मॉड्यूल:affix

"इस मॉड्यूल हेतु प्रलेख मॉड्यूल:affix/doc पर बनाया जा सकता है"

local export = {}

local debug_force_cat = false -- if set to true, always display categories even on userspace pages

local m_links = require("Module:links")
local m_utilities = require("Module:utilities")
local m_table = require("Module:table")
local affix_lang_data_module_prefix = "Module:affix/lang-data/"

local u = mw.ustring.char
local rsub = mw.ustring.gsub
local usub = mw.ustring.sub
local ulen = mw.ustring.len
local rfind = mw.ustring.find
local rmatch = mw.ustring.match


-- Export this so the category code in [[Module:category tree/poscatboiler/data/terms by etymology]] can access it.
export.langs_with_lang_specific_data = {
	["fi"] = true,
	["tr"] = true,
}

local default_pos = "term"

--[=[
About different types of hyphens ("template", "display" and "lookup"):
----------------------------------------------------------------------

* The "template hyphen" is the per-script hyphen character that is used in template calls to indicate that a term is an
  affix. This is always a single Unicode char, but there may be multiple possible hyphens for a given script. Normally
  this is just the regular hyphen character "-", but for some non-Latin-script languages (currently only right-to-left
  languages), it is different.

* The "display hyphen" is the string (which might be an empty string) that is added onto a term as displayed and linked,
  to indicate that a term is an affix. Currently this is always either the same as the template hyphen or an empty
  string, but the code below is written generally enough to handle arbitrary display hyphens. Specifically:

  (1) For East Asian languages, the display hyphen is always blank.
  (2) For Arabic-script languages, either tatweel (ـ) or ZWNJ (zero-width non-joiner) are allowed as template hyphens,
 	  where ZWNJ is supported primarily for Farsi, because some suffixes have non-joining behavior. The display hyphen
	  corresponding to tatweel is also tatweel, but the display hyphen corresponding to ZWNJ is blank (tatweel is also
	  the default display hyphen, for calls to {{prefix}}/{{suffix}}/etc. that don't include an explicit hyphen).

* The "lookup hyphen" is the hyphen that is used when looking up language-specific affix mappings. (These mappings are
  discussed in more detail below when discussing link affixes.) It depends only on the script of the affix in question.
  Most scripts (including East Asian scripts) use a regular hyphen "-" as the lookup hyphen, but Hebrew and Arabic
  have their own lookup hyphens (respectively maqqef and tatweel). Note that for Arabic in particular, there are
  three possible template hyphens that are recognized (tatweel, ZWNJ and regular hyphen), but mappings must use tatweel.


About different types of affixes ("template", "display", "link", "lookup" and "category"):
------------------------------------------------------------------------------------------

