			return false;
+		pos += chlen;
+	}
+	return true;
+}