Modül:ru-alfabeçeviri

local export = {}

--[=[

FIXME:

1. (DONE) If you write '''Б'''ез, it transliterates to '''B'''jez instead of
   '''B'''ez.
2. (DONE) Convert ъ to nothing before comma or other non-letter particle, e.g.
   in Однимъ словомъ, идешь на чтеніе.
3. (DONE) Make special-casing for adjectives in -го and for что (and friends)
    be the default, and implement transformations in Cyrillic rather than after
    translit so that we can display the transformed Cyrillic in the
    "phonetic respelling" notation of {{ru-IPA}}.
]=]

local u = mw.ustring.char
local rfind = mw.ustring.find
local rsub = mw.ustring.gsub -- WARNING: Don't return this directly in a function, or surround in parens
local rmatch = mw.ustring.match
local rsplit = mw.text.split
local ulower = mw.ustring.lower
local usub = mw.ustring.sub

local GR = u(0x0300) -- grave =  ̀
local TEMP_G = u(0xFFF1) -- substitute to preserve g from changing to v

local function ine(x) -- if not empty
	if x == "" then return nil else return x end
end

-- In this table, we now map Cyrillic е and э to je and e, and handle the
-- post-consonant version (plain e and ɛ) specially.
local tab = {
	="A", ="B", ="V", ="G", ="D", ="Ye", ="Yó", ="J", ="Z", ="İ", ="Y",
	="K", ="L", ="M", ="N", ="O", ="P", ="R", ="S", ="T", ="U", ="F",
	="H", ="Ts", ="Ç", ="Ş", ="Şç", ="ʺ", ="I", ="ʹ", ="E", ="Yu", ="Ya",
	='a', ='b', ='v', ='g', ='d', ='ye', ='yó', ='j', ='z', ='i', ='y',
	='k', ='l', ='m', ='n', ='o', ='p', ='r', ='s', ='t', ='u', ='f',
	='h', ='ts', ='ç', ='ş', ='şç', ='ʺ', ='ı', ='ʹ', ='e', ='yu', ='ya',
	-- Russian style quotes
	='“', ='”',
	-- archaic, pre-1918 letters
	='I', ='i', ='F', ='f',
	='Yě', ='yě', ='I', ='i',
}

-- following based on ru-common for use with is_monosyllabic()
-- any Cyrillic or Latin vowel, including ёЁ and composed Cyrillic vowels with grave accent;
-- not including accented Latin vowels except ě (FIXME, might want to change this)
local vowels = "аеиоуяэыюіѣѵүАЕИОУЯЭЫЮІѢѴҮѐЀѝЍёЁAEIOUYĚƐaeiouyěɛ"

-- FIXME! Doesn't work with ɣ, which gets included in this character set
local non_consonants = ""
local consonants = ""

local map_to_plain_e_map = { = "E",  = "e",  = "Ě",  = "ě",  = "Ɛ",  = "ɛ"}
local function map_to_plain_e(pre, e)
	return pre .. map_to_plain_e_map
end

local map_to_je_map = { = "Je",  = "je",  = "Jě",  = "jě",  = "E",  = "e"}
local function map_to_je(pre, e)
	if e == nil then
		e = pre
		pre = ""
	end
	return pre .. map_to_je_map
end

-- decompose composed grave chars; they will map to uncomposed Latin letters for
-- consistency with other char+grave combinations, and we do this early to
-- avoid problems converting to e or je
local decompose_grave_map = { = 'е' .. GR,  = 'Е' .. GR,  = 'и' .. GR,  = 'И' .. GR}

-- True if Cyrillic or decomposed Latin word has no more than one vowel;
-- includes non-syllabic stems such as льд-; copied from ru-common and modified
-- to avoid having to import that module (which would slow things down
-- significantly)
local function is_monosyllabic(word)
	return not rfind(word, ".*")
end

-- Apply transformations to the Cyrillic to more closely match pronunciation.
-- Return two arguments: the "original" text (after decomposing composed
-- grave characters), and the transformed text. If the two are different,
-- {{ru-IPA}} should display a "phonetic respelling" notation. 
-- NOADJ disables special-casing for adjectives in -го, while FORCEADJ forces
-- special-casing for adjectives, including those in -аго (pre-reform spelling)
-- and disables checking for exceptions (e.g. много, ого). NOSHTO disables
-- special-casing for что and related words.
function export.apply_tr_fixes(text, noadj, noshto, forceadj)
	-- decompose composed grave characters before we convert Cyrillic е to
	-- Latin e or je
	text = rsub(text, "", decompose_grave_map)