* A "template affix" is an affix in its source form as it appears in a template call. Generally, a template affix has
  an attached template hyphen (see above) to indicate that it is an affix and indicate what type of affix it is
  (prefix, suffix, interfix/infix or circumfix), but some of the older-style templates such as {{tl|suffix}},
  {{tl|prefix}}, {{tl|confix}}, etc. have "positional" affixes where the presence of the affix in a certain position
  (e.g. the second or third parameter) indicates that it is a certain type of affix, whether or not it has an attached
  template hyphen.
* A "display affix" is the corresponding affix as it is actually displayed to the user. The display affix may differ
  from the template affix for various reasons:
  (1) The display affix may be specified explicitly using the |altN= parameter, the <alt:...> inline modifier or a
	  piped link of the form e.g. [[-kas|-käs]] (here indicating that the affix should display as '-käs' but be linked
	  as '-kas'). Here, the template affix is arguably the entire piped link, while the display affix is '-käs'.
  (2) Even in the absence of |altN= parameters, <alt:...> inline modifiers and piped links, certain languages have
	  differences between the "template hyphen" specified in the template (which always needs to be specified somehow
	  or other in templates like {{affix}}, to indicate that the term is an affix and what type of affix it is) and the
	  display hyphen (see above), with corresponding differences between template and display affixes.
* A "link affix" is the affix that is linked to when the affix is shown to the user. The link affix is usually the same
  as the display affix, but will differ in one of three circumstances:
  (1) The display and link affixes are explicitly made different using |altN= parameters, <alt:...> inline modifiers or
	  piped links, as described above under "display affix".
  (2) For certain languages, certain affixes are mapped to canonical form using language-specific mappings. For example,
      in Finnish, the adjective-forming suffix [[-kas]] appears as [[-käs]] after front vowels, but logically both
	  forms are the same suffix and should be linked and categorized the same. Similarly, in Latin, the negative and
	  intensive prefixes spelled [[in-]] (etymologically two distinct prefixes) appear variously as [[il-]], [[im-]] or
	  [[ir-]] before certain consonants. Mappings are supplied in [[Module:affix/lang-data/LANGCODE]] to convert
	  Finnish [[-käs]] to [[-kas]] for linking and categorization purposes. Note that the affixes in the mappings use
	  "lookup hyphens" to indicate the different types of affixes, which is usually the same as the template hyphen but
	  differs for Arabic scripts, because there are multiple possible template hyphens recognized but only one lookup
	  hyphen (tatweel). The form of the affix as used to look up in the mapping tables is called the "lookup affix";
	  see below.
  (3) The link affix is passed through the language's makeEntryName() function, which may strip certain diacritics:
      e.g. macrons in Latin and Old English (indicating length); acute and grave accents in Russian and various other
	  Slavic languages (indicating stress); vowel diacritics in most Arabic-script languages; and also tatweel in some
	  Arabic-script languages (currently, for example, Persian, Arabic and Urdu strip tatweel, but Ottoman Turkish does
	  not).
* A "lookup affix" is the form of the affix as it is looked up in the language-specific lookup mappings described above
  under link affixes. There are actually two lookup stages:
  (1) First, the affix is looked up in a modified display form (specifically, the same as the display affix but using
      lookup hyphens). Note that this lookup does not occur if an explicit display form is given using |altN= or an
	  <alt:...> inline modifier, or if the template affix contains a piped or embedded link.
  (2) If no entry is found, the affix is then looked up in a modified link form (specifically, the modified display
      form passed through the language's makeEntryName() function, which strips out certain diacritics, but with the
	  lookup hyphen re-added if it was stripped out, as in the case of tatweel in many Arabic-script languages).
  The reason for this double lookup procedure is to allow for mappings that are sensitive to the extra diacritics, but
  also allow for mappings that are not sensitive in this fashion (e.g. Russian [[-ливый]] occurs both stressed and
  unstressed, but is the same prefix either way).
* A "category affix" is the affix as it appears in categories such as [[:Category:Finnish terms suffixed with -kas]].
  The category affix is currently always the same as the link affix.
]=]

-----------------------------------------------------------------------------------------
--                               Template and display hyphens                          --
-----------------------------------------------------------------------------------------

--[=[
Per-script template hyphens. The template hyphen is what appears in the {{affix}}/{{prefix}}/{{suffix}}/etc. template
(in the wikicode). See above.

They key below is a script code, after removing a hyphen and anything preceding. Hence, script codes like 'fa-Arab'
and 'ur-Arab' will match 'Arab'.

The value below is a string consisting of one or more hyphen characters. If there is more than one character, the
default hyphen must come last and a non-default function must be specified for the script in display_hyphens[] so
the correct display hyphen will be specified when no template hyphen is given (in {{suffix}}/{{prefix}}/etc.).

Script detection is normally done when linking, but we need to do it earlier. However, under most circumstances we
don't need to do script detection. Specifically, we only need to do script detection for a given language if

(a) the language has multiple scripts; and
(b) at least one of those scripts is listed below or in display_hyphens.
]=]

local ZWNJ = u(0x200C) -- zero-width non-joiner
local template_hyphens = {
	["Hebr"] = "־", -- Hebrew-specific hyphen termed "maqqef"
	-- This covers all Arabic scripts. See above.
	["Arab"] = "ـ" .. ZWNJ .. "-", -- tatweel + zero-width non-joiner + regular hyphen
	-- FIXME! What about the following right-to-left scripts?
	-- Adlm (Adlam)
	-- Armi (Imperial Aramaic)
	-- Avst (Avestan)
	-- Cprt (Cypriot)
	-- Khar (Kharoshthi)
	-- Mand (Mandaic/Mandaean)
	-- Mani (Manichaean)
	-- Mend (Mende/Mende Kikakui)
	-- Narb (Old North Arabian)
	-- Nbat (Nabataean/Nabatean)
	-- Nkoo (N'Ko)
	-- Orkh (Orkhon runes)
	-- Phli (Inscriptional Pahlavi)
	-- Phlp (Psalter Pahlavi)
	-- Phlv (Book Pahlavi)
	-- Phnx (Phoenician)
	-- Prti (Inscriptional Parthian)
	-- Rohg (Hanifi Rohingya)
	-- Samr (Samaritan)
	-- Sarb (Old South Arabian)
	-- Sogd (Sogdian)
	-- Sogo (Old Sogdian)
	-- Syrc (Syriac)
	-- Thaa (Thaana)
}

