ข้ามไปเนื้อหา

มอดูล:de-verb

จาก วิกิพจนานุกรม พจนานุกรมเสรี

This module implements {{de-conj}} and {{de-conj-table}}. It is not meant to be invoked directly from user code. If you have questions about it, please contact User:Benwing2.


local export = {}


--[=[

Authorship: Ben Wing <benwing2>

]=]

--[=[

TERMINOLOGY:

-- "slot" = A particular combination of tense/mood/person/number/etc.
	 Example slot names for verbs are "pres_1s" (present first singular) and
	 "subc_subii_3p" (subordinate-clause subjunctive II third plural).
	 Each slot is filled with zero or more forms.

-- "form" = The conjugated German form representing the value of a given slot.

-- "lemma" = The dictionary form of a given German term. For German, always the infinitive.
]=]

--[=[

FIXME:

1. Handle fensterln. (DONE)
2. Handle spie- past tense of speien. (DONE)
3. Make sure rathen, verheirathen work. (DONE)
4. Modify einfix to better handle managen, framen. (DONE)
5. Use variant codes so variant imperatives with and without -e match up in conjoined verbs e.g. [[ausschneiden und einfügen]].
6. In conjoined verbs e.g. [[ausschneiden und einfügen]], don't repeat auxiliaries.
--]=]

local lang = require("Module:languages").getByCode("de")
local m_string_utilities = require("Module:string utilities")
local m_links = require("Module:links")
local m_table = require("Module:table")
local iut = require("Module:inflection utilities")

local rfind = mw.ustring.find
local rmatch = mw.ustring.match
local rsplit = mw.text.split

local function link_term(term, face)
	return m_links.full_link({ lang = lang, term = term }, face)
end


local vowel = "aeiouyäëïöüÿáéíóúýàèìòùỳâêîôûŷãẽĩõũỹ"
local vowel_c = "[" .. vowel .. "]"
local not_vowel_c = "[^" .. vowel .. "]"

local function ends_in_dt(stem)
	return stem:find("[dt]h?$")
end

local inseparable_prefixes = {
	"be", "emp", "ent", "er", "ge", "miss", "miß", "ver", "zer",
	-- can also be separable
	"durch", "hinter", "über", "um", "unter", "voll", "wider", "wieder",
}

local wuerde = link_term("würde", "term")
local past_subjunctive_single_word_preferred = m_table.listToSet {
	"haben", "sein", "können", "müssen", "dürfen", "mögen", "sollen", "wollen", "werden"
}
local past_subjunctive_single_word_preferred_footnote = "[แนะนำให้ใช้, เลี่ยงรูปประกอบ " .. wuerde .. "]"

local past_subjunctive_single_word_often_used = m_table.listToSet {
	"bleiben", "brauchen", "bringen", "fallen", "finden", "geben", "gehen", "halten", "heißen", "heissen",
	"kommen", "laufen", "lassen", "liegen", "nehmen", "sehen", "stehen", "tun", "thun", "wissen"
}
local past_subjunctive_single_word_often_used_footnote = "[พบได้ทั้งรูปนี้และรูปประกอบ " .. wuerde .. "]"

local past_subjunctive_single_word_rare_footnote = "[พบได้น้อยมากยกเว้นในบริบทที่เป็นทางการมาก, แนะนำให้ใช้รูปประกอบ " .. wuerde .. "]"

local schst_footnote = "[ใช้ได้ในลักษณะเป็นทางการในออสเตรีย, ใช้ในภาษาปากในทุกพื้นที่ที่พูดภาษาเยอรมัน]"


local all_persons_numbers = {
	["1s"] = "1|s",
	["2s"] = "2|s",
	["3s"] = "3|s",
	["1p"] = "1|p",
	["2p"] = "2|p",
	["3p"] = "3|p",
}

local person_number_list = { "1s", "2s", "3s", "1p", "2p", "3p", }
local persnum_to_index = {}
for k, v in pairs(person_number_list) do
	persnum_to_index[v] = k
end
local imp_person_number_list = { "2s", "2p", }

local verb_slots_basic = {
	{"infinitive", "inf"},
	{"infinitive_linked", "inf"},
	{"pres_part", "pres|part"},
	{"perf_part", "perf|part"},
	{"zu_infinitive", "zu"}, -- will be handled specially by [[Module:accel/de]]
	{"aux", "-"},
}

local verb_slots_subordinate_clause = {
}

local verb_slots_composed = {
}

-- Add entries for a slot with person/number variants.
-- `verb_slots` is the table to add to.
-- `slot_prefix` is the prefix of the slot, typically specifying the tense/aspect.
-- `tag_suffix` is the set of inflection tags to add after the person/number tags,
-- or "-" to use "-" as the inflection tags (which indicates that no accelerator entry
-- should be generated).
local function add_slot_personal(verb_slots, slot_prefix, tag_suffix)
	for persnum, persnum_tag in pairs(all_persons_numbers) do
		local slot = slot_prefix .. "_" .. persnum
		if tag_suffix == "-" then
			table.insert(verb_slots, {slot, "-"})
		else
			table.insert(verb_slots, {slot, persnum_tag .. "|" .. tag_suffix})
		end
	end
end

add_slot_personal(verb_slots_basic, "pres", "pres")
add_slot_personal(verb_slots_basic, "subi", "sub|I")
add_slot_personal(verb_slots_basic, "pret", "pret")
add_slot_personal(verb_slots_basic, "subii", "sub|II")
table.insert(verb_slots_basic, {"imp_2s", "s|imp"})
table.insert(verb_slots_basic, {"imp_2p", "p|imp"})
add_slot_personal(verb_slots_subordinate_clause, "subc_pres", "dep|pres")
add_slot_personal(verb_slots_subordinate_clause, "subc_subi", "dep|sub|I")
add_slot_personal(verb_slots_subordinate_clause, "subc_pret", "dep|pret")
add_slot_personal(verb_slots_subordinate_clause, "subc_subii", "dep|sub|II")
add_slot_personal(verb_slots_composed, "perf_ind", "-")
add_slot_personal(verb_slots_composed, "perf_sub", "-")
add_slot_personal(verb_slots_composed, "plup_ind", "-")
add_slot_personal(verb_slots_composed, "plup_sub", "-")
table.insert(verb_slots_composed, {"futi_inf", "-"})
add_slot_personal(verb_slots_composed, "futi_subi", "-")
add_slot_personal(verb_slots_composed, "futi_ind", "-")
add_slot_personal(verb_slots_composed, "futi_subii", "-")
table.insert(verb_slots_composed, {"futii_inf", "-"})
add_slot_personal(verb_slots_composed, "futii_subi", "-")
add_slot_personal(verb_slots_composed, "futii_ind", "-")
add_slot_personal(verb_slots_composed, "futii_subii", "-")


local all_verb_slots = {}
for _, slot_and_accel in ipairs(verb_slots_basic) do
	table.insert(all_verb_slots, slot_and_accel)
end
for _, slot_and_accel in ipairs(verb_slots_subordinate_clause) do
	table.insert(all_verb_slots, slot_and_accel)
end
for _, slot_and_accel in ipairs(verb_slots_composed) do
	table.insert(all_verb_slots, slot_and_accel)
end


local pronouns = { "ich", "du", "er", "wir", "ihr", "sie", }

irreg_verbs = {
	["haben"] = {
		["pres"] = { "habe", "hast", "hat", "haben", "habt", "haben", },
		["pret"] = { "hatte", "hattest", "hatte", "hatten", "hattet", "hatten", },
		["subi"] = { "habe", "habest", "habe", "haben", "habet", "haben", },
		["subii"] = { "hätte", "hättest", "hätte", "hätten", "hättet", "hätten", },
		["imp"] = { {"hab", "habe"}, "habt", },
		["presp"] = "habend",
		["pp"] = "habt",
	},
	["sein"] = {
		["pres"] = { "bin", "bist", "ist", "sind", "seid", "sind", },
		["pret"] = { "war", "warst", "war", "waren", "wart", "waren", },
		["subi"] = { "sei", {"seist", "seiest"}, "sei", "seien", "seiet", "seien", },
		["subii"] = { "wäre", {"wärst", "wärest"}, "wäre", "wären", {"wärt", "wäret"}, "wären", },
		["imp"] = { "sei", "seid", },
		["presp"] = "seiend",
		["pp"] = "wesen",
	},
	["tun"] = {
		["pres"] = { "tue", "tust", "tut", "tun", "tut", "tun", },
		["pret"] = { "tat", {"tatest", "tatst"}, "tat", "taten", "tatet", "taten", },
		["subi"] = { "tue", "tuest", "tue", "tuen", "tuet", "tuen", },
		["subii"] = { "täte", "tätest", "täte", "täten", "tätet", "täten", },
		["imp"] = { {"tu", "tue"}, "tut", },
		["presp"] = "tuend",
		["pp"] = "tan",
	},
	-- FIXME, maybe we should construct this automatically from "tun".
	["thun"] = {
		["pres"] = { "thue", "thust", "thut", "thun", "thut", "thun", },
		["pret"] = { "that", {"thatest", "thatst"}, "that", "thaten", "thatet", "thaten", },
		["subi"] = { "thue", "thuest", "thue", "thuen", "thuet", "thuen", },
		["subii"] = { "thäte", "thätest", "thäte", "thäten", "thätet", "thäten", },
		["imp"] = { {"thu", "thue"}, "thut", },
		["presp"] = "thuend",
		["pp"] = "than",
	},
	["werden"] = {
		["pres"] = { "werde", "wirst", "wird", "werden", "werdet", "werden", },
		["pret"] = {
			{"wurde", {form = "ward", footnotes = {"[โบราณ]"}}},
			{"wurdest", {form = "wardst", footnotes = {"[โบราณ]"}}},
			{"wurde", {form = "ward", footnotes = {"[โบราณ]"}}},
			"wurden",
			"wurdet",
			"wurden",
		},
		["subi"] = { "werde", "werdest", "werde", "werden", "werdet", "werden", },
		["subii"] = { "würde", "würdest", "würde", "würden", "würdet", "würden", },
		["imp"] = { {"werd", "werde"}, "werdet", },
		["presp"] = "werdend",
		["pp"] = "worden",
	},
}


local sein_forms = {
	["sein"] = {"mein", "dein", "sein", "unser", "euer", "ihr"},
	["seine"] = {"meine", "deine", "seine", "unsere", "eure", "ihre"},
	["seinen"] = {"meinen", "deinen", "seinen", "unseren", "euren", "ihren"},
	["seinem"] = {"meinem", "deinem", "seinem", "unserem", "eurem", "ihrem"},
	["seiner"] = {"meiner", "deiner", "seiner", "unserer", "eurer", "ihrer"},
	["seines"] = {"meines", "deines", "seines", "unseses", "eures", "ihres"},
}


