Module:User:Benwing2/csb-adjective

local export = {}

--[=[

Authorship: Ben Wing

]=]

--[=[

TERMINOLOGY:

-- "slot" = A particular combination of case/gender/number. Example slot names for adjectives are "gen_f" (genitive feminine singular) and "nom_mp_pers" (personal nominative masculine plural). Each slot is filled with zero or more forms.

-- "form" = The declined Kashubian form representing the value of a given slot.

-- "lemma" = The dictionary form of a given Kashubian term. Generally the nominative masculine singular, but may occasionally be another form if the nominative masculine singular is missing. ]=]

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

local current_title = mw.title.getCurrentTitle local NAMESPACE = current_title.nsText local PAGENAME = current_title.text

local u = mw.ustring.char local rsplit = mw.text.split local rfind = mw.ustring.find local rmatch = mw.ustring.match local rgmatch = mw.ustring.gmatch local rsubn = mw.ustring.gsub local ulen = mw.ustring.len local uupper = mw.ustring.upper

-- version of rsubn that discards all but the first return value local function rsub(term, foo, bar) local retval = rsubn(term, foo, bar) return retval end

-- version of rsubn that returns a 2nd argument boolean indicating whether -- a substitution was made. local function rsubb(term, foo, bar) local retval, nsubs = rsubn(term, foo, bar) return retval, nsubs > 0 end

-- All slots that are used by any of the different tables. The key is the slot and the value is a list of the tables -- that use the slot. "regular" = regular, "plonly" = indicator 'plonly', "dva" = dwa and òba. local input_adjective_slots = { nom_m = {"regular"}, nom_f = {"regular"}, nom_n = {"regular"}, nom_mp_pers = {"regular", "dva", "plonly"}, nom_p_not_mp_pers = {"regular", "plonly"}, nom_mp_npers = {"dva"}, nom_fp = {"dva"}, nom_np = {"dva"}, gen_mn = {"regular"}, gen_f = {"regular"}, gen_p = {"regular", "plonly", "dva"}, dat_mn = {"regular"}, dat_f = {"regular"}, dat_p = {"regular", "plonly", "dva"}, acc_m_an = {"regular"}, acc_m_in = {"regular"}, acc_f = {"regular"}, acc_n = {"regular"}, acc_mp_pers = {"regular", "dva", "plonly"}, acc_p_not_mp_pers = {"regular", "plonly"}, acc_mp_npers = {"dva"}, acc_fp = {"dva"}, acc_np = {"dva"}, ins_mn = {"regular"}, ins_f = {"regular"}, ins_p = {"regular", "plonly", "dva"}, loc_mn = {"regular"}, loc_f = {"regular"}, loc_p = {"regular", "plonly", "dva"}, short = {"regular"}, }

local output_adjective_slots = { nom_m = "nom|m|s", nom_m_linked = "nom|m|s", -- used in Module:csb-noun? nom_f = "nom|f|s", nom_n = "nom|n|s", nom_mp_pers = "pr|nom|m|p", nom_mp_pers_linked = "pr|nom|m|p", -- used in Module:csb-noun? nom_p_not_mp_pers = "np|nom|p", nom_mp_npers = "np|nom|m|p", nom_mp_npers_linked = "pr|nom|m|p", -- used in Module:csb-noun? nom_fp = "nom|f|p", nom_np = "nom|n|p", gen_mn = "gen|m//n|s", gen_f = "gen|f|s", gen_p = "gen|p", dat_mn = "dat|m//n|s", dat_f = "dat|f|s", dat_p = "dat|p", acc_m_an = "an|acc|m|s", acc_m_in = "in|acc|m|s", acc_f = "acc|f|s", acc_n = "acc|n|s", acc_mp_pers = "pr|acc|m|p", acc_p_not_mp_pers = "np|acc|p", acc_mp_npers = "np|acc|m|p", acc_fp = "acc|f|p", acc_np = "acc|n|p", ins_mn = "ins|m//n|s", ins_f = "ins|f|s", ins_p = "ins|p", loc_mn = "loc|m//n|s", loc_f = "loc|f|s", loc_p = "loc|p", short = "short", }

local potential_lemma_slots = {"nom_m", "nom_mp_npers", "nom_mp_pers"}

local function get_output_adjective_slots(alternant_multiword_spec) return output_adjective_slots end