-- Hyphens used when looking up an affix in a lang-specific affix mapping. Defaults to regular hyphen (-). The keys
-- are script codes, after removing a hyphen and anything preceding. Hence, script codes like 'fa-Arab' and 'ur-Arab'
-- will match 'Arab'. The value should be a single character.
local lookup_hyphens = {
	["Hebr"] = "־",
	-- This covers all Arabic scripts. See above.
	["Arab"] = "ـ",
}

-- Default display-hyphen function.
local function default_display_hyphen(script, hyph)
	if not hyph then
		return template_hyphens[script] or "-"
	end
	return hyph
end

local function arab_get_display_hyphen(script, hyph)
	if not hyph then
		return "ـ" -- tatweel
	elseif hyph == ZWNJ then
		return ""
	else
		return hyph
	end
end

local function no_display_hyphen(script, hyph)
	return ""
end

-- Per-script function to return the correct display hyphen given the script and template hyphen. The function should
-- also handle the case where the passed-in template hyphen is nil, corresponding to the situation in
-- {{prefix}}/{{suffix}}/etc. where no template hyphen is specified. The key is the script code after removing a hyphen
-- and anything preceding, so 'fa-Arab', 'ur-Arab' etc. will match 'Arab'.
local display_hyphens = {
	-- This covers all Arabic scripts. See above.
	["Arab"] = arab_get_display_hyphen,
	["Hani"] = no_display_hyphen,
	-- The following is a mixture of several scripts. Hopefully the specs here are correct!
	["Jpan"] = no_display_hyphen,
	["Laoo"] = no_display_hyphen,
	["Nshu"] = no_display_hyphen,
	["Thaa"] = no_display_hyphen,
	["Thai"] = no_display_hyphen,
}

-----------------------------------------------------------------------------------------
--                                 Basic Utility functions                             --
-----------------------------------------------------------------------------------------

local function glossary_link(entry, text)
	text = text or entry
	return "[[Appendix:Glossary#" .. entry .. "|" .. text .. "]]"
end


local function track(page)
	if type(page) == "table" then
		for i, pg in ipairs(page) do
			page[i] = "affix/" .. pg
		end
	else
		page = "affix/" .. page
	end
	require("Module:debug/track")(page)
end


-----------------------------------------------------------------------------------------
--                                      Compound types                                 --
-----------------------------------------------------------------------------------------

local function make_compound_type(typ, alttext)
	return {
		text = glossary_link(typ, alttext) .. " compound",
		cat = typ .. " compounds",
	}
end


-- Make a compound type entry with a simple rather than glossary link.
-- These should be replaced with a glossary link when the entry in the glossary
-- is created.
local function make_non_glossary_compound_type(typ, alttext)
	local link = alttext and "[[" .. typ .. "|" .. alttext .. "]]" or "[[" .. typ .. "]]"
	return {
		text = link .. " compound",
		cat = typ .. " compounds",
	}
end