local sich_forms = {
	["accpron"] = {"mich", "dich", "sich", "uns", "euch", "sich"},
	["datpron"] = {"mir", "dir", "sich", "uns", "euch", "sich"},
}


local function skip_slot(base, slot)
	if not slot:find("[123]") then
		-- Don't skip non-personal slots.
		return false
	end

	if base.nofinite then
		return true
	end

	if base.only3s and not slot:find("3s") or
		base.only3sp and not slot:find("3[sp]") then
		return true
	end

	return false
end


local function strip_spaces(text)
	return text:gsub("^%s*(.-)%s*", "%1")
end


local function escape_sein_sich_indicators(arg1)
	if not arg1:find("pron>") then
		return arg1
	end
	local segments = iut.parse_balanced_segment_run(arg1, "<", ">")
	-- Loop over every other segment. The even-numbered segments are angle-bracket specs while
	-- the odd-numbered segments are the text between them.
	for i = 2, #segments - 1, 2 do
		if segments[i] == "<accpron>" then
			segments[i] = "⦃⦃accpron⦄⦄"
		elseif segments[i] == "<datpron>" then
			segments[i] = "⦃⦃datpron⦄⦄"
		elseif segments[i] == "<pron>" then
			segments[i] = "⦃⦃pron⦄⦄"
		end
	end
	return table.concat(segments)
end


local function undo_escape_form(form)
	-- assign to var to throw away second value
	local newform = form:gsub("⦃⦃", "<"):gsub("⦄⦄", ">")
	return newform
end


local function remove_sein_sich_indicators(form)
	-- assign to var to throw away second value
	local newform = form:gsub("⦃⦃.-⦄⦄", "")
	return newform
end


local function replace_sein_sich_indicators(slot, form)
	if not form:find("⦃") then
		return form
	end
	local persnum = slot:match("^.*_([123][sp])$")
	local index
	if persnum then
		index = persnum_to_index[persnum]
	else
		index = 3
	end
	form = form:gsub("sich(%]*)⦃⦃accpron⦄⦄", function(brackets)
		return sich_forms.accpron[index] .. brackets
	end)
	form = form:gsub("sich(%]*)⦃⦃datpron⦄⦄", function(brackets)
		return sich_forms.datpron[index] .. brackets
	end)
	form = form:gsub("(seine?[mnrs]?)(%]*)⦃⦃pron⦄⦄", function(sein_form, brackets)
		if sein_forms[sein_form] then
			return sein_forms[sein_form][index] .. brackets
		else
			error("Unrecognized sein-form '" .. sein_form .. "' in slot " .. slot .. ": " .. undo_escape_form(form))
		end
	end)
	form = form:gsub("sein%]%](e?[mnrs]?)⦃⦃pron⦄⦄", function(sein_ending, brackets)
		local sein_form = "sein" .. sein_ending
		if sein_forms[sein_form] then
			return sein_forms["sein"][index] .. "|" .. sein_forms[sein_form][index] .. "]]"
		else
			error("Unrecognized sein-form '" .. sein_form .. "' in slot " .. slot .. ": " .. undo_escape_form(form))
		end
	end)
	if form:find("⦃⦃") or form:find("⦄⦄") then
		error("Unrecognized pronoun substitution in slot " .. slot .. ": " .. undo_escape_form(form))
	end
	return form
end


local function combine_stem_ending(slot, stem, ending)
	local ending_with_pound = ending .. "#"
	if ending_with_pound:find("^st[ #]") then
		if rfind(stem, "[sxzß]$") then
			ending = ending:gsub("^s", "")
		elseif stem:find("st$") then
			-- bersten
			ending = ending:gsub("^st", "")
		end
	elseif ending_with_pound:find("^t[ #]") and stem:find("th?$") then
		ending = ending:gsub("^t", "")
	end
	return replace_sein_sich_indicators(slot, stem .. ending)
end


local function add(base, slot, stems, endings, footnotes)
	if skip_slot(base, slot) then
		return
	end
	local function do_combine_stem_ending(stem, ending)
		return combine_stem_ending(slot, stem, ending)
	end
	iut.add_forms(base.forms, slot, stems, endings, do_combine_stem_ending, nil, nil,
		iut.combine_footnotes(footnotes, base.all_footnotes))
end


local function add_multi(base, slot, stems_and_endings, footnotes)
	if skip_slot(base, slot) then
		return
	end
	local function do_combine_stem_ending(stem, ending)
		return combine_stem_ending(slot, stem, ending)
	end
	iut.add_multiple_forms(base.forms, slot, stems_and_endings, do_combine_stem_ending, nil, nil,
		iut.combine_footnotes(footnotes, base.all_footnotes))
end


local function add3(base, slot, stems1, stems2, endings, footnotes)
	return add_multi(base, slot, {stems1, stems2, endings})
end


local function add4(base, slot, stems1, stems2, stems3, endings, footnotes)
	return add_multi(base, slot, {stems1, stems2, stems3, endings})
end


local function add5(base, slot, stems1, stems2, stems3, stems4, endings, footnotes)
	return add_multi(base, slot, {stems1, stems2, stems3, stems4, endings})
end


local function get_subii_note(base)
	if base.from_headword then
		-- don't include autogenerated note if from_headword, so it doesn't end up in the headword qualifiers
		return nil
	end
	if past_subjunctive_single_word_preferred[base.base_verb] then
		return {past_subjunctive_single_word_preferred_footnote}
	elseif past_subjunctive_single_word_often_used[base.base_verb] then
		return {past_subjunctive_single_word_often_used_footnote}
	end
	return {past_subjunctive_single_word_rare_footnote}
end


local function add_zu_infinitive(base)
	if base.any_pre_pref then
		local zu
		if base.pre_pref == "" or base.pre_pref:find(" $") then
			zu = "zu "
		else
			zu = "zu"
		end
		add(base, "zu_infinitive", base.pre_pref .. zu, base.bare_infinitive)
	end
end


local function add_present(base, pretpres)
	local stems = base.infstem
	local stems23 = base.pres_23 or stems

	local function doadd(slot_pref, form_pref)
		-- Do forms based off the infinitive stem.
		for _, stemform in ipairs(stems) do
			local prefixed_stem = form_pref .. stemform.form
			local syncopated_stem = base.unstressed_el_er and prefixed_stem:gsub("e([lr])$", "%1")

			local function addit(slot, stem, ending, footnotes)
				add(base, slot_pref .. slot, stem, ending .. (slot_pref == "" and base.post_pref or ""),
					iut.combine_footnotes(footnotes, stemform.footnotes))
			end

			-- first singular present indicative
			if not pretpres then -- pres_1s for preterite-present verbs used pres23 stem
				if base.unstressed_el_er then
					addit("pres_1s", syncopated_stem, "e")
					addit("pres_1s", prefixed_stem, "e")
					addit("pres_1s", prefixed_stem, "")
				else
					addit("pres_1s", prefixed_stem, "e")
				end
			end

			-- plural present indicative
			local e_in_2p = ends_in_dt(prefixed_stem) or base.unstressed_e_infix
			if e_in_2p then
				addit("pres_2p", prefixed_stem, "et")
			else
				addit("pres_2p", prefixed_stem, "t")
			end
			if base.unstressed_el_er or base.unstressed_erl then
				addit("pres_1p", prefixed_stem, "n")
				addit("pres_3p", prefixed_stem, "n")
			else
				addit("pres_1p", prefixed_stem, "en")
				addit("pres_3p", prefixed_stem, "en")
			end

			-- subjunctive I
			addit("subi_1s", prefixed_stem, "e")
			addit("subi_2s", prefixed_stem, "est")
			addit("subi_3s", prefixed_stem, "e")
			addit("subi_2p", prefixed_stem, "et")
			if base.unstressed_el_er then
				addit("subi_1s", syncopated_stem, "e")
				addit("subi_2s", syncopated_stem, "est")
				addit("subi_3s", syncopated_stem, "e")
				addit("subi_2p", syncopated_stem, "et")
			end
			if base.unstressed_el_er then
				addit("subi_1p", prefixed_stem, "n")
				addit("subi_3p", prefixed_stem, "n")
			else
				addit("subi_1p", prefixed_stem, "en")
				addit("subi_3p", prefixed_stem, "en")
			end

			-- imperative plural; most preterite-present verbs don't have it, and [[wissen]] is handled below
			if slot_pref == "" and not pretpres then
				if e_in_2p then
					addit("imp_2p", prefixed_stem, "et")
				else
					addit("imp_2p", prefixed_stem, "t")
				end
			end
		end

		-- Do forms based off of pres23 stem (also includes pres_1sg in preterite-present verbs).
		for _, stem23form in ipairs(stems23) do
			local prefixed_stem23 = form_pref .. stem23form.form

			local function addit(slot, stem, ending, footnotes)
				add(base, slot_pref .. slot, stem, ending .. (slot_pref == "" and base.post_pref or ""),
					iut.combine_footnotes(footnotes, stem23form.footnotes))
			end

			if pretpres then
				-- Totally different code for preterite-present singular and imperative.
				addit("pres_1s", prefixed_stem23, "")
				addit("pres_2s", prefixed_stem23, "st")
				addit("pres_3s", prefixed_stem23, "")
				if slot_pref == "" and base.base_verb == "wissen" then
					-- Only [[wissen]] among the preterite-present verbs has an imperative.
					addit("imp_2s", "wiss", "e")
					-- dewikt mentions 'wisset' as an alternative, but not in Duden
					addit("imp_2p", "wiss", "t")
				end
			else
				-- Normal code path.
				local stem23_is_same_as_stem
				for _, stemform in ipairs(stems) do
					if stemform.form == stem23form.form then
						stem23_is_same_as_stem = true
						break
					end
				end

				-- present 2/3 singular
				local e_in_23s = base.unstressed_e_infix or stem23_is_same_as_stem and ends_in_dt(prefixed_stem23)
				if e_in_23s then
					addit("pres_2s", prefixed_stem23, "est")
					addit("pres_3s", prefixed_stem23, "et")
				else
					addit("pres_2s", prefixed_stem23, "st")
					if prefixed_stem23:find("sch$") then
						addit("pres_2s", prefixed_stem23, "t",
							not base.from_headword and {schst_footnote} or nil)
					end
					addit("pres_3s", prefixed_stem23, "t")
				end

