Module:User:Erutuon/UTF-8

local export = {}

local m_debug = require("Module:debug")

-- Excludes null byte, which is supposed to be able to be included in Lua strings,
-- but causes patterns to fail.
local continuationByte = ""
---[[
local UTF8Char = "" .. continuationByte .. "*"
local nonASCII = "" .. continuationByte .. "+"
--]]
local escapePatt = "\\%d"
local hexPatt = "0x%X"

local floor = math.floor

local function highlight(text)
	return m_debug.highlight(text, { inline = true })
end

local function hex(number)
	return hexPatt:format(number)
end

local function byteEscape(number)
	return escapePatt:format(number)
end

local function escapeBytes(str)
	local out = {}
	for i, byte in ipairs{ string.byte(str, 1, -1) } do
		if byte < 128 then
			table.insert(out, string.char(byte))
		else
			table.insert(out, byteEscape(byte))
		end
	end
	return table.concat(out)
end

-- Based on the helpful byte chart at ].
local function getLeadingContinuation(codepoint)
	if codepoint < 0x80 then
		error("ASCII does not have leading bytes.")
	elseif codepoint < 0x800 then
		return 194 + floor((codepoint - 0x80) / 0x40), 1
	elseif codepoint < 0x10000 then
		return 224 + floor((codepoint - 0x800) / 0x1000), 2
	elseif codepoint < 0x11000 then
		return 240 + floor((codepoint - 0x10000) / 0x40000), 3
	else
		error(("Codepoint U+%X is outside valid range."):format(codepoint))
	end
end

function export.makeUTF8Pattern(lower, higher)
	local codepoint1, codepoint2 = mw.ustring.codepoint(lower), mw.ustring.codepoint(higher)
	local leading1, continuationCount1 = getLeadingContinuation(codepoint1)
	local leading2, continuationCount2 = getLeadingContinuation(codepoint2)
	local continuationSequence = string.rep(continuationByte, continuationCount1)
	if continuationCount1 < continuationCount2 then
		continuationSequence = continuationSequence .. "+"
	elseif continuationCount1 > continuationCount2 then
		error(string.format("The first character to makeUTF8Pattern (U+%X) should have a lower codepoint than the second (U+%X).", codepoint1, codepoint2))
	end
	local leading = leading1 == leading2 and byteEscape(leading1) or ""
	return lower .. "-" .. higher ..
		" (" .. highlight(hex(codepoint1) .. "-" .. hex(codepoint2)) .. "): " ..
		highlight(leading .. escapeBytes(continuationSequence))
end

function export.makeRange(characters)
	characters = string.gsub(characters, "%-", "")
	local firstChar = string.match(characters, UTF8Char)
	local lower, higher = firstChar, firstChar
	for character in string.gmatch(characters, UTF8Char) do
		if character < lower then
			lower = character
		elseif character > higher then
			higher = character
		end
	end
	return lower, higher
end

function export.charPatternForScript(scCode)
	local sc = require("Module:scripts").getByCode(scCode)
	local characters = sc:getCharacters()
	return export.makeUTF8Pattern(export.makeRange(characters))
end

function export.show(frame)
	return export.charPatternForScript(frame.args or "polytonic")
end

return export