export.compound_types = {
	["alliterative"] = make_non_glossary_compound_type("alliterative"),
	["allit"] = "alliterative",
	["antonymous"] = make_non_glossary_compound_type("antonymous"),
	["ant"] = "antonymous",
	["bahuvrihi"] = make_compound_type("bahuvrihi", "bahuvrīhi"),
	["bahu"] = "bahuvrihi",
	["bv"] = "bahuvrihi",
	["coordinative"] = make_compound_type("coordinative"),
	["coord"] = "coordinative",
	["descriptive"] = make_compound_type("descriptive"),
	["desc"] = "descriptive",
	["determinative"] = make_compound_type("determinative"),
	["det"] = "determinative",
	["dvandva"] = make_compound_type("dvandva"),
	["dva"] = "dvandva",
	["endocentric"] = make_compound_type("endocentric"),
	["endo"] = "endocentric",
	["exocentric"] = make_compound_type("exocentric"),
	["exo"] = "exocentric",
	["karmadharaya"] = make_compound_type("karmadharaya", "karmadhāraya"),
	["karma"] = "karmadharaya",
	["kd"] = "karmadharaya",
	["rhyming"] = make_non_glossary_compound_type("rhyming"),
	["rhy"] = "rhyming",
	["synonymous"] = make_non_glossary_compound_type("synonymous"),
	["syn"] = "synonymous",
	["tatpurusa"] = make_compound_type("tatpurusa", "tatpuruṣa"),
	["tat"] = "tatpurusa",
	["tp"] = "tatpurusa",
}


local function process_compound_type(typ, nocap, notext, has_parts)
	local text_sections = {}
	local categories = {}

	if typ then
		local typdata = export.compound_types[typ]
		if type(typdata) == "string" then
			typdata = export.compound_types[typdata]
		end
		if not typdata then
			error("Internal error: Unrecognized type '" .. typ .. "'")
		end
		local text = typdata.text
		if not nocap then
			text = require("Module:string utilities").ucfirst(text)
		end
		local cat = typdata.cat
		local oftext = typdata.oftext or " of"

		if not notext then
			table.insert(text_sections, text)
			if has_parts then
				table.insert(text_sections, oftext)
				table.insert(text_sections, " ")
			end
		end
		table.insert(categories, cat)
	end

	return text_sections, categories
end


-----------------------------------------------------------------------------------------
--                                     Utility functions                               --
-----------------------------------------------------------------------------------------

local function glossary_link(entry, text)
	text = text or entry
	return "[[Appendix:Glossary#" .. entry .. "|" .. text .. "]]"
end


local function track(page)
	if type(page) == "table" then
		for i, pg in ipairs(page) do
			page[i] = "affix/" .. pg
		end
	else
		page = "affix/" .. page
	end
	require("Module:debug/track")(page)
end


-- Iterate an array up to the greatest integer index found.
local function ipairs_with_gaps(t)
	local indices = require("Module:table").numKeys(t)
	local max_index = #indices > 0 and math.max(unpack(indices)) or 0
	local i = 0
	return function()
		while i < max_index do
			i = i + 1
			return i, t[i]
		end
	end
end

export.ipairs_with_gaps = ipairs_with_gaps


--[=[
Concatenate formatted parts together with any overall lit= spec plus categories, which are formatted by prepending the
language name. The value of an entry in CATEGORIES can be either a string (which is formatted using SORT_KEY) or a
table of the form {cat=CATEGORY, sort_key=SORT_KEY, sort_base=SORT_BASE}, specifying the sort key and sort base to use
when formatting the category. If NOCAT is given, no categories are added.
]=]
function export.concat_parts(lang, parts_formatted, categories, nocat, sort_key, lit, force_cat)
	local cattext
	if nocat then
		cattext = ""
	else
		for i, cat in ipairs(categories) do
			if type(cat) == "table" then
				categories[i] = m_utilities.format_categories({lang:getNonEtymologicalName() .. " " .. cat.cat}, lang,
					cat.sort_key, cat.sort_base, force_cat or debug_force_cat)
			else
				categories[i] = m_utilities.format_categories({lang:getNonEtymologicalName() .. " " .. cat}, lang,
					sort_key, nil, force_cat or debug_force_cat)
			end
		end
		cattext = table.concat(categories)
	end
	return table.concat(parts_formatted, " +&lrm; ") .. (lit and ", literally " .. m_links.mark(lit, "gloss") or "") ..
		cattext
end


local function pluralize(pos)
	if pos ~= "nouns" and usub(pos, -5) ~= "verbs" and usub(pos, -4) ~= "ives" then
		if pos:find("[sx]$") then
			pos = pos .. "es"
		else
			pos = pos .. "s"
		end
	end
	return pos
end


local function make_entry_name_no_links(lang, term)
	-- Remove links and call lang:makeEntryName(term).
	return (lang:makeEntryName(m_links.remove_links(term)))
end