				-- imperative singular; this may or may not be based off the pres23 stem
				-- Specifically, if the pres23 stem is different from the infinitive stem and does not have an ä or ö
				-- in it ([[fahren]], er [[fährt]] but imperative [[fahr]]/[[fahre]]; [[stoßen]], er [[stößt]] but
				-- imperative [[stoß]]/[[stoße]]), use it. Don't add -e unless '.longimp' is given (for [[sehen]], with
				-- imperatives [[sieh]] and [[siehe]]; but normally [[geben]] with imperative only [[gib]], similarly
				-- for [[treten]], [[gelten]], [[bergen]], [[etc.]]). In all other cases, use the infinitive stem, and
				-- under normal circumstances include two variants, without -e and with -e. We include only the variant
				-- with -e if '.e' is given ([[atmen]], [[zeichnen]], etc.), and we include only the variant without -e
				-- if '.shortimp' is given (unclear if this is needed for any verb).
				if slot_pref == "" then
					if not rfind(prefixed_stem23, "[äö]") and not stem23_is_same_as_stem then
						addit("imp_2s", prefixed_stem23, "")
						if base.longimp then
							addit("imp_2s", prefixed_stem23, "e")
						end
					else
						assert(form_pref == "")
						for _, stemform in ipairs(stems) do
							local function addimp(stem, ending)
								add(base, "imp_2s", stem, ending .. base.post_pref, stemform.footnotes)
							end
							local stem = stemform.form
							if base.unstressed_el_er then
								local syncopated_stem = stem:gsub("e([lr])$", "%1")
								addimp(syncopated_stem, "e")
							end
							if base.unstressed_e_infix then
								addimp(stem, "e")
							elseif base.shortimp then
								addimp(stem, "")
							else
								addimp(stem, "")
								addimp(stem, "e")
							end
						end
					end
				end
			end
		end
	end

	-- Do the basic forms
	doadd("", "")
	-- Also do the subordinate clause forms if any alternants have a prefix.
	if base.any_pre_pref then
		doadd("subc_", base.pre_pref)
	end

	-- Do the miscellaneous non-finite forms
	add3(base, "pres_part", base.pre_pref, base.bare_infinitive, "d")
	add_zu_infinitive(base)
end


-- Add the past forms and/or subjunctive II forms. `is_past` should be true if we're conjugating the past tense
-- (not the past subjunctive), in which case the stems in `stem` should be the actual 1s/3s forms, ending in -te
-- or -de if weak. If `is_past` is false, we're conjugating the past subjunctive. In that case, the stems passed
-- in should include -e if that needs to be preserved in all forms (specifically, for [[spie]] past of [[speien]];
-- syncopated #'spist' is not allowed). (In practice, we chop off the final -e of explicitly user-specified past
-- subjunctive forms and otherwise pass in the actual past tense. This means that user-specified past subjunctives
-- in -te (e.g. [[brächte]]) will be passed in as 'brächt' but past subjunctives defaulted from the past tense
-- will be passed in e.g. as 'lachte'. This works because for stems in -t and -d, the result is the same.)
local function add_past_or_subii(base, slot_pref, stems, is_past)
	local function doadd(full_slot_pref, form_pref)
		for _, stemform in ipairs(stems) do
			local prefixed_stem = form_pref .. stemform.form
			local ends_in_e = prefixed_stem:find("e$")
			-- Normally, the weak past tense ends in -te and the strong past does not end in -e. An exception is
			-- [[spie]] (past tense of [[speien]]), which will be treated as weak (superseded spellings like
			-- 'spieen' will not be generated).
			local is_strong = is_past and not ends_in_e
			prefixed_stem = prefixed_stem:gsub("e$", "")
			local stem_ends_in_dt = ends_in_dt(prefixed_stem)

			local function addit(slot, ending, footnotes)
				local subii_footnotes
				if full_slot_pref:find("subii") then
					subii_footnotes = get_subii_note(base)
				end
				add(base, full_slot_pref .. slot, prefixed_stem, ending .. (full_slot_pref == slot_pref and base.post_pref or ""),
					iut.combine_footnotes(subii_footnotes, iut.combine_footnotes(footnotes, stemform.footnotes)))
			end

			if is_strong then
				addit("1s", "")
				addit("3s", "")
				if not stem_ends_in_dt then
					if rfind(prefixed_stem, "[sxzß]$") then
						addit("2s", "est")
					end
					addit("2s", "st")
					addit("2p", "t")
				else
					-- bitten -> batest or batst, similarly for laden, raten, etc.
					addit("2s", "est")
					if not prefixed_stem:find("st$") then
						-- bersten
						addit("2s", "st")
					end
					addit("2p", "et")
				end
			else
				-- Weak past and past subjunctive have same endings when ending in -d or -t (always the case for weak past).
				addit("1s", "e")
				addit("3s", "e")
				if not stem_ends_in_dt and not ends_in_e then
					addit("2s", "est") -- more formal; FIXME: should we footnote this?
					addit("2p", "et") -- more formal; FIXME: should we footnote this?
					addit("2s", "st")
					addit("2p", "t")
				else
					addit("2s", "est")
					addit("2p", "et")
				end
			end

			-- Both pasts, as well as past subjunctive, have same endings in 1p and 3p.
			addit("1p", "en")
			addit("3p", "en")
		end
	end

	-- Do the basic forms
	doadd(slot_pref, "")
	-- Also do the subordinate clause forms if any alternants have a prefix.
	if base.any_pre_pref then
		doadd("subc_" .. slot_pref, base.pre_pref)
	end
end


local conjs = {}
local conjprops = {}

conjs["normal"] = function(base)
	add_present(base)
	add_past_or_subii(base, "pret_", base.past, "past")
	add_past_or_subii(base, "subii_", base.past_sub or base.past)
end


conjs["pretpres"] = function(base)
	add_present(base, "pretpres")
	add_past_or_subii(base, "pret_", base.past, "past")
	add_past_or_subii(base, "subii_", base.past_sub or base.past)
end


conjs["irreg"] = function(base)
	local function doadd(slot_pref)
		local function addit(slot, forms, footnotes)
			if slot_pref == "" then
				add3(base, slot, base.insep_prefix, forms, base.post_pref, footnotes)
			else
				add(base, slot_pref .. slot, base.pre_pref .. base.insep_prefix, forms, footnotes)
			end
		end

		-- Do present, preterite, subjunctive I and II.
		for _, slot_tense in ipairs({"pres", "pret", "subi", "subii"}) do
			for index, forms in ipairs(base.irregverbobj[slot_tense]) do
				local persnum = person_number_list[index]
				local footnotes
				if slot_tense == "subii" then
					footnotes = get_subii_note(base)
				end
				addit(slot_tense .. "_" .. persnum, forms, footnotes)
			end
		end

		-- Do imperative.
		if slot_pref == "" then
			for index, forms in ipairs(base.irregverbobj["imp"]) do
				local persnum = imp_person_number_list[index]
				addit("imp_" .. persnum, forms)
			end
		end
	end

	-- Do the basic forms
	doadd("")
	-- Also do the subordinate clause forms if any alternants have a prefix.
	if base.any_pre_pref then
		doadd("subc_")
	end

	-- Do the miscellaneous non-finite forms
	iut.add_multiple_forms(base, "pp", {base.ge_prefix, base.insep_prefix, base.irregverbobj["pp"]},
		-- We don't want to use combine_stem_ending because we want sein-sich indicators to be left alone,
		-- so they get replaced later when constructing composed forms.
		function(stem, ending) return stem .. ending end)
	add(base, "perf_part", base.pre_pref, base.pp)
	-- only [[werden]] by itself; not [[loswerden]], [[fertigwerden]], etc.
	if base.lemma == "werden" then
		iut.insert_form(base.forms, "perf_part", {form = "worden", footnotes = {"[เมื่อเป็นคำกริยานุเคราะห์]"}})
	end
	add(base, "pres_part", base.pre_pref .. base.insep_prefix, base.irregverbobj["presp"])
	add_zu_infinitive(base)
end


local function add_composed_forms(base)
	local forms = base.forms

	local function add_composed(tense_mood, index, persnum, auxforms, participle, suffix, footnotes)
		local pers_auxforms = iut.convert_to_general_list_form(auxforms[index])
		local linked_pers_auxforms = iut.map_forms(pers_auxforms, function(form) return "[[" .. form .. "]] " end)
		add4(base, tense_mood .. "_" .. persnum, linked_pers_auxforms, "[[" .. base.pre_pref, participle, "]]" .. suffix, footnotes)
	end

	local function add_composed_perf(tense_mood, index, persnum, haben_auxforms, sein_auxforms, haben_suffix, sein_suffix)
		for _, auxform in ipairs(base.aux) do
			if auxform.form == "haben" then
				add_composed(tense_mood, index, persnum, haben_auxforms, base.pp, haben_suffix, auxform.footnotes)
			end
			if auxform.form == "sein" then
				add_composed(tense_mood, index, persnum, sein_auxforms, base.pp, sein_suffix, auxform.footnotes)
			end
		end
	end

	local haben_forms = irreg_verbs["haben"]
	local sein_forms = irreg_verbs["sein"]
	local werden_forms = irreg_verbs["werden"]
	for index, persnum in ipairs(person_number_list) do
		add_composed_perf("perf_ind", index, persnum, haben_forms["pres"], sein_forms["pres"], "", "")
		add_composed_perf("perf_sub", index, persnum, haben_forms["subi"], sein_forms["subi"], "", "")
		add_composed_perf("plup_ind", index, persnum, haben_forms["pret"], sein_forms["pret"], "", "")
		add_composed_perf("plup_sub", index, persnum, haben_forms["subii"], sein_forms["subii"], "", "")
		for _, mood in ipairs({"ind", "subi", "subii"}) do
			local tense = mood == "ind" and "pres" or mood
			add_composed("futi_" .. mood, index, persnum, werden_forms[tense], base.bare_infinitive, "")
			add_composed_perf("futii_" .. mood, index, persnum, werden_forms[tense], werden_forms[tense], " [[haben]]", " [[sein]]")
		end
	end

	add3(base, "futi_inf", "[[" .. base.pre_pref, base.bare_infinitive, "]] [[werden]]")
	add5(base, "futii_inf", "[[" .. base.pre_pref, base.pp, "]] [[", base.aux, "]] [[werden]]")
end


local function handle_derived_slots(base)
	-- Compute linked versions of potential lemma slots, for use in {{de-verb}}.
	-- We substitute the original lemma (before removing links) for forms that
	-- are the same as the lemma, if the original lemma has links.
	for _, slot in ipairs({"infinitive"}) do
		iut.insert_forms(base.forms, slot .. "_linked", iut.map_forms(base.forms[slot], function(form)
			if form == base.lemma and rfind(base.linked_lemma, "%[%[") then
				return base.linked_lemma
			else
				return form
			end
		end))
	end
end


local function conjugate_verb(base)
	if not conjs[base.conj] then
		error("Internal error: Unrecognized conjugation type '" .. base.conj .. "'")
	end
	conjs[base.conj](base)
	add_composed_forms(base)
	-- No overrides implemented currently.
	-- process_slot_overrides(base)
	handle_derived_slots(base)
end


local function parse_indicator_spec(angle_bracket_spec)
	local base = {}
	local function parse_err(msg)
		error(msg .. ": " .. angle_bracket_spec)
	end
	local function fetch_footnotes(separated_group)
		local footnotes
		for j = 2, #separated_group - 1, 2 do
			if separated_group[j + 1] ~= "" then
				parse_err("Extraneous text after bracketed footnotes: '" .. table.concat(separated_group) .. "'")
			end
			if not footnotes then
				footnotes = {}
			end
			table.insert(footnotes, separated_group[j])
		end
		return footnotes
	end

	local function fetch_specs(comma_separated_group, transform_form)
		if not comma_separated_group then
			return {{}}
		end
		local specs = {}
		
		local colon_separated_groups = iut.split_alternating_runs(comma_separated_group, ":")
		for _, colon_separated_group in ipairs(colon_separated_groups) do
			local form = colon_separated_group[1]
			if transform_form then
				form = transform_form(form)
			end
			table.insert(specs, {form = form, footnotes = fetch_footnotes(colon_separated_group)})
		end
		return specs
	end

	local inside = angle_bracket_spec:match("^<(.*)>$")
	assert(inside)
	if inside == "" then
		return base
	end
	local segments = iut.parse_balanced_segment_run(inside, "[", "]")
	local dot_separated_groups = iut.split_alternating_runs(segments, "%.")
	for i, dot_separated_group in ipairs(dot_separated_groups) do
		local comma_separated_groups = iut.split_alternating_runs(dot_separated_group, "%s*[,#]%s*", "preserve splitchar")
		local first_element = comma_separated_groups[1][1]
		if first_element == "haben" or first_element == "sein" then
			for j = 1, #comma_separated_groups, 2 do
				if j > 1 and strip_spaces(comma_separated_groups[j - 1][1]) ~= "," then
					parse_err("Separator of # not allowed with haben or sein")
				end
				local aux = comma_separated_groups[j][1]
				if aux ~= "haben" and aux ~= "sein" then
					parse_err("Unrecognized auxiliary '" .. aux .. "'")
				end
				if base.aux then
					for _, existing_aux in ipairs(base.aux) do
						if existing_aux.form == aux then
							parse_err("Auxiliary '" .. aux .. "' specified twice")
						end
					end
				else
					base.aux = {}
				end
				table.insert(base.aux, {form = aux, footnotes = fetch_footnotes(comma_separated_groups[j])})
			end
		elseif first_element == "-ge" or first_element == "+ge" then
			for j = 1, #comma_separated_groups, 2 do
				if j > 1 and strip_spaces(comma_separated_groups[j - 1][1]) ~= "," then
					parse_err("Separator of # not allowed with +ge or -ge")
				end
				local prefix = comma_separated_groups[j][1]
				if prefix ~= "+ge" and prefix ~= "-ge" then
					parse_err("Unrecognized ge- prefix '" .. prefix .. "'")
				end
				local ge_prefix
				if prefix == "+ge" then
					ge_prefix = "ge"
				else
					ge_prefix = ""
				end
				if base.ge_prefix then
					for _, existing_prefix in ipairs(base.ge_prefix) do
						if existing_prefix.form == ge_prefix then
							parse_err("Ge- prefix '" .. prefix .. "' specified twice")
						end
					end
				else
					base.ge_prefix = {}
				end
				table.insert(base.ge_prefix, {form = ge_prefix, footnotes = fetch_footnotes(comma_separated_groups[j])})
			end
		elseif #comma_separated_groups > 1 then
			-- principal parts specified
			if base.past then
				parse_err("Can't specify principal parts twice")
			end
			local parts = {}
			assert(#comma_separated_groups[2] == 1)
			local past_index
			local first_separator = strip_spaces(comma_separated_groups[2][1])
			if first_separator == "#" then
				-- present 3rd singular specified
				base.pres_23 = fetch_specs(comma_separated_groups[1], function(form)
					local stem
					if base.conj == "pretpres" then
						stem = form
					else
						stem = form:match("^(.-)%-$")
						if not stem then
							stem = form:match("^(.-)e?t$")
						end
					end
					if stem then
						return stem
					else
						parse_err("Present 3sg form '" .. form .. "' should end in - (for the stem) or -t")
					end
				end)
				past_index = 3
			else
				past_index = 1
			end

			base.past = fetch_specs(comma_separated_groups[past_index], function(form)
				return form
			end)

			if #comma_separated_groups < past_index + 2 then
				parse_err("Missing past participle spec")
			end
			assert(#comma_separated_groups[past_index + 1] == 1)
			if strip_spaces(comma_separated_groups[past_index + 1][1]) ~= "," then
				parse_err("Only first separator can be a #")
			end
			base.pp = fetch_specs(comma_separated_groups[past_index + 2], function(form)
				if form:find("e[nd]$") or form:find("t$") then
					return form
				else
					parse_err("Past participle '" .. form .. "' should end in -en, -t, or -ed")
				end
			end)

			if #comma_separated_groups > past_index + 2 then
				assert(#comma_separated_groups[past_index + 3] == 1)
				if strip_spaces(comma_separated_groups[past_index + 3][1]) ~= "," then
					parse_err("Only first separator can be a #")
				end
				base.past_sub = fetch_specs(comma_separated_groups[past_index + 4], function(form)
					local stem = form:match("^(.-)e$")
					if not stem then
						parse_err("Past subjunctive '" .. form .. "' should end in -e")
					end
					return stem
				end)
				if #comma_separated_groups > past_index + 4 then
					parse_err("Too many specs given")
				end
			end
		elseif first_element == "pretpres" or first_element == "irreg" then
			if #comma_separated_groups[1] > 1 then
				parse_err("No footnotes allowed with '" .. first_element .. "' spec")
			end
			if base.conj then
				parse_err("Conjugation specified as '" .. first_element .. "' but already specified or autodetermined as '" .. base.conj .. "'")
			end
			base.conj = first_element
		elseif first_element == "einfix" or first_element == "-einfix" then
			if #comma_separated_groups[1] > 1 then
				parse_err("No footnotes allowed with '" .. first_element .. "' spec")
			end
			base.unstressed_e_infix = first_element == "einfix"
		elseif first_element == "shortimp" or first_element == "longimp" or
			first_element == "only3s" or first_element == "only3sp" or
			first_element == "nofinite" then
			if #comma_separated_groups[1] > 1 then
				parse_err("No footnotes allowed with '" .. first_element .. "' spec")
			end
			base[first_element] = true
		elseif first_element == "" or first_element == "inf" then
			local footnotes = fetch_footnotes(comma_separated_groups[1])
			if not footnotes then
				parse_err("Empty spec and 'inf' spec without footnotes not allowed")
			end
			if first_element == "inf" then
				base.infstem_footnotes = footnotes
			else
				base.all_footnotes = footnotes
			end
		else
			parse_err("Unrecognized spec '" .. comma_separated_groups[1][1] .. "'")
		end
	end

	return base
end


-- Normalize all lemmas, splitting off separable prefixes and substituting the pagename for blank lemmas.
local function normalize_all_lemmas(alternant_multiword_spec, from_headword)
	local any_pre_pref
	iut.map_word_specs(alternant_multiword_spec, function(base)
		if base.lemma == "" then
			local PAGENAME = mw.title.getCurrentTitle().text
			base.lemma = PAGENAME
		end
		if base.lemma:find("_") and not base.lemma:find("%[%[") then
			-- If lemma is multiword and has no links, add links automatically.
			base.lemma= "[[" .. base.lemma:gsub("_", "]]_[[") .. "]]"
		end
		base.orig_lemma = base.lemma
		base.orig_lemma_no_links = m_links.remove_links(base.lemma)
		-- Normalize the linked lemma by removing dot, underscore, and <pron> and such indicators.
		base.linked_lemma = remove_sein_sich_indicators(base.lemma:gsub("%.", ""):gsub("_", " "))
		base.lemma = m_links.remove_links(base.linked_lemma)
		local lemma = base.orig_lemma_no_links
		base.pre_pref, base.post_pref = "", ""
		local prefix, verb = lemma:match("^(.*)_(.-)$")
		if prefix then
			prefix = prefix:gsub("_", " ") -- in case of multiple preceding words
			base.pre_pref = base.pre_pref .. prefix .. " "
			base.post_pref = base.post_pref .. " " .. prefix
		else
			verb = lemma
		end
		prefix, base.base_verb = verb:match("^(.*)%.(.-)$")
		if prefix then
			-- There may be multiple separable prefixes (e.g. [[wiedergutmachen]], ich mache wieder gut)
			base.pre_pref = base.pre_pref .. prefix:gsub("%.", "")
			base.post_pref = base.post_pref .. " " .. prefix:gsub("%.", " ")
		else
			base.base_verb = verb
		end
		if base.pre_pref ~= "" then
			any_pre_pref = true
		end
		if base.only3s then
			alternant_multiword_spec.only3s = true
		end
		if base.only3sp then
			alternant_multiword_spec.only3sp = true
		end
		-- Remove <pron> indicators and such.
		local reconstructed_lemma = remove_sein_sich_indicators(base.pre_pref .. base.base_verb)
		if reconstructed_lemma ~= base.lemma then
			error("Internal error: Raw lemma '" .. base.lemma .. "' differs from reconstructed lemma '" .. reconstructed_lemma .. "'")
		end
		base.from_headword = from_headword
	end)
	if any_pre_pref then
		iut.map_word_specs(alternant_multiword_spec, function(base)
			base.any_pre_pref = true
		end)
	end
	if alternant_multiword_spec.only3s then
		iut.map_word_specs(alternant_multiword_spec, function(base)
			if not base.only3s then
				error("If some alternants specify 'only3s', all must")
			end
		end)
	end
	if alternant_multiword_spec.only3sp then
		iut.map_word_specs(alternant_multiword_spec, function(base)
			if not base.only3sp then
				error("If some alternants specify 'only3sp', all must")
			end
		end)
	end
end


local function detect_verb_type(base, verb_types)
	local this_verb_types = {}

	local function set_verb_type()
		base.verb_types = this_verb_types
	
		if verb_types then
			for _, verb_type in ipairs(this_verb_types) do
				m_table.insertIfNot(verb_types, verb_type)
			end
		end
	end

	if base.conj == "pretpres" then
		m_table.insertIfNot(this_verb_types, "pretpres")
		set_verb_type()
		return
	elseif base.conj == "irreg" then
		m_table.insertIfNot(this_verb_types, "irreg")
		set_verb_type()
		return
	end

	local infstem = m_table.deepcopy(base.infstem)
	local past = m_table.deepcopy(base.past)
	local pp = m_table.deepcopy(base.pp)

	local function matches_forms(forms, expected, ending_to_chop)
		expected = expected:gsub("C", not_vowel_c) .. "$"
		local seen = false
		for _, form in ipairs(forms) do
			local stem
			if ending_to_chop then
				stem = rmatch(form.form, "^(.*)" .. ending_to_chop .. "$")
			else
				stem = form.form
			end
			if stem and rfind("#" .. stem, expected) then
				seen = true
				form.seen = form.seen or "maybe"
			end
		end
		return seen
	end

	local function reset_maybes(forms, value)
		for _, form in ipairs(forms) do
			if form.seen == "maybe" then
				form.seen = value
			end
		end
	end

	local function reset_all_maybes(value)
		reset_maybes(infstem, value)
		reset_maybes(past, value)
		reset_maybes(pp, value)
	end

	local function has_unseen_weak_pp()
		for _, form in ipairs(pp) do
			if not form.seen and form.form:find("[dt]$") then
				return true
			end
		end
		return false
	end

	local function has_unseen_strong_pp()
		for _, form in ipairs(pp) do
			if not form.seen and form.form:find("n$") then
				return true
			end
		end
		return false
	end

	local function check(verbtype, infre, pastre, ppre, exclude)
		if exclude then
			for _, form in ipairs(infstem) do
				if exclude(form.form) then
					return
				end
			end
		end
		if matches_forms(infstem, infre) and
			matches_forms(past, pastre) and
			matches_forms(pp, ppre, "en") then
			m_table.insertIfNot(this_verb_types, verbtype)
			reset_all_maybes(true)
		else
			reset_all_maybes(false)
		end
	end

	local function check_strong()
		check("1", "Ce[iy]C*", "CieC*", "Cie?C*") -- beigen, bleiben, gedeihen, leihen, meiden, preisen, reiben, reihen,
			-- scheiden, scheinen, schreiben, schreien, schweigen, speiben, speien, speisen, steigen, treiben, weisen,
			-- zeihen; use 'Cie?C*' for past participle to handle 'schrien', 'spien'
		check("1", "Ce[iy]C*", "CiC*", "CiC*") -- beißen/beissen/beyßen/beyssen, bleichen, fleißen/fleissen, gleichen,
			-- gleiten, greifen, kneifen, kreischen, leiden, pfeifen, reißen/reissen, reiten, scheißen/scheissen,
			-- schleichen, schleifen, schleißen/schleissen, schmeißen/schmeissen, schneiden/schneyden, schreiten,
			-- spleißen/spleissen, streichen, streiten, weichen
		check("2", "CieC*", "CoC*", "CoC*") -- biegen, bieten, fliegen, fliehen, fließen/fliessen, frieren,
			-- genießen/geniessen, gießen/giessen, kiesen, kriechen, riechen, schieben, schießen/schiessen, schliefen,
			-- schließen/schliessen, sieden, sprießen/spriessen, stieben, triefen, verdrießen/verdriessen, verlieren,
			-- wiegen, ziehen
		check("2", "CauC*", "CoC*", "CoC*") -- krauchen, saufen, saugen
		check("2", "CüC", "CoC", "CoC") -- lügen, trügen
		local function exclude_nehmen_sprechen(form)
			-- need to exclude nehmen, stehlen, befehlen/empfehlen, sprechen, brechen, stechen
			return rfind(form, vowel_c .. "ch$") or rfind(form, vowel_c .. "h" .. not_vowel_c .. "$")
		end
		check("3", "C[ei]CC+", "CaCC+", "C[ou]CC+", exclude_nehmen_sprechen) -- [with e, + o in pp]: bergen, bersten,
			-- gelten, helfen, schelten, sterben, verderben, werben, werfen; [with i, + u in pp]: binden, brinnen, dringen,
			-- finden, gelingen, klingen, misslingen, ringen, schlingen, schwinden, schwingen, singen, sinken, springen,
			-- stinken, trinken, winden, wringen, zwingen; [with i, + o in pp]: rinnen, gewinnen, schwimmen, sinnen,
			-- spinnen
		check("3", "C[eiaö]CC+", "CoCC+", "CoCC+", exclude_nehmen_sprechen)
			-- [with e]: dreschen, fechten, flechten, melken, quellen, schmelzen, schwellen; [with i]: glimmen, klimmen;
			-- [with a]: schallen (geschallt), erschallen; [with ö]: erlöschen
		check("3", "quell", "quoll", "quoll") -- need to special-case quellen due to u preceding e
		check("3", "schind", "schund", "schund") -- need to special-case due to 'u' in past
		check("4", "C[eäo]C*", "Cah?Ch?", "CoC*") -- [with e]: befehlen, brechen, schrecken, nehmen, sprechen, stechen,
			-- stecken (gesteckt), stehlen, treffen; [with ä]: gebären; [with o]: kommen
		check("4", "C[äe]C", "Coh?C", "Coh?C", function(form) return form:find("heb$") end)
			-- [with ä]: gären, wägen, schwären; [with e]: bewegen, weben, scheren (but not heben)
		check("5", "C[ei]C*", "CaC", "CeC*") -- [with e, one C]: geben, genesen, geschehen, lesen, meßen, sehen, treten;
			-- [with e, two C]: essen, fressen, messen, vergessen; [with i, two C]: bitten, sitzen
		check("5", "C[ei]C*", "Cass", "Cess") -- essen, fressen, messen, sitzen in Swiss spelling
		check("5", "CieC", "CaC", "CeC") -- liegen
		check("6", "CaC*", "CuC*", "CaC*") -- backen, fragen (gefragt), graben, laden, mahlen, schaffen, schlagen,
			-- tragen, wachsen, waschen
		check("6", "heb", "h[ou]b", "hob") -- we need to special-case this because heben (class 6 per Wikipedia) has the
			-- exact same vowels as weben (class 4 per Wikipedia)
		check("6", "schwör", "schw[ou]r", "schwor") -- only strong verb with these vowels
		check("7", "CaC*", "CieC*", "CaC*") -- blasen, braten, fallen, halten, lassen, raten/rathen, schlafen
		check("7", "C[aäe]C*", "CiC*", "CaC*") -- [with a]: fangen; [with ä]: hängen; [with e]: gehen
		check("7", "Ce[iy]C*", "CieC*", "Ce[iy]C*") -- heißen/heissen/heyßen/heyssen
		check("7", "CauC*", "CieC*", "CauC*") -- hauen, laufen
		check("7", "CoC*", "CieC*", "CoC*") -- stoßen/stossen
		check("7", "CuC*", "CieC*", "CuC*") -- rufen
	end

	for _, form in ipairs(past) do
		local past_stem = form.form:match("^(.*)te$")
		if past_stem then
			if matches_forms(infstem, "#" .. past_stem) then
				-- Need to run matches_forms() on all possibilities even if earlier ones match,
				-- to mark the seen forms correctly.
				local matches_pp = matches_forms(pp, "#" .. past_stem .. "t")
				matches_pp = matches_forms(pp, "#ge" .. past_stem .. "t") or matches_pp
				if matches_pp then
					m_table.insertIfNot(this_verb_types, "weak")
					form.seen = true
					reset_all_maybes(true)
				else
					reset_all_maybes(false)
				end
			end
		end
		if not form.seen and form.form:find("ete$") then
			if matches_forms(infstem, "#" .. past_stem:gsub("e$", "")) then
				-- Need to run matches_forms() on all possibilities even if earlier ones match,
				-- to mark the seen forms correctly.
				local matches_pp = matches_forms(pp, "#" .. past_stem .. "t")
				matches_pp = matches_forms(pp, "#ge" .. past_stem .. "t") or matches_pp
				matches_pp = matches_forms(pp, "#" .. past_stem .. "d") or matches_pp
				matches_pp = matches_forms(pp, "#ge" .. past_stem .. "d") or matches_pp
				if matches_pp then
					m_table.insertIfNot(this_verb_types, "weak")
					form.seen = true
					reset_all_maybes(true)
				else
					reset_all_maybes(false)
				end
			end
		end
		if past_stem and not form.seen then
			if not has_unseen_weak_pp() and has_unseen_strong_pp() then
				m_table.insertIfNot(this_verb_types, "mixed")
			else
				m_table.insertIfNot(this_verb_types, "irregweak")
			end
			matches_forms(pp, "#" .. past_stem .. "t")
			matches_forms(pp, "#ge" .. past_stem .. "t")
			form.seen = true
			reset_all_maybes(true)
		end
		if not form.seen then
			check_strong()
		end
		if not form.seen then
			if not has_unseen_strong_pp() and has_unseen_weak_pp() then
				m_table.insertIfNot(this_verb_types, "mixed")
			else
				m_table.insertIfNot(this_verb_types, "irregstrong")
			end
		end
	end

	for _, form in ipairs(pp) do
		if not form.seen then
			if form.form:find("n$") then
				if m_table.contains(this_verb_types, "strong") then
					m_table.insertIfNot(this_verb_types, "irregstrong")
				elseif m_table.contains(this_verb_types, "weak") then
					m_table.insertIfNot(this_verb_types, "mixed")
				end
			elseif form.form:find("[dt]$") then
				if m_table.contains(this_verb_types, "weak") then
					m_table.insertIfNot(this_verb_types, "irregweak")
				elseif m_table.contains(this_verb_types, "strong") then
					m_table.insertIfNot(this_verb_types, "mixed")
				end
			end
		end
	end

	base.verb_types = this_verb_types

	if verb_types then
		for _, verb_type in ipairs(this_verb_types) do
			m_table.insertIfNot(verb_types, verb_type)
		end
	end

	set_verb_type()
end


local function detect_indicator_spec(base)
	base.forms = {}
	base.aux = base.aux or {{form = "haben"}}
	base.bare_infinitive = {{form = base.base_verb, footnotes = base.infstem_footnotes}}
	add(base, "infinitive", base.pre_pref, base.bare_infinitive)

	if base.only3s and base.only3sp then
		error("'only3s' and 'only3sp' cannot both be specified")
	end

	if base.conj == "irreg" then
		for irregverb, verbobj in pairs(irreg_verbs) do
			base.insep_prefix = base.base_verb:match("^(.-)" .. irregverb .. "$")
			if base.insep_prefix then
				base.irregverb = irregverb
				base.irregverbobj = verbobj
				if not base.ge_prefix then
					if base.insep_prefix ~= "" then
						base.ge_prefix = {{form = ""}}
					else
						base.ge_prefix = {{form = "ge"}}
					end
				end
				return
			end
		end
		error("Unrecognized irregular base verb '" .. base.base_verb .. "'")
	end

	-- The following applies to everything but 'irreg' verbs.

	local infstem, infroot = base.base_verb:match("^((.*)e[lr])n$")
	if infstem then
		base.unstressed_el_er = true
	else
		infstem, infroot = base.base_verb:match("^((.*)erl)n$") -- [[fensterln]]
		if infstem then
			base.unstressed_erl = true
		else
			infstem = base.base_verb:match("^(.*)en$")
			infroot = infstem
			if not infstem then
				error("Unrecognized infinitive, should end in -en, -eln, -ern or -erln: '" .. base.base_verb .. "'")
			end
		end
	end
	base.infstem = {{form = infstem, footnotes = base.infstem_footnotes}}

	if base.unstressed_e_infix == nil then
		-- Autodetect whether we need an -e- infix in the pres_2s and pres_3s ([[atmen]], [[eignen]], etc.).
		-- Almost all such cases have -Cmen or -Cnen where C is a consonant other than r or l and other than the
		-- following m or n (hence [[meinen]], [[lernen]], [[filmen]], [[schwimmen]] should be excluded); we also
		-- need to exclue -Vhmen and -Vhnen ([[wohnen]], [[rühmen]]), but not -Chmen and -Chnen ([[zeichnen]]).
		if base.base_verb:find("[mn]en$") and not base.base_verb:find("([mn])%1en$") and
			not rfind(base.base_verb, vowel_c .. "[hrl]?[mn]en$") then
			base.unstressed_e_infix = true
		end
	end

	if not base.conj then
		base.conj = "normal"
	end
	if base.conj == "normal" then
		local weak_past
		if not base.past then
			if base.unstressed_e_infix or ends_in_dt(infstem) then
				weak_past = infstem .. "et"
			else
				weak_past = infstem .. "t"
			end
			base.past = {{form = weak_past .. "e"}}
		end
		if not base.pp then
			if not weak_past then
				error("Internal error: past was explicitly given but not past participle")
			end
			if not base.ge_prefix then
				local no_ge
				for _, insep_prefix in ipairs(inseparable_prefixes) do
					-- There must be a vowel following the inseparable prefix; excludes beben, bechern, belfern, bellen, bessern,
					-- beten, betteln, betten, erben, erden, ernten, erzen, entern, gecken, gehren, gellen, gerben, geten, missen,
					-- zergen, zerren, etc.
					if rfind(infroot, "^" .. insep_prefix .. ".*" .. vowel_c .. ".*") and
						-- Exclude cases like beigen, beichten, beugen, beulen, geifern; this also wrongly excludes
						-- beirren, which needs -ge.
						not rfind(infroot, "^[bg]e[iu]" .. not_vowel_c .. "*$") then
						no_ge = true
						break
					end
					-- Check for -ier preceded by a vowel (excludes bieren, frieren, gieren, schmieren, stieren, zieren, etc.)
					if not base.unstressed_el_er and not base.unstressed_erl and rfind(infroot, "^.*" .. vowel_c .. ".*ier$") then
						no_ge = true
						break
					end
				end
				if no_ge then
					base.ge_prefix = {{form = ""}}
				else
					base.ge_prefix = {{form = "ge"}}
				end
			end
			base.pp = iut.map_forms(base.ge_prefix, function(form)
				if base.unstressed_el_er or base.unstressed_erl then
					return form .. base.base_verb:gsub("n$", "") .. "t"
				else
					return form .. weak_past
				end
			end)
		end
	else
		if not base.pp then
			error("For '" .. base.conj .. "' type verbs, past participle must be explicitly given")
		end
	end

	add(base, "perf_part", base.pre_pref, base.pp)
end


local function detect_all_indicator_specs(alternant_multiword_spec)
	iut.map_word_specs(alternant_multiword_spec, function(base)
		detect_indicator_spec(base)
		detect_verb_type(base)
	end)
end


-- Set the overall auxiliary or auxiliaries. We can't do this using the normal inflection
-- code as it will produce e.g. '[[haben]] und [[haben]]' for conjoined verbs.
local function compute_auxiliary(alternant_multiword_spec)
	iut.map_word_specs(alternant_multiword_spec, function(base)
		iut.insert_forms(alternant_multiword_spec.forms, "aux", base.aux)
	end)
end


function export.process_verb_classes(classes)
	local class_descs = {}
	local cats = {}

	local function insert_desc(desc)
		m_table.insertIfNot(class_descs, desc)
	end

	local function insert_cat(cat)
		m_table.insertIfNot(cats, cat .. "ภาษาเยอรมัน")
	end

	for _, class in ipairs(classes) do
		if class == "weak" then
			insert_desc("[[ภาคผนวก:อภิธานศัพท์#คำกริยาอ่อน|คำกริยาอ่อน]]")
			insert_cat("คำกริยาอ่อน")
		elseif class == "irregweak" then
			insert_desc("[[ภาคผนวก:อภิธานศัพท์#คำกริยาอ่อน|คำกริยาอ่อน]][[ภาคผนวก:อภิธานศัพท์#ไม่ปรกติ|ไม่ปรกติ]]")
			insert_cat("คำกริยาอ่อน")
			insert_cat("คำกริยาอ่อนไม่ปรกติ")
		elseif class == "pretpres" then
			insert_desc("[[Appendix:Glossary#preterite-present verb|preterite-present]]")
			insert_cat("preterite-present verbs")
		elseif class == "irreg" then
			insert_desc("[[ภาคผนวก:อภิธานศัพท์#ไม่ปรกติ|ไม่ปรกติ]]")
			insert_cat("คำกริยาผิดปกติ")
		elseif class == "mixed" then
			insert_desc("คำกริยาผสม")
			insert_cat("คำกริยาผสม")
		elseif class == "irregstrong" then
			insert_desc("[[ภาคผนวก:อภิธานศัพท์#คำกริยาแข็ง|คำกริยาแข็ง]][[ภาคผนวก:อภิธานศัพท์#ไม่ปรกติ|ไม่ปรกติ]]")
			insert_cat("คำกริยาแข็ง")
			insert_cat("คำกริยาแข็งไม่ปรกติ")
		elseif class:find("^[1-7]$") then
			insert_desc("[[ภาคผนวก:อภิธานศัพท์#คำกริยาแข็ง|คำกริยาแข็ง]]กลุ่ม " .. class)
			insert_cat("คำกริยาแข็ง")
			insert_cat("คำกริยาแข็งกลุ่ม " .. class .. "&nbsp;")
		else
			error("คำกริยาไม่ทราบกลุ่ม '" .. class .. "'")
		end
	end

	return class_descs, cats
end


local function add_categories_and_annotation(alternant_multiword_spec, base, from_headword, manual)
	local function insert_cat(full_cat)
		m_table.insertIfNot(alternant_multiword_spec.categories, full_cat)
	end

	if not from_headword then
		for _, slot_and_accel in ipairs(all_verb_slots) do
			local slot = slot_and_accel[1]
			local forms = base.forms[slot]
			local must_break = false
			if forms then
				for _, form in ipairs(forms) do
					if not form.form:find("%[%[") then
						local title = mw.title.new(form.form)
						if title and not title.exists then
							insert_cat("คำกริยาที่มีลิงก์แดงในตารางการผันรูปภาษาเยอรมัน")
							must_break = true
							break
						end
					end
				end
			end
			if must_break then
				break
			end
		end
	end

	if manual then
		return
	end

	local class_descs, cats = export.process_verb_classes(base.verb_types)
	for _, desc in ipairs(class_descs) do
		m_table.insertIfNot(alternant_multiword_spec.verb_types, desc)
	end
	-- Don't place multiword terms in categories like 'German class 4 strong verbs' to avoid spamming the
	-- categories with such terms.
	if from_headword and not base.lemma:find(" ") then
		for _, cat in ipairs(cats) do
			insert_cat(cat)
		end
	end

	for _, aux in ipairs(base.aux) do
		m_table.insertIfNot(alternant_multiword_spec.auxiliaries, link_term(aux.form, "term"))
		if from_headword and not base.lemma:find(" ") then -- see above
			insert_cat("คำกริยาที่ใช้ " .. aux.form .. " เป็นคำกริยานุเคราะห์ภาษาเยอรมัน")
			-- Set flags for use below in adding 'German verbs using haben and sein as auxiliary'
			alternant_multiword_spec["saw_" .. aux.form] = true
		end
	end
end


-- Compute the categories to add the verb to, as well as the annotation to display in the
-- conjugation title bar. We combine the code to do these functions as both categories and
-- title bar contain similar information.
local function compute_categories_and_annotation(alternant_multiword_spec, from_headword, manual)
	alternant_multiword_spec.categories = {}
	alternant_multiword_spec.verb_types = {}
	alternant_multiword_spec.auxiliaries = {}
	iut.map_word_specs(alternant_multiword_spec, function(base)
		add_categories_and_annotation(alternant_multiword_spec, base, from_headword)
	end)
	if manual then
		alternant_multiword_spec.annotation = ""
		return
	end
	local ann_parts = {}
	table.insert(ann_parts, table.concat(alternant_multiword_spec.verb_types, " หรือ "))
	if #alternant_multiword_spec.auxiliaries > 0 then
		table.insert(ann_parts, ", คำกริยานุเคราะห์ " .. table.concat(alternant_multiword_spec.auxiliaries, " หรือ "))
	end
	if from_headword and alternant_multiword_spec.saw_haben and alternant_multiword_spec.saw_sein then
		m_table.insertIfNot(alternant_multiword_spec.categories, "คำกริยาที่ใช้ haben และ sein เป็นคำกริยานุเคราะห์ภาษาเยอรมัน")
	end
	alternant_multiword_spec.annotation = table.concat(ann_parts)
end


local function show_forms(alternant_multiword_spec)
	local lemmas = iut.map_forms(alternant_multiword_spec.forms.infinitive,
		remove_sein_sich_indicators)
	alternant_multiword_spec.lemmas = lemmas -- save for later use in make_table()
	local linked_pronouns = {}
	for index, pronoun in ipairs(pronouns) do
		-- use 'es' instead of 'er' for 3s-only verbs
		if index == 3 and alternant_multiword_spec.only3s then
			linked_pronouns[index] = link_term("es")
		else
			linked_pronouns[index] = link_term(pronoun)
		end
	end
	dass = link_term("dass") .. " "

	local function generate_link(data)
		local link = m_links.full_link {
			lang = lang, term = data.form.formval_for_link, tr = "-", accel = data.form.accel_obj
		}
		local footnote_text = iut.get_footnote_text(data.form.footnotes, data.footnote_obj)
		local persnum = data.slot:match("^imp_(2[sp])$")
		if persnum then
			link = link .. " (" .. linked_pronouns[persnum_to_index[persnum]] .. ")"
		else
			persnum = data.slot:match("^.*_([123][sp])$")
			if persnum then
				link = linked_pronouns[persnum_to_index[persnum]] .. " " .. link
			end
			if data.slot:find("^subc_") then
				link = dass .. link
			end
		end
		return link .. footnote_text
	end

	local function join_spans(data)
		if data.slot == "aux" then
			return table.concat(data.formval_spans, " หรือ ")
		else
			return table.concat(data.formval_spans, "<br />")
		end
	end

	local props = {
		lang = lang,
		lemmas = lemmas,
		generate_link = generate_link,
		join_spans = join_spans,
	}
	props.slot_list = verb_slots_basic
	iut.show_forms(alternant_multiword_spec.forms, props)
	alternant_multiword_spec.footnote_basic = alternant_multiword_spec.forms.footnote
	props.slot_list = verb_slots_subordinate_clause
	iut.show_forms(alternant_multiword_spec.forms, props)
	alternant_multiword_spec.footnote_subordinate_clause = alternant_multiword_spec.forms.footnote
	props.slot_list = verb_slots_composed
	iut.show_forms(alternant_multiword_spec.forms, props)
	alternant_multiword_spec.footnote_composed = alternant_multiword_spec.forms.footnote
end


local notes_template = [=[
<div style="width:100%;text-align:left;background:#d9ebff">
<div style="display:inline-block;text-align:left;padding-left:1em;padding-right:1em">
{footnote}
</div></div>
]=]

local zu_infinitive_table = [=[
|-
! colspan="2" style="background:#d0d0d0" | รูปกริยากลาง zu
| colspan="4" | {zu_infinitive}
]=]

local basic_table = [=[
<div class="NavFrame" style="">
<div class="NavHead" style="">รูปผันคำกริยา {title}</div>
<div class="NavContent">
{\op}| border="1px solid #000000" style="border-collapse:collapse; background:#fafafa; text-align:center; width:100%" class="inflection-table"
|-
! colspan="2" style="background:#d0d0d0" | <span title="Infinitiv">รูปกริยากลาง</span>
| colspan="4" | {infinitive}
|-
! colspan="2" style="background:#d0d0d0" | <span title="Partizip I (Partizip Präsens)">พาร์ทิซิเพิลปัจจุบันกาล</span>
| colspan="4" | {pres_part}
|-
! colspan="2" style="background:#d0d0d0" | <span title="Partizip II (Partizip Perfekt)">พาร์ทิซิเพิลอดีตกาล</span>
| colspan="4" | {perf_part}
{zu_infinitive_table}|-
! colspan="2" style="background:#d0d0d0" | <span title="Hilfsverb">คำกริยานุเคราะห์</span>
| colspan="4" | {aux}
|-
| rowspan="2" style="background:#a0ade3" |
! colspan="2" style="background:#a0ade3" | <span title="Indikativ">มาลาบอกเล่า</span>
| rowspan="2" style="background:#a0ade3" |
! colspan="2" style="background:#a0ade3" | <span title="Konjunktiv">มาลาสมมุติ</span>
|-
! style="background:#a0ade3" | เอกพจน์
! style="background:#a0ade3" | พหูพจน์
! style="background:#a0ade3" | เอกพจน์
! style="background:#a0ade3" | พหูพจน์
|-
! rowspan="3" style="background:#c0cfe4; width:7em" | <span title="Präsens">ปัจจุบันกาล</span>
| {pres_1s}
| {pres_1p}
! rowspan="3" style="background:#c0cfe4; width:7em" | <span title="Konjunktiv I (Konjunktiv Präsens)">I</span>
| {subi_1s}
| {subi_1p}
|-
| {pres_2s}
| {pres_2p}
| {subi_2s}
| {subi_2p}
|-
| {pres_3s}
| {pres_3p}
| {subi_3s}
| {subi_3p}
|-
| colspan="6" style="background:#d5d5d5; height: .25em" | 
|-
! rowspan="3" style="background:#c0cfe4" | <span title="Präteritum">อดีตกาล</span>
| {pret_1s}
| {pret_1p}
! rowspan="3" style="background:#c0cfe4" | <span title="Konjunktiv II (Konjunktiv Präteritum)">II</span>
| {subii_1s}
| {subii_1p}
|-
| {pret_2s}
| {pret_2p}
| {subii_2s}
| {subii_2p}
|-
| {pret_3s}
| {pret_3p}
| {subii_3s}
| {subii_3p}
|-
| colspan="6" style="background:#d5d5d5; height: .25em" | 
|-
! style="background:#c0cfe4" | <span title="Imperativ">มาลาสั่ง</span>
| {imp_2s}
| {imp_2p}
| colspan="3" style="background:#e0e0e0" |
|{\cl}{notes_clause}</div></div>
]=]

local subordinate_clause_table = [=[
<div class="NavFrame" style="">
<div class="NavHead" style="">รูปผันคำกริยา {title} เมื่ออยู่ในอนุประโยค</div>
<div class="NavContent">
{\op}| border="1px solid #000000" style="border-collapse:collapse; background:#fafafa; text-align:center; width:100%" class="inflection-table"
|-
| style="background:#a0ade3" |
! colspan="2" style="background:#a0ade3" | <span title="Indikativ">มาลาบอกเล่า</span>
| style="background:#a0ade3" |
! colspan="2" style="background:#a0ade3" | <span title="Konjunktiv">มาลาสมมุติ</span>
|-
! rowspan="3" style="background:#c0cfe4; width:7em" | <span title="Präsens">ปัจจุบัน</span>
| {subc_pres_1s}
| {subc_pres_1p}
! rowspan="3" style="background:#c0cfe4; width:7em" | <span title="Konjunktiv I (Konjunktiv Präsens)">I</span> 
| {subc_subi_1s}
| {subc_subi_1p}
|-
| {subc_pres_2s}
| {subc_pres_2p}
| {subc_subi_2s}
| {subc_subi_2p}
|-
| {subc_pres_3s}
| {subc_pres_3p}
| {subc_subi_3s}
| {subc_subi_3p}
|-
| colspan="6" style="background:#d5d5d5; height: .25em" | 
|-
! rowspan="3" style="background:#c0cfe4" | <span title="Präteritum">อดีต</span>
| {subc_pret_1s}
| {subc_pret_1p}
! rowspan="3" style="background:#c0cfe4" | <span title="Konjunktiv II (Konjunktiv Präteritum)">II</span>
| {subc_subii_1s}
| {subc_subii_1p}
|-
| {subc_pret_2s}
| {subc_pret_2p}
| {subc_subii_2s}
| {subc_subii_2p}
|-
| {subc_pret_3s}
| {subc_pret_3p}
| {subc_subii_3s}
| {subc_subii_3p}
|{\cl}{notes_clause}</div></div>
]=]

local composed_table = [=[
<div class="NavFrame" style="">
<div class="NavHead" style="">รูปประกอบของคำกริยา {title}</div>
<div class="NavContent">
{\op}| border="1px solid #000000" style="border-collapse:collapse; background:#fafafa; text-align:center; width:100%" class="inflection-table"
|-
! colspan="6" style="background:#99cc99" | <span title="Perfekt">การณ์ลักษณะสมบูรณ์</span>
|-
! style="background:#99cc99" |
! style="background:#99cc99" | เอกพจน์
! style="background:#99cc99" | พหูพจน์
! style="background:#99cc99" |
! style="background:#99cc99" | เอกพจน์
! style="background:#99cc99" | พหูพจน์
|-
! rowspan="3" style="background:#cfedcc; width:7em" | <span title="Indikativ">มาลาบอกเล่า</span>
| {perf_ind_1s}
| {perf_ind_1p}
! rowspan="3" style="background:#cfedcc; width:7em" | <span title="Konjunktiv">มาลาสมมุติ</span>
| {perf_sub_1s}
| {perf_sub_1p}
|-
| {perf_ind_2s}
| {perf_ind_2p}
| {perf_sub_2s}
| {perf_sub_2p}
|-
| {perf_ind_3s}
| {perf_ind_3p}
| {perf_sub_3s}
| {perf_sub_3p}
|-
! colspan="6" style="background:#99CC99" | <span title="Plusquamperfekt">การณ์ลักษณะอภิสมบูรณ์</span>
|-
! rowspan="3" style="background:#cfedcc" | <span title="Indikativ">มาลาบอกเล่า</span>
| {plup_ind_1s}
| {plup_ind_1p}
! rowspan="3" style="background:#cfedcc" | <span title="Konjunktiv">มาลาสมมุติ</span>
| {plup_sub_1s}
| {plup_sub_1p}
|-
| {plup_ind_2s}
| {plup_ind_2p}
| {plup_sub_2s}
| {plup_sub_2p}
|-
| {plup_ind_3s}
| {plup_ind_3p}
| {plup_sub_3s}
| {plup_sub_3p}
|-
! colspan="6" style="background:#9999DF" | <span title="Futur I">อนาคตกาล I</span>
|-
! rowspan="3" style="background:#ccccff" | <span title="Infinitiv">รูปกริยากลาง</span>
| rowspan="3" colspan="2" | {futi_inf}
! rowspan="3" style="background:#ccccff" | <span title="Konjunktiv I (Konjunktiv Präsens)">มาลาสมมุติ I</span>
| {futi_subi_1s}
| {futi_subi_1p}
|-
| {futi_subi_2s}
| {futi_subi_2p}
|-
| {futi_subi_3s}
| {futi_subi_3p}
|-
! colspan="6" style="background:#d5d5d5; height: .25em" |
|-
! rowspan="3" style="background:#ccccff" | <span title="Indikativ">มาลาบอกเล่า</span>
| {futi_ind_1s}
| {futi_ind_1p}
! rowspan="3" style="background:#ccccff" | <span title="Konjunktiv II (Konjunktiv Präteritum)">มาลาสมมุติ II</span>
| {futi_subii_1s}
| {futi_subii_1p}
|-
| {futi_ind_2s}
| {futi_ind_2p}
| {futi_subii_2s}
| {futi_subii_2p}
|-
| {futi_ind_3s}
| {futi_ind_3p}
| {futi_subii_3s}
| {futi_subii_3p}
|-
! colspan="6" style="background:#9999DF" | <span title="Futur II">อนาคตกาล II</span>
|-
! rowspan="3" style="background:#ccccff" | <span title="Infinitiv">รูปกริยากลาง</span>
| rowspan="3" colspan="2" | {futii_inf}
! rowspan="3" style="background:#ccccff" | <span title="Konjunktiv I (Konjunktiv Präsens)">มาลาสมมุติ I</span>
| {futii_subi_1s}
| {futii_subi_1p}
|-
| {futii_subi_2s}
| {futii_subi_2p}
|-
| {futii_subi_3s}
| {futii_subi_3p}
|-
! colspan="6" style="background:#d5d5d5; height: .25em" |
|-
! rowspan="3" style="background:#ccccff" | <span title="Indikativ">มาลาบอกเล่า</span>
| {futii_ind_1s}
| {futii_ind_1p}
! rowspan="3" style="background:#ccccff" | <span title="Konjunktiv II (Konjunktiv Präteritum)">มาลาสมมุติ II</span>
| {futii_subii_1s}
| {futii_subii_1p}
|-
| {futii_ind_2s}
| {futii_ind_2p}
| {futii_subii_2s}
| {futii_subii_2p}
|-
| {futii_ind_3s}
| {futii_ind_3p}
| {futii_subii_3s}
| {futii_subii_3p}
|{\cl}{notes_clause}</div></div>]=]


local function make_table(alternant_multiword_spec)
	local forms = alternant_multiword_spec.forms

	forms.title = link_term(alternant_multiword_spec.lemmas[1].form, "term")
	if alternant_multiword_spec.annotation ~= "" then
		forms.title = forms.title .. " (" .. alternant_multiword_spec.annotation .. ")"
	end

