မေႃႇၵျူး:grc-translit
Appearance
Documentation for this module may be created at မေႃႇၵျူး:grc-translit/doc
local export = {}
local m_data = require('Module:grc-utilities/data')
local tokenize = require('Module:grc-utilities').tokenize
local ufind = mw.ustring.find
local ugsub = mw.ustring.gsub
local U = mw.ustring.char
local ulower = mw.ustring.lower
local uupper = mw.ustring.upper
local UTF8char = '[%z\1-\127\194-\244][\128-\191]*'
-- Diacritics
local diacritics = m_data.named
-- Greek
local acute = diacritics.acute
local grave = diacritics.grave
local circumflex = diacritics.circum
local diaeresis = diacritics.diaeresis
local smooth = diacritics.smooth
local rough = diacritics.rough
local macron = diacritics.macron
local breve = diacritics.breve
local subscript = diacritics.subscript
-- Latin
local hat = diacritics.Latin_circum
local macron_diaeresis = macron .. diaeresis .. "?" .. hat
local a_subscript = '^[αΑ].*' .. subscript .. '$'
local velar = 'κγχξ'
local tt = {
-- Vowels
["α"] = "a",
["ε"] = "e",
["η"] = "e" .. macron,
["ι"] = "i",
["ο"] = "o",
["υ"] = "u",
["ω"] = "o" .. macron,
-- Consonants
["β"] = "b",
["γ"] = "g",
["δ"] = "d",
["ζ"] = "z",
["θ"] = "th",
["κ"] = "k",
["λ"] = "l",
["μ"] = "m",
["ν"] = "n",
["ξ"] = "x",
["π"] = "p",
["ρ"] = "r",
["σ"] = "s",
["ς"] = "s",
["τ"] = "t",
["φ"] = "ph",
["χ"] = "kh",
["ψ"] = "ps",
-- Archaic letters
["ϝ"] = "w",
["ϻ"] = "ś",
["ϙ"] = "q",
["ϡ"] = "š",
["ͷ"] = "v",
-- Incorrect characters: see [[Wiktionary:About Ancient Greek#Miscellaneous]].
-- These are tracked by [[Module:script utilities]].
["ϐ"] = "b",
["ϑ"] = "th",
["ϰ"] = "k",
["ϱ"] = "r",
["ϲ"] = "s",
["ϕ"] = "ph",
-- Diacritics
-- unchanged: macron, diaeresis, grave, acute
[breve] = '',
[smooth] = '',
[rough] = '',
[circumflex] = hat,
[subscript] = 'i',
}
function export.tr(text, lang, sc)
if text == '῾' then
return 'h'
end
--[[
Replace semicolon or Greek question mark with regular question mark,
except after an ASCII alphanumeric character (to avoid converting
semicolons in HTML entities).
]]
text = ugsub(text, "([^A-Za-z0-9])[;" .. U(0x37E) .. "]", "%1?")
-- Handle the middle dot. It is equivalent to semicolon or colon, but semicolon is probably more common.
text = text:gsub("·", ";")
local tokens = tokenize(text)
--now read the tokens
local output = {}
for i, token in pairs(tokens) do
-- Convert token to lowercase and substitute each character
-- for its transliteration
local translit = ulower(token):gsub(UTF8char, tt)
local next_token = tokens[i + 1]
if token == 'γ' and next_token and velar:find(next_token, 1, true) then
-- γ before a velar should be <n>
translit = 'n'
elseif token == 'ρ' and tokens[i - 1] == 'ρ' then
-- ρ after ρ should be <rh>
translit = 'rh'
elseif ufind(token, a_subscript) then
-- add macron to ᾳ
translit = ugsub(translit, '([aA])', '%1' .. macron)
end
if token:find(rough) then
if ufind(token, '^[Ρρ]') then
translit = translit .. 'h'
else -- vowel
translit = 'h' .. translit
end
end
-- Remove macron from a vowel that has a circumflex.
if ufind(translit, macron_diaeresis) then
translit = translit:gsub(macron, '')
end
-- Capitalize first character of transliteration.
if token ~= ulower(token) then
translit = translit:gsub("^" .. UTF8char, uupper)
end
table.insert(output, translit)
end
output = table.concat(output)
return output
end
return export