	local origtext = text
	-- the second half of the if-statement below is an optimization; see above.
	if not noadj and text:find("го") then
		if not forceadj then
			-- handle много
			text = rsub(text, "%f(но?)го%f", "%1" .. TEMP_G .. "о")
			-- handle немного, намного
			text = rsub(text, "%f(мно?)го%f", "%1" .. TEMP_G .. "о")
			-- handle до́рого 
			text = rsub(text, "%f(о?ро)го%f", "%1" .. TEMP_G .. "о")
			-- handle недо́рого 
			text = rsub(text, "%f(едо?ро)го%f", "%1" .. TEMP_G .. "о")
			-- handle стро́го
			text = rsub(text, "%f(тро?)го%f", "%1" .. TEMP_G .. "о")
			-- handle нестро́го
			text = rsub(text, "%f(естро?)го%f", "%1" .. TEMP_G .. "о")
			-- handle убо́го
			text = rsub(text, "%f(бо?)го%f", "%1" .. TEMP_G .. "о")
			-- handle поло́го
			text = rsub(text, "%f(оло?)го%f", "%1" .. TEMP_G .. "о")
			-- handle длинноно́го
			text = rsub(text, "%f(линноно?)го%f", "%1" .. TEMP_G .. "о")
			-- handle коротконо́го
			text = rsub(text, "%f(оротконо?)го%f", "%1" .. TEMP_G .. "о")
			-- handle кривоно́го
			text = rsub(text, "%f(ривоно?)го%f", "%1" .. TEMP_G .. "о")
			-- handle пе́го 
			text = rsub(text, "%f(е?)го%f", "%1" .. TEMP_G .. "о")
			-- handle лого, сого, ого
			text = rsub(text, "%f(??)г(о?)%f", "%1" .. TEMP_G .. "%2")
			-- handle Того, То́го (but not того or Того́, which have /v/)
			text = rsub(text, "%f(То́?)го%f", "%1" .. TEMP_G .. "о")
			-- handle лего
			text = rsub(text, "%f(е?)го%f", "%1" .. TEMP_G .. "о")
			-- handle игого, огого; note, we substitute TEMP_G for both г's
			-- because otherwise the ого- at the beginning gets converted to ово
			text = rsub(text, "%f()гог(о?)%f", "%1" .. TEMP_G .. "о" .. TEMP_G .. "%2")
			-- handle Диего
			text = rsub(text, "%f(Дие́?)го%f", "%1" .. TEMP_G .. "о")
		end
		--handle genitive/accusative endings, which are spelled -ого/-его/-аго
		-- (-ogo/-ego/-ago) but transliterated -ovo/-evo/-avo; only for adjectives
		-- and pronouns, excluding words like много, ого (-аго occurs in
		-- pre-reform spelling); \204\129 is an acute accent, \204\128 is a grave accent
		local pattern = "(?)()(?)"
		local reflexive = "(?)"
		local v = { = "в",  = "В"}
		local repl = function(e, g, o, sja) return e .. v .. o .. (sja or "") end
		text = rsub(text, pattern .. "%f", repl)
		text = rsub(text, pattern .. reflexive .. "%f", repl)
		-- handle сегодня
		text = rsub(text, "%f(е)г(о?дня)%f", "%1в%2")
		-- handle сегодняшн-
		text = rsub(text, "%f(е)г(о?дняшн)", "%1в%2")
		-- replace TEMP_G with g; must be done after the -go -> -vo changes
		text = rsub(text, TEMP_G, "г")
	end

	-- the second half of the if-statement below is an optimization; see above.
	if not noshto and text:find("то") then
		local ch2sh = { = "ш",  = "Ш"}
		-- Handle что
		text = rsub(text, "%f()(то?)%f",
			function(ch, to) return ch2sh .. to end)
		-- Handle чтобы, чтоб
		text = rsub(text, "%f()(то?бы?)%f",
			function(ch, to) return ch2sh .. to end)
		-- Handle ничто
		text = rsub(text, "%f(и)ч(то?)%f", "%1ш%2")
	end

	text = rsub(text, "(?)г()", "%1х%2")

	return origtext, text
end

-- Transliterate after the pronunciation-related transformations of
-- export.apply_tr_fixes() have been applied. Called from {{ru-IPA}}.
-- INCLUDE_MONOSYLLABIC_JO_ACCENT is as in export.tr().
function export.tr_after_fixes(text, include_monosyllabic_jo_accent)
	-- Remove word-final hard sign, either utterance-finally or followed by
	-- a non-letter character such as space, comma, period, hyphen, etc.
	text = rsub(text, "$", "")
	text = rsub(text, "()", "%1")