	-- Maybe format the subordinate clause table.
	local formatted_subordinate_clause_table
	if forms.subc_pres_3s ~= "—" then -- use 3s in case of only3s verb
		forms.zu_infinitive_table = m_string_utilities.format(zu_infinitive_table, forms)
		forms.footnote = alternant_multiword_spec.footnote_subordinate_clause
		forms.notes_clause = forms.footnote ~= "" and m_string_utilities.format(notes_template, forms) or ""
		formatted_subordinate_clause_table = m_string_utilities.format(subordinate_clause_table, forms)
	else
		forms.zu_infinitive_table = ""
		formatted_subordinate_clause_table = ""
	end

	-- Format the basic table.
	forms.footnote = alternant_multiword_spec.footnote_basic
	forms.notes_clause = forms.footnote ~= "" and m_string_utilities.format(notes_template, forms) or ""
	local formatted_basic_table = m_string_utilities.format(basic_table, forms)

	-- Format the composed table.
	forms.footnote = alternant_multiword_spec.footnote_composed
	forms.notes_clause = forms.footnote ~= "" and m_string_utilities.format(notes_template, forms) or ""
	local formatted_composed_table = m_string_utilities.format(composed_table, forms)

	-- Paste them together.
	return formatted_basic_table .. formatted_subordinate_clause_table .. formatted_composed_table
end


-- Externally callable function to parse and conjugate a verb given user-specified arguments.
-- Return value is WORD_SPEC, an object where the conjugated forms are in `WORD_SPEC.forms`
-- for each slot. If there are no values for a slot, the slot key will be missing. The value
-- for a given slot is a list of objects {form=FORM, footnotes=FOOTNOTES}.
function export.do_generate_forms(parent_args, from_headword, def)
	local params = {
		[1] = {},
	}

	if from_headword then
		params["lemma"] = {list = true}
		params["id"] = {}
	end

	local args = require("Module:parameters").process(parent_args, params)
	local PAGENAME = mw.title.getCurrentTitle().text

	if not args[1] then
		if PAGENAME == "de-conj" or PAGENAME == "de-verb" then
			args[1] = def or "aus.fahren<fährt#fuhr,gefahren,führe.haben,sein>"
		else
			args[1] = PAGENAME
			-- If pagename has spaces in it, add links around each word
			if args[1]:find(" ") then
				args[1] = "[[" .. args[1]:gsub(" ", "]] [[") .. "]]"
			end
		end
	end
	local parse_props = {
		parse_indicator_spec = parse_indicator_spec,
		lang = lang,
		allow_default_indicator = true,
		allow_blank_lemma = true,
	}
	local escaped_arg1 = escape_sein_sich_indicators(args[1])
	local alternant_multiword_spec = iut.parse_inflected_text(escaped_arg1, parse_props)
	alternant_multiword_spec.pos = pos or "verbs"
	alternant_multiword_spec.args = args
	normalize_all_lemmas(alternant_multiword_spec, from_headword)
	detect_all_indicator_specs(alternant_multiword_spec)
	local inflect_props = {
		slot_list = all_verb_slots,
		lang = lang,
		inflect_word_spec = conjugate_verb,
		-- We add links around the generated verbal forms rather than allow the entire multiword
		-- expression to be a link, so ensure that user-specified links get included as well.
		include_user_specified_links = true,
	}
	iut.inflect_multiword_or_alternant_multiword_spec(alternant_multiword_spec, inflect_props)
	compute_auxiliary(alternant_multiword_spec)
	compute_categories_and_annotation(alternant_multiword_spec, from_headword)
	return alternant_multiword_spec
end


-- Entry point for {{de-conj}}. Template-callable function to parse and conjugate a verb given
-- user-specified arguments and generate a displayable table of the conjugated forms.
function export.show(frame)
	local parent_args = frame:getParent().args
	local alternant_multiword_spec = export.do_generate_forms(parent_args)
	show_forms(alternant_multiword_spec)
	return make_table(alternant_multiword_spec) .. require("Module:utilities").format_categories(alternant_multiword_spec.categories, lang)
end


-- Concatenate all forms of all slots into a single string of the form
-- "SLOT=FORM,FORM,...|SLOT=FORM,FORM,...|...". Embedded pipe symbols (as might occur
-- in embedded links) are converted to <!>. If INCLUDE_PROPS is given, also include
-- additional properties (currently, none). This is for use by bots.
local function concat_forms(alternant_multiword_spec, include_props)
	local ins_text = {}
	for _, slot_and_accel in ipairs(all_verb_slots) do
		local slot = slot_and_accel[1]
		local formtext = iut.concat_forms_in_slot(alternant_multiword_spec.forms[slot])
		if formtext then
			table.insert(ins_text, slot .. "=" .. formtext)
		end
	end
	if include_props then
		local verb_types = {}
		iut.map_word_specs(alternant_multiword_spec, function(base)
			detect_verb_type(base, verb_types)
		end)
		table.insert(ins_text, "class=" .. table.concat(verb_types, ","))
	end
	return table.concat(ins_text, "|")
end


local numbered_params = {
	-- required params
	[1] = "infinitive",
	[2] = "pres_part",
	[3] = "perf_part",
	[4] = "aux",
	[5] = "pres_1s",
	[6] = "pres_2s",
	[7] = "pres_3s",
	[8] = "pres_1p",
	[9] = "pres_2p",
	[10] = "pres_3p",
	[11] = "pret_1s",
	[12] = "pret_2s",
	[13] = "pret_3s",
	[14] = "pret_1p",
	[15] = "pret_2p",
	[16] = "pret_3p",
	[17] = "subi_1s",
	[18] = "subi_2s",
	[19] = "subi_3s",
	[20] = "subi_1p",
	[21] = "subi_2p",
	[22] = "subi_3p",
	[23] = "subii_1s",
	[24] = "subii_2s",
	[25] = "subii_3s",
	[26] = "subii_1p",
	[27] = "subii_2p",
	[28] = "subii_3p",
	[29] = "imp_2s",
	[30] = "imp_2p",
	-- [31] formerly the 2nd variant of imp_2s; now no longer allowed (use comma-separated 29=)
	-- [32] formerly indicated whether the 2nd variant of imp_2s was present
	-- optional params
	[33] = "subc_pres_1s",
	[34] = "subc_pres_2s",
	[35] = "subc_pres_3s",
	[36] = "subc_pres_1p",
	[37] = "subc_pres_2p",
	[38] = "subc_pres_3p",
	[39] = "subc_pret_1s",
	[40] = "subc_pret_2s",
	[41] = "subc_pret_3s",
	[42] = "subc_pret_1p",
	[43] = "subc_pret_2p",
	[44] = "subc_pret_3p",
	[45] = "subc_subi_1s",
	[46] = "subc_subi_2s",
	[47] = "subc_subi_3s",
	[48] = "subc_subi_1p",
	[49] = "subc_subi_2p",
	[50] = "subc_subi_3p",
	[51] = "subc_subii_1s",
	[52] = "subc_subii_2s",
	[53] = "subc_subii_3s",
	[54] = "subc_subii_1p",
	[55] = "subc_subii_2p",
	[56] = "subc_subii_3p",
	[57] = "zu_infinitive",
}

local max_required_param = 30



-- Externally callable function to parse and conjugate a verb where all forms are given manually.
-- Return value is WORD_SPEC, an object where the conjugated forms are in `WORD_SPEC.forms`
-- for each slot. If there are no values for a slot, the slot key will be missing. The value
-- for a given slot is a list of objects {form=FORM, footnotes=FOOTNOTES}.
function export.do_generate_forms_manual(parent_args)
	local params = {
		["generate_forms"] = {type = "boolean"},
	}
	for paramnum, _ in pairs(numbered_params) do
		params[paramnum] = {required = paramnum <= max_required_param}
	end

	local args = require("Module:parameters").process(parent_args, params)

	local base = {
		forms = {},
		manual = true,
	}
	local function process_numbered_param(paramnum)
		local argval = args[paramnum]
		if paramnum == 4 then
			if argval == "h" then
				base.aux = {{form = "haben"}}
			elseif argval == "s" then
				base.aux = {{form = "sein"}}
			elseif argval == "hs" then
				base.aux = {{form = "haben"}, {form = "sein"}}
			elseif argval == "sh" then
				base.aux = {{form = "sein"}, {form = "haben"}}
			elseif not argval then
				error("Missing auxiliary in 4=")
			else
				error("Unrecognized auxiliary 4=" .. argval)
			end
		elseif argval and argval ~= "-" then
			local split_vals = rsplit(argval, "%s*,%s*")
			for _, val in ipairs(split_vals) do
				-- FIXME! This won't work with commas or brackets in footnotes.
				-- To fix this, use functions from [[Module:inflection utilities]].
				local form, footnote = val:match("^(.-)%s*(%[[^%]%[]-%])$")
				local footnotes
				if form then
					footnotes = {footnote}
				else
					form = val
				end
				local slot = numbered_params[paramnum]
				--if slot:find("subii") then
				--	local subii_footnotes = get_subii_note(base)
				--	footnotes = iut.combine_footnotes(subii_footnotes, footnotes)
				--end
				iut.insert_form(base.forms, slot, {form = form, footnotes = footnotes})
			end
		end
	end

	-- Do the infinitive first as we need to reference it in subjunctive II footnotes.
	process_numbered_param(1)
	for paramnum, _ in pairs(numbered_params) do
		if paramnum ~= 1 then
			process_numbered_param(paramnum)
		end
	end

	add_composed_forms(base)
	compute_categories_and_annotation(base, nil, "manual")
	return base, args.generate_forms
end


-- Entry point for {{de-conj-table}}. Template-callable function to parse and conjugate a verb given
-- manually-specified inflections and generate a displayable table of the conjugated forms.
function export.show_manual(frame)
	local parent_args = frame:getParent().args
	local base, generate_forms = export.do_generate_forms_manual(parent_args)
	if generate_forms then
		return concat_forms(base)
	end
	show_forms(base)
	return make_table(base) .. require("Module:utilities").format_categories(base.categories, lang)
end


-- Template-callable function to parse and conjugate a verb given user-specified arguments and return
-- the forms as a string "SLOT=FORM,FORM,...|SLOT=FORM,FORM,...|...". Embedded pipe symbols (as might
-- occur in embedded links) are converted to <!>. If |include_props=1 is given, also include
-- additional properties (currently, none). This is for use by bots.
function export.generate_forms(frame)
	local include_props = frame.args["include_props"]
	local parent_args = frame:getParent().args
	local alternant_multiword_spec = export.do_generate_forms(parent_args)
	return concat_forms(alternant_multiword_spec, include_props)
end


return export