local function combine_stem_ending(stem, ending) if stem == "?" then return "?" else return stem .. ending end end

local function add(base, slot, stems, endings, footnote) if stems then stems = iut.combine_form_and_footnotes(stems, footnote) end iut.add_forms(base.forms, slot, stems, endings, combine_stem_ending) end

local function add_normal_decl(base, stems,	nom_m, nom_f, nom_n, nom_mp_pers, nom_p_not_mp_pers,	gen_mn, gen_f, gen_p,	dat_mn, dat_f, dat_p,	acc_f,	ins_mn, ins_f, ins_p,	loc_mn, loc_f, loc_p,	footnote) if stems then stems = iut.combine_form_and_footnotes(stems, footnote) end add(base, "nom_m", stems, nom_m) add(base, "nom_f", stems, nom_f) add(base, "nom_n", stems, nom_n) add(base, "nom_mp_pers", stems, nom_mp_pers) add(base, "nom_p_not_mp_pers", stems, nom_p_not_mp_pers) add(base, "gen_mn", stems, gen_mn) add(base, "gen_f", stems, gen_f) add(base, "gen_p", stems, gen_p) add(base, "dat_mn", stems, dat_mn) add(base, "dat_f", stems, dat_f) add(base, "dat_p", stems, dat_p) add(base, "acc_f", stems, acc_f) add(base, "ins_mn", stems, ins_mn) add(base, "ins_f", stems, ins_f) add(base, "ins_p", stems, ins_p) add(base, "loc_mn", stems, loc_mn) add(base, "loc_f", stems, loc_f) add(base, "loc_p", stems, loc_p) end

local function add_normal_oblique_hard_y_decl(base, stem) add_normal_decl(base, stem,		nil, "ô", "é", "y", "é",		"égò", "y", "ëch",		"émù", "y", "ym",		"ą",		"ym", "ą", "yma",		"ym", "y", "ëch"	) end

local decls = {}

local function stem_is_soft(stem) return rfind(stem, "[ńż]$") or rfind(stem, "[cs]z$") end

decls["normal"] = function(base) local stem, suffix

-- hard in -y stem, suffix = rmatch(base.lemma, "^(.*)(y)$") if stem then add_normal_decl(base, stem, "y") add_normal_oblique_hard_y_decl(base, stem) return end

-- hard or soft in -i stem, suffix = rmatch(base.lemma, "^(.*)(i)$") if stem then local soft = stem_is_soft(stem) local cztok = rfind(stem, "cz$") and base.cztok local hard_stem = cztok and rsub(stem, "cz$", "k") or stem -- Because there may be two stems, do them separately. add_normal_decl(base, stem,			"i", nil, "é", "i", "é",			"égò", "i", soft and "ich" or "ëch",			"émù", "i", "im",			nil,			"im", nil, "ima",			"im", "i", soft and "ich" or "ëch"		) add_normal_decl(base, hard_stem,			nil, "ô", nil, nil, nil,			nil, nil, nil,			nil, nil, nil,			"ą",			nil, "ą", nil,			nil, nil, nil		) return end

-- possessive in -ův stem, suffix = rmatch(base.lemma, "^(.*)(ów)$") if stem then add_normal_decl(base, stem,			"ów", "owa", {"owò", "owé"}, "owi", "owé",			"owégò", "owi", "owëch",			"owémù", "owi", "owim",			"ową",			"owim", "ową", "owima",			"owim", "owi", "owëch"		) return end

-- possessive in -in stem, suffix = rmatch(base.lemma, "^(.*)(in)$") if stem then add_normal_decl(base, stem, "in", nil, "ino") add_normal_oblique_hard_y_decl(base, stem .. "in") return end

error("Unrecognized adjective lemma, should end in '-y', '-i', '-ów' or '-in': '" .. base.lemma .. "'") end

decls["irreg"] = function(base) local lemma = base.lemma

if lemma == "dwa" then for _, slot_value in ipairs { {"nom_mp_pers", {"dwaj", "dwaji"}}, {"nom_mp_npers", "dwa"}, {"nom_fp", "dwie"}, {"nom_np", "dwa"}, {"gen_p", "dwùch"}, {"dat_p", {"dwùma", "dwùm"}}, {"ins_p", "dwùm"}, {"loc_p", "dwùch"}, } do			local slot, value = unpack(slot_value) add(base, slot, "", value) end return end

local stem = lemma:match("^([st]w)ój$") if not stem then stem = lemma:match("^(m)ój$") end if stem then add_normal_decl(base, stem,			"ój", "òja", "òje", "òji", "òje",			"òjégò", "òji", "òjich",			"òjémù", "òji", "òjim",			"òjã",			"òjim", "òją", "òjima",			"òjim", "òji", "òjich"		) add_normal_decl(base, stem,			nil, "a", nil, nil, nil,			"égò", "i", {"ich", "ëch"},			"émù", "i", "im",			"ã",			"im", "ą", "ima",			"im", "i", {"ich", "ëch"},			"[rare]"		) return end

local stem1, stem2 if lemma == "naj" or lemma == "naji" or lemma == "nasz" then stem1 = "naj" stem2 = "nasz" elseif lemma == "waj" or lemma == "waji" or lemma == "wasz" then stem1 = "waj" stem2 = "wasz" elseif lemma == "Wasz" then stem2 = "Wasz" end local function add_plural_pron_endings(stem) add_normal_decl(base, stem,			"", "a", "e", "i", "e",			"égò", "i", "ich",			"émù", "i", "im",			"ã",			"im", "ą", "ima",			"im", "i", "ich"		) end if stem1 then add_plural_pron_endings(stem1) add_normal_decl(base, stem1, "i") end if stem2 then add_plural_pron_endings(stem2) return end

stem = lemma:match("^(jed)en$") if stem then add_normal_decl(base, stem, "en") add_normal_decl(base, stem .. "n",			nil, "a", "o", "y", "e",			"égò", "y", "ëch",			"émù", "y", "ym",			"ã",			"ym", "ą", "yma",			"ym", "y", "ëch"		) return end

stem = lemma:match("^(niżód)en$") if stem then add_normal_decl(base, stem, "en") add_normal_oblique_hard_y_decl(base, stem .. "n") return end

error(("Unrecognized irregular lemma '%s'"):format(base.lemma)) end

local function fetch_footnotes(separated_group) local footnotes for j = 2, #separated_group - 1, 2 do		if separated_group[j + 1] ~= "" then error("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

--[=[ Parse a single override spec (e.g. 'ins_mn:autodráhou:autodrahou[rare]' or 'gen_p+loc_p:čobotyx') and return two values: the slot(s) the override applies to, and an object describing the override spec. The input is actually a list where the footnotes have been separated out; for example, given the spec 'ins_p:čobotami:čobotámi[rare]:čobitmi[archaic]', the input will be a list {"ins_p:čobotami:čobotámi", "[rare]", ":čobitmi", "[archaic]", ""}. The object returned for 'ins_mn:autodráhou:autodrahou[rare]' looks like this:

{ values = { {	 form = "autodráhou" },	{	 form = "autodrahou", footnotes = {"[rare]"} } } } ]=] local function parse_override(segments) local retval = {values = {}} local slots = {} local colon_separated_groups = iut.split_alternating_runs_and_strip_spaces(segments, ":") for i, colon_separated_group in ipairs(colon_separated_groups) do		if i == 1 then if #colon_separated_group > 1 then error(("Footnotes not allowed after slot name: '%s'"):format(table.concat(segments))) end local slotspec = colon_separated_group[1] for _, slot in ipairs(rsplit(slotspec, "%+")) do				if not input_adjective_slots[slot] then error(("Unrecognized slot '%s': '%s'"):format(slot, table.concat(segments))) end table.insert(slots, slot) end else local value = {} local form = colon_separated_group[1] if form == "" then error(("Use - to indicate an empty ending for slot%s '%s': '%s'"):format(#slots > 1 and "s" or "", table.concat(slots), table.concat(segments))) elseif form == "-" then value.form = "" else value.form = form end value.footnotes = fetch_footnotes(colon_separated_group) table.insert(retval.values, value) end end return slots, retval end