	 -- the if-statement below isn't necessary but may speed things up,
	 -- particularly when include_monosyllabic_jo_accent isn't set, in that
	 -- in the majority of cases where ё doesn't occur, we avoid a pattern find
	 -- (in is_monosyllabic()) and three pattern subs. The translit module needs
	 -- to be as fast as possible since it may be called hundreds or
	 -- thousands of times on some pages.
	 if rfind(text, "") then
		-- We need to special-case ё after a "hushing" consonant, which becomes
		-- ó (or o), without j. We also need special cases for monosyllabic ё
		-- when INCLUDE_MONOSYLLABIC_JO_ACCENT isn't set, so we don't add the
		-- accent mark that we would otherwise include.
		if not include_monosyllabic_jo_accent and is_monosyllabic(text) then
			text = rsub(text, "()ё","%1o")
			text = text:gsub("ё", "jo")
			text = text:gsub("Ё", "Jo")
		else
			text = rsub(text, "()ё","%1ó")
			-- conversion of remaining ё will occur as a result of 'tab'.
		end
	end

	-- ю after ж and ш becomes u (e.g. брошюра, жюри)
	text = rsub(text, "()ю","%1u")

	 -- the if-statement below isn't necessary but may speed things up in that
	 -- in the majority of cases where the letters below don't occur, we avoid
	 -- six pattern subs.
	 if rfind(text, "") then
		-- е after a dash at the beginning of a word becomes e, and э becomes ɛ
		-- (like after a consonant)
		text = rsub(text, "^(%-)()", map_to_plain_e)
		text = rsub(text, "(%s%-)()", map_to_plain_e)
		-- don't get confused by single quote or parens between consonant and е;
		-- e.g. Б'''ез''', американ(ец)
		text = rsub(text, "(" .. consonants .. "*)()", map_to_plain_e)

		-- This is now the default
		-- е after a vowel or at the beginning of a word becomes je, and э becomes e
		-- text = rsub(text, "^()", map_to_je)
		-- text = rsub(text, "(" .. non_consonants .. ")()", map_to_je)
		-- -- need to do it twice in case of sequences of such vowels
		-- text = rsub(text, "^()", map_to_je)
		-- text = rsub(text, "(" .. non_consonants .. ")()", map_to_je)
	end

	text = (rsub(text,'.',tab))
	return text
end

-- Transliterates text, which should be a single word or phrase. It should
-- include stress marks, which are then preserved in the transliteration.
-- ё is a special case: it is rendered (j)ó in multisyllabic words and
-- monosyllabic words in multi-word phrases, but rendered (j)o without an
-- accent in isolated monosyllabic words, unless INCLUDE_MONOSYLLABIC_JO_ACCENT
-- is specified. (This is used in conjugation and declension tables.)
-- NOADJ disables special-casing for adjectives in -го, while FORCEADJ forces
-- special-casing for adjectives and disables checking for exceptions
-- (e.g. много). NOSHTO disables special-casing for что and related words.
function export.tr(text, lang, sc, include_monosyllabic_jo_accent, noadj, noshto, forceadj)
	local origtext, subbed_text = export.apply_tr_fixes(text, noadj, noshto, forceadj)
	return export.tr_after_fixes(subbed_text, include_monosyllabic_jo_accent)
end

-- translit with various special-case substitutions; NOADJ disables
-- special-casing for adjectives in -го, while FORCEADJ forces special-casing
-- for adjectives and disables checking for expections (e.g. много).
-- NOSHTO disables special-casing for что and related words. SUB is used
-- to implement arbitrary substitutions in the Cyrillic text before other
-- transformations are applied and before translit. It is of the form
-- FROM/TO,FROM/TO,...
function export.tr_sub(text, include_monosyllabic_jo_accent, noadj, noshto, sub,
	forceadj)
	if type(text) == 'table' then -- called directly from a template
		include_monosyllabic_jo_accent = ine(text.args.include_monosyllabic_jo_accent)
		noadj = ine(text.args.noadj)
		noshto = ine(text.args.noshto)
		sub = ine(text.args.sub)
		text = text.args
	end

	if sub then
		local subs = rsplit(sub, ",")
		for _, subpair in ipairs(subs) do
			local subsplit = rsplit(subpair, "/")
			text = rsub(text, subsplit, subsplit)
		end
	end

	return export.tr(text, nil, nil, include_monosyllabic_jo_accent, noadj, noshto, forceadj)
end

--for adjectives, pronouns
function export.tr_adj(text, include_monosyllabic_jo_accent)
	if type(text) == 'table' then -- called directly from a template
		include_monosyllabic_jo_accent = ine(text.args.include_monosyllabic_jo_accent)
		text = text.args
	end

	-- we have to include "forceadj" because typically when tr_adj() is called
	-- from the noun or adjective modules, it's called with suffix ого, which
	-- would otherwise trigger the exceptional case and be transliterated as ogo
	return export.tr(text, nil, nil, include_monosyllabic_jo_accent, false,
		"noshto", "forceadj")
end

return export

-- For Vim, so we get 4-space tabs
-- vim: set ts=4 sw=4 noet: