From feb6ad276ecd25563d5172083305adc4ab7b3aac Mon Sep 17 00:00:00 2001
From: Lars Mueller <appgurulars@gmx.de>
Date: Sat, 1 Oct 2022 19:40:22 +0200
Subject: [PATCH] Add utf8.codepoint (inverse of utf8.char)

---
 utf8.lua | 40 +++++++++++++++++++++++++++++++++++++++-
 1 file changed, 39 insertions(+), 1 deletion(-)

diff --git a/utf8.lua b/utf8.lua
index 6d5b47c..9bc0b05 100644
--- a/utf8.lua
+++ b/utf8.lua
@@ -34,7 +34,11 @@ end
 
 function utf8.char(...)
 	local n_args = select("#", ...)
-	if n_args == 0 then return end
+	if n_args == 0 then
+		return
+	end if n_args == 1 then
+		return string_char(utf8_bytes(...))
+	end
 	local chars = {}
 	for i = 1, n_args do
 		chars[i] = string_char(utf8_bytes(select(i, ...)))
@@ -45,4 +49,38 @@ end
 -- Overly permissive pattern that greedily matches a single UTF-8 codepoint
 utf8.charpattern = "[%z-\127\194-\253][\128-\191]*"
 
+local function utf8_codepoint(str)
+	local first_byte = str:byte()
+	if first_byte < 0x80 then
+		assert(#str == 1, "invalid UTF-8")
+		return first_byte
+	end
+
+	local len, head_bits
+	if first_byte >= 0xC0 and first_byte <= 0xDF then -- 110_00000 to 110_11111
+		len, head_bits = 2, first_byte % 0x20 -- last 5 bits
+	elseif first_byte >= 0xE0 and first_byte <= 0xEF then -- 1110_0000 to 1110_1111
+		len, head_bits = 3, first_byte % 0x10 -- last 4 bits
+	elseif first_byte >= 0xF0 and first_byte <= 0xF7 then -- 11110_000 to 11110_111
+		len, head_bits = 4, first_byte % 0x8 -- last 3 bits
+	else error"invalid UTF-8" end
+	assert(#str == len, "invalid UTF-8")
+
+	local codepoint = 0
+	local pow = 1
+	for i = len, 2, -1 do
+		local byte = str:byte(i)
+		local val_bits = byte % 0x40 -- extract last 6 bits xxxxxx from 10xxxxxx
+		assert(byte - val_bits == 0x80) -- assert that first two bits are 10
+		codepoint = codepoint + val_bits * pow
+		pow = pow * 0x40
+	end
+	return codepoint + head_bits * pow
+end
+
+function utf8.codepoint(...)
+	if select("#", ...) == 0 then return end
+	return utf8_codepoint(...), utf8.codepoint(select(2, ...))
+end
+
 return utf8
\ No newline at end of file