local function parse_indicator_spec(angle_bracket_spec) local inside = rmatch(angle_bracket_spec, "^<(.*)>$") assert(inside) local base = {overrides = {}, forms = {}} if inside ~= "" then local segments = iut.parse_balanced_segment_run(inside, "[", "]") local dot_separated_groups = iut.split_alternating_runs_and_strip_spaces(segments, "%.") for i, dot_separated_group in ipairs(dot_separated_groups) do			local part = dot_separated_group[1] if rfind(part, ":") then local slots, override = parse_override(dot_separated_group) for _, slot in ipairs(slots) do					if base.overrides[slot] then error(("Two overrides specified for slot '%s'"):format(slot)) else base.overrides[slot] = {override} end end elseif part == "cztok" or part == "irreg" or part == "dva" or part == "plonly" then if #dot_separated_group > 1 then error("Footnotes only allowed with slot overrides or by themselves: '" ..						table.concat(dot_separated_group) .. "'") end if base[part] then error("Can't specify '" .. part .. "' twice: '" .. inside .. "'") end base[part] = true elseif part == "" then if #dot_separated_group == 1 then error("Blank indicator: '" .. inside .. "'") end base.footnotes = fetch_footnotes(dot_separated_group) else error("Unrecognized indicator '" .. part .. "': '" .. inside .. "'") end end end return base end

local function normalize_all_lemmas(alternant_multiword_spec, pagename) iut.map_word_specs(alternant_multiword_spec, function(base)		if base.lemma == "" then			base.lemma = pagename		end		base.orig_lemma = base.lemma		base.orig_lemma_no_links = m_links.remove_links(base.lemma)		base.lemma = base.orig_lemma_no_links	end) end

local function detect_indicator_spec(base) if base.irreg then base.decl = "irreg" if base.lemma == "dwa" or base.lemma == "òba" then base.dva = true end else base.decl = "normal" end if base.cztok and not base.lemma:find("czi$") then error(("Indicator 'cztok' can only be specified with lemmas ending in -czi, but saw '%s'"):format(base.lemma)) end end

local function detect_all_indicator_specs(alternant_multiword_spec) iut.map_word_specs(alternant_multiword_spec, function(base)		detect_indicator_spec(base)		local special = base.plonly and "plonly" or base.dva and "dva" or "regular"		if alternant_multiword_spec.special == nil then			alternant_multiword_spec.special = special		elseif alternant_multiword_spec.special ~= special then			-- We do this because we have a special table with its own slots for each of these special variants.			if special == "dva" then				error("If some alternants are irregular dva or òba, all must be")			elseif special == "plonly" then				error("If some alternants use indicator 'plonly', all must")			else				error("Can't mix regular declension with 'plonly' indicator or with irregular dva or òba")			end		end	end) end

local function process_slot_overrides(base, do_slot) for slot, overrides in pairs(base.overrides) do		local special = base.plonly and "plonly" or base.dva and "dva" or "regular" local allowed_specials = input_adjective_slots[slot] if not allowed_specials then error(("Internal error: Encountered unrecognized slot '%s' not caught during parse_indicator_spec"):format( slot)) end if not m_table.contains(allowed_specials, special) then error(("Override specified for slot '%s' not compatible with table type '%s'"):format(slot, special)) end if do_slot(slot) then base.slot_overridden[slot] = true base.forms[slot] = nil for _, override in ipairs(overrides) do				for _, value in ipairs(override.values) do					local form = value.form local combined_notes = iut.combine_footnotes(base.footnotes, value.footnotes) iut.insert_form(base.forms, slot, {form = form, footnotes = combined_notes}) end end end end end

local function handle_derived_slots_and_overrides(base) local function is_derived_slot(slot) return slot:find("^acc") end

local function is_non_derived_slot(slot) return not is_derived_slot(slot) end

base.slot_overridden = {} -- Handle overrides for the non-derived slots. Do this before generating the derived -- slots so overrides of the source slots (e.g. nom_p) propagate to the derived slots. process_slot_overrides(base, is_non_derived_slot)

local function copy_if(from_slot, to_slot) if not base.forms[to_slot] then iut.insert_forms(base.forms, to_slot, base.forms[from_slot]) end end

copy_if("nom_n", "acc_n") copy_if("gen_mn", "acc_m_an") copy_if("nom_m", "acc_m_in") copy_if("gen_p", "acc_mp_pers") copy_if("nom_p_not_mp_pers", "acc_p_not_mp_pers") copy_if("nom_mp_npers", "acc_mp_npers") copy_if("nom_fp", "acc_fp") copy_if("nom_np", "acc_np")

-- Handle overrides for derived slots, to allow them to be overridden. process_slot_overrides(base, is_derived_slot)

