Module:Syl-translit

-- Transliteration for Sylheti in Sylheti Nagri script

local export = {} local gsub = mw.ustring.gsub local match = mw.ustring.match local conv = { -- consonants ['ꠇ'] = 'x',	['ꠈ'] = 'x',	['ꠉ'] = 'g',	['ꠊ'] = 'g', ['ꠌ'] = 's',	['ꠍ'] = 's',	['ꠎ'] = 'z',	['ꠏ'] = 'z', ['ꠐ'] = 'ṭ',	['ꠑ'] = 'ṭ',	['ꠒ'] = 'ḍ',	['ꠓ'] = 'ḍ', ['ꠔ'] = 't',	['ꠕ'] = 't',	['ꠖ'] = 'd',	['ꠗ'] = 'd',	['ꠘ'] = 'n', ['ꠙ'] = 'f',	['ꠚ'] = 'f',	['ꠛ'] = 'b',	['ꠜ'] = 'b',	['ꠝ'] = 'm', ['ꠞ'] = 'r',	['ꠟ'] = 'l',	['ꠠ'] = 'ṛ', ['ꠡ'] = 'ś',   ['ꠢ'] = 'h', ['ꠋ'] = 'ṅ',

-- vowel diacritics ['ꠣ'] = 'a', ['ꠤ'] = 'i',	['ꠥ'] = 'u', ['ꠦ'] = 'e', ['ꠧ'] = 'o',

-- vowel signs ['ꠀ'] = 'a', ['ꠁ'] = 'i', ['ꠃ'] = 'u', ['ꠄ'] = 'e', ['ꠅ'] = 'o',

-- virama ['꠆'] = '',	-- anusvar [' ꠋ'] = 'ṅ',

-- dvisvara ['ꠂ'] = 'i',

-- numerals ['০'] = '0', ['১'] = '1', ['২'] = '2', ['৩'] = '3', ['৪'] = '4', ['৫'] = '5', ['৬'] = '6', ['৭'] = '7', ['৮'] = '8', ['৯'] = '9',	-- punctuation ['꠪'] = '.', -- dari/purn virama ['꠫'] = '.' }

local aspirates = "ꠈꠊꠍꠏꠑꠓꠕꠗꠚꠜ" local consonant, vowel, vowel_sign = "ꠇ-ꠊꠌ-ꠢ", "ꠣ-ꠧ", "ꠀꠁꠃ-ꠅ" local c = "[" .. consonant .. "]" local v = "[" .. vowel .. vowel_sign .. "]" local syncope_pattern = "(" .. v .. c .. v .. c .. ")ô(" .. c .. "ঁ?" .. v .. ")"

local function rev_string(text) local result, length = "", mw.ustring.len(text) for i = 1, length do result = result .. mw.ustring.sub(text, -i, -i) end return result end

function export.tr(text, lang, sc) -- from MOD:as-translit text = gsub(text, "(" .. c .. ")([" .. vowel .. "’?꠆]?)", function(a, b)		local res = a .. (b == "" and "ô" or b)		if match(a, "[" .. aspirates .. "]") then res = res .. "’" end		return res end) for word in mw.ustring.gmatch(text, "[ꠀ-ꠧô’]+") do		local orig_word = word word = rev_string(word) word = gsub(word, "^ô(" .. c .. ")(" .. v .. ")", "%1%2")		while match(word, syncope_pattern) do			word = gsub(word, syncope_pattern, "%1%2") end text = gsub(text, orig_word, rev_string(word)) end text = gsub(text, ".", conv)

-- ô is really just o	text = mw.ustring.gsub(text, 'ô', 'o') -- velars text = mw.ustring.gsub(text, 'x([iu])', 'k%1') text = mw.ustring.gsub(text, '([iu])x', '%1k') text = mw.ustring.gsub(text, 'xx', 'kk') -- palatals (I think?) text = mw.ustring.gsub(text, 'ss', 'cc') -- affricates (I'm just guessing now) text = mw.ustring.gsub(text, 'jj', 'zz') text = mw.ustring.gsub(text, 'zs', 'jc') -- final r/l text = mw.ustring.gsub(text, '([xrl])o$', '%1') text = mw.ustring.gsub(text, '([xrl])o ', '%1 ')

-- tone text = gsub(text, '’', '́') return text end return export