-- Compute linked versions of potential lemma slots, for use in. -- 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(potential_lemma_slots) do iut.insert_forms(base.forms, slot .. "_linked", iut.map_forms(base.forms[slot], function(form) if form == base.orig_lemma_no_links and rfind(base.orig_lemma, "%[%[") then return base.orig_lemma else return form end end)) end end

local function decline_adjective(base) if not decls[base.decl] then error("Internal error: Unrecognized declension type '" .. base.decl .. "'") end decls[base.decl](base) handle_derived_slots_and_overrides(base) end

local function get_decl_from_lemma(base) if base.decl == "irreg" then return "irregular" elseif rfind(base.lemma, "y$") then return "hard" elseif rfind(base.lemma, "i$") then local stem = rmatch(base.lemma, "^(.*)i$") return stem_is_soft(stem) and "soft" or "hard" else return "possessive" end end

local function add_categories(alternant_multiword_spec) local cats = {} local plpos = m_string_utilities.pluralize(alternant_multiword_spec.pos or "adjective") local function insert(cattype) m_table.insertIfNot(cats, "Kashubian " .. cattype .. " " .. plpos) end iut.map_word_specs(alternant_multiword_spec, function(base)		insert(get_decl_from_lemma(base))		if base.overrides.short then			table.insert(cats, "Kashubian " .. plpos .. " with short forms")		end	end) alternant_multiword_spec.categories = cats end

local function show_forms(alternant_multiword_spec) local lemmas = {} for _, slot in ipairs(potential_lemma_slots) do		if alternant_multiword_spec.forms[slot] then for _, formobj in ipairs(alternant_multiword_spec.forms[slot]) do -- FIXME, now can support footnotes as qualifiers in headwords? table.insert(lemmas, formobj.form) end break end end local props = { lemmas = lemmas, slot_table = get_output_adjective_slots(alternant_multiword_spec), lang = lang, }	iut.show_forms(alternant_multiword_spec.forms, props) end

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

local function template_prelude(min_width) return rsub([===[ {title}{annotation} {\op}| border="1px solid #000000" style="border-collapse:collapse;background:#F9F9F9;text-align:center; min-width:MINWIDTHem" class="inflection-table" ]===], "MINWIDTH", min_width) end

local function template_postlude return [=[ end
 * {\cl}{notes_clause}  ]=]

local table_spec_both = template_prelude("75") .. [=[ ! style="background:#d9ebff" colspan=5 | singular ! style="background:#d9ebff" colspan=2 | plural ! style="background:#d9ebff" | ! style="background:#d9ebff" | masculine animate ! style="background:#d9ebff" | masculine inanimate ! style="background:#d9ebff" | feminine ! style="background:#d9ebff" | neuter ! style="background:#d9ebff" | virile (= masculine personal) ! style="background:#d9ebff" | non-virile ! style="background:#eff7ff" | nominative ! style="background:#eff7ff" | genitive ! style="background:#eff7ff" | dative ! style="background:#eff7ff" | accusative ! style="background:#eff7ff" | instrumental ! style="background:#eff7ff" | locative ]=] .. template_postlude
 * colspan=2 | {nom_m}
 * {nom_f}
 * {nom_n}
 * {nom_mp_pers}
 * {nom_p_not_mp_pers}
 * colspan=2 | {gen_mn}
 * {gen_f}
 * {gen_mn}
 * colspan=2 | {gen_p}
 * colspan=2 | {dat_mn}
 * {dat_f}
 * {dat_mn}
 * colspan=2 | {dat_p}
 * {acc_m_an}
 * {acc_m_in}
 * {acc_f}
 * {acc_n}
 * {acc_mp_pers}
 * {acc_p_not_mp_pers}
 * colspan=2 | {ins_mn}
 * {ins_f}
 * {ins_mn}
 * colspan=2 | {ins_p}
 * colspan=2 | {loc_mn}
 * {loc_f}
 * {loc_mn}
 * colspan=2 | {loc_p}{short_clause}

local table_spec_plonly = template_prelude("40") .. [=[ ! style="background:#d9ebff" colspan=5 | plural ! style="background:#d9ebff" | ! style="background:#d9ebff" | virile (= masculine personal) ! style="background:#d9ebff" colspan=3 | non-virile ! style="background:#eff7ff" | nominative ! style="background:#eff7ff" | genitive ! style="background:#eff7ff" | dative ! style="background:#eff7ff" | accusative ! style="background:#eff7ff" | instrumental ! style="background:#eff7ff" | locative ]=] .. template_postlude
 * {nom_mp_pers}
 * colspan=3 | {nom_p_not_mp_pers}
 * colspan=4 | {gen_p}
 * colspan=4 | {dat_p}
 * {acc_mp_pers}
 * colspan=3 | {acc_p_not_mp_pers}
 * colspan=4 | {ins_p}
 * colspan=4 | {loc_p}

local table_spec_dva = template_prelude("55") .. [=[ ! style="width:40%;background:#d9ebff" colspan="2" | ! style="background:#d9ebff" colspan="4" | plural ! style="width:40%;background:#d9ebff" colspan="2" | ! style="background:#d9ebff" | masculine personal ! style="background:#d9ebff" | masculine nonpersonal ! style="background:#d9ebff" | feminine ! style="background:#d9ebff" | neuter ! style="background:#eff7ff" colspan="2" | nominative ! style="background:#eff7ff" colspan="2" | genitive ! style="background:#eff7ff" colspan="2" | dative ! style="background:#eff7ff" colspan="2" | accusative ! style="background:#eff7ff" colspan="2" | instrumental ! style="background:#eff7ff" colspan="2" | locative ]=] .. template_postlude
 * {nom_mp_pers}
 * {nom_mp_npers}
 * {nom_fp}
 * {nom_np}
 * colspan="4" | {gen_p}
 * colspan="4" | {dat_p}
 * {acc_mp_pers}
 * {acc_mp_npers}
 * {acc_fp}
 * {acc_np}
 * colspan="4" | {ins_p}
 * colspan="4" | {loc_p}

local short_template = [=[

! style="background:#eff7ff" | short ]=]
 * colspan=4 | {short}

local notes_template = [===[ {footnote} ]===]

if alternant_multiword_spec.title then forms.title = alternant_multiword_spec.title else forms.title = 'Declension of ' .. forms.lemma .. '' end

local ann_parts = {} local decls = {} iut.map_word_specs(alternant_multiword_spec, function(base)		m_table.insertIfNot(decls, get_decl_from_lemma(base))	end) table.insert(ann_parts, table.concat(decls, " // ")) forms.annotation = " (" .. table.concat(ann_parts, ", ") .. ")"

forms.notes_clause = forms.footnote ~= "" and m_string_utilities.format(notes_template, forms) or "" forms.short_clause = forms.short and forms.short ~= "—" and m_string_utilities.format(short_template, forms) or "" return m_string_utilities.format(		alternant_multiword_spec.special == "plonly" and table_spec_plonly or		alternant_multiword_spec.special == "dva" and table_spec_dva or		table_spec_both, forms	) end

-- Externally callable function to parse and decline an adjective given user-specified arguments. Return value is -- ALTERNANT_MULTIWORD_SPEC, an object where the declined forms are in `ALTERNANT_MULTIWORD_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, pos, from_headword) local params = { [1] = {},		pos = {}, json = {type = "boolean"}, -- for use with bots title = {}, pagename = {}, }	for slot, _ in pairs(input_adjective_slots) do		params[slot] = {} end

-- Only default param 1 when displaying the template. local args = require("Module:parameters").process(parent_args, params) local SUBPAGE = mw.title.getCurrentTitle.subpageText local pagename = args.pagename or SUBPAGE if not args[1] then if SUBPAGE == "csb-adecl" then args[1] = "bëlny<>" else args[1] = pagename end end local parse_props = { parse_indicator_spec = parse_indicator_spec, allow_default_indicator = true, allow_blank_lemma = true, }	local alternant_multiword_spec = iut.parse_inflected_text(args[1], parse_props) alternant_multiword_spec.pos = args.pos alternant_multiword_spec.title = args.title alternant_multiword_spec.forms = {} normalize_all_lemmas(alternant_multiword_spec, pagename) detect_all_indicator_specs(alternant_multiword_spec) local inflect_props = { slot_table = get_output_adjective_slots(alternant_multiword_spec), inflect_word_spec = decline_adjective, }	iut.inflect_multiword_or_alternant_multiword_spec(alternant_multiword_spec, inflect_props) add_categories(alternant_multiword_spec) if args.json and not from_headword then return require("Module:JSON").toJSON(alternant_multiword_spec) end return alternant_multiword_spec end

-- Entry point for. Template-callable function to parse and decline an adjective given user-specified -- arguments and generate a displayable table of the declined 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

return export