--[=[
WARNING: This destructively writes into `terminfo`.
]=]
function export.link_term(terminfo, lang, sc, sort_key, force_cat, nocat)
	local result

	terminfo.sc = terminfo.sc or sc
	if terminfo.lang then
		result = require("Module:etymology").format_derived(lang, terminfo, sort_key, nocat,
			force_cat or debug_force_cat)
	else
		terminfo.lang = lang
		result = m_links.full_link(terminfo, "term", false)
	end

	if terminfo.q then
		track("q-after-result")
		error("Please use qqN= or <qq:...> instead of qN= or <q:...> to put a qualifier after the term; qN= will be changing to put the qualifier before the term")
		result = result .. " " .. require("Module:qualifier").format_qualifier(terminfo.q)
	end

	if terminfo.qq then
		result = result .. " " .. require("Module:qualifier").format_qualifier(terminfo.qq)
	end

	return result
end


local function canonicalize_script_code(scode)
	-- Convert fa-Arab, ur-Arab etc. to Arab.
	return (scode:gsub("^.*%-", ""))
end


-----------------------------------------------------------------------------------------
--                                  Affix-handling functions                           --
-----------------------------------------------------------------------------------------

-- Figure out the appropriate script for the given affix and language (unless the script is explicitly passed in), and
-- return the values of template_hyphens[], display_hyphens[] and lookup_hyphens[] for that script, substituting
-- default values as appropriate. Four values are returned:
--	DETECTED_SCRIPT, TEMPLATE_HYPHEN, DISPLAY_HYPHEN, LOOKUP_HYPHEN
local function detect_script_and_hyphens(text, lang, sc)
	local scode
	-- 1. If the script is explicitly passed in, use it.
	if sc then
		scode = sc:getCode()
	else
		local possible_script_codes = lang:getScriptCodes()
		-- YUCK! `possible_script_codes` comes from loadData() so #possible_scripts doesn't work (always returns 0).
		local num_possible_script_codes = m_table.length(possible_script_codes)
		if num_possible_script_codes == 0 then
			-- This shouldn't happen; if the language has no script codes,
			-- the list {"None"} should be returned.
			error("Something is majorly wrong! Language " .. lang:getNonEtymologicalName() .. " has no script codes.")
		end
		if num_possible_script_codes == 1 then
			-- 2. If the language has only one possible script, use it.
			scode = possible_script_codes[1]
		else
			-- 3. Check if any of the possible scripts for the language have non-default values for template_hyphens[]
			--    or display_hyphens[]. If so, we need to do script detection on the text. If not, just use "Latn",
			--    which may not be technically correct but produces the right results because Latn has all default
			--    values for template_hyphens[] and display_hyphens[].
			local may_have_nondefault_hyphen = false
			for _, script_code in ipairs(possible_script_codes) do
				script_code = canonicalize_script_code(script_code)
				if template_hyphens[script_code] or display_hyphens[script_code] then
					may_have_nondefault_hyphen = true
					break
				end
			end
			if not may_have_nondefault_hyphen then
				scode = "Latn"
			else
				scode = lang:findBestScript(text):getCode()
			end
		end
	end
	scode = canonicalize_script_code(scode)
	local template_hyphen = template_hyphens[scode] or "-"
	local lookup_hyphen = lookup_hyphens[scode] or "-"
	local display_hyphen = display_hyphens[scode] or default_display_hyphen
	return scode, template_hyphen, display_hyphen, lookup_hyphen
end


--[=[
Given a template affix `term` and an affix type `affix_type`, change the relevant template hyphen(s) in the affix to
the display or lookup hyphen specified in `new_hyphen`, or add them if they are missing. `new_hyphen` can be a string,
specifying a fixed hyphen, or a function of two arguments (the script code `scode` and the discovered template hyphen,
or nil of no relevant template hyphen is present). `thyph_re` is a Lua pattern (which must be enclosed in parens) that
matches the possible template hyphens. Note that not all template hyphens present in the affix are changed, but only
the "relevant" ones (e.g. for a prefix, a relevant template hyphen is one coming at the end of the affix).
]=]
local function reconstruct_term_per_hyphens(term, affix_type, scode, thyph_re, new_hyphen)
	local function get_hyphen(hyph)
		if type(new_hyphen) == "string" then
			return new_hyphen
		end
		return new_hyphen(scode, hyph)
	end

	if not affix_type then
		return term
	elseif affix_type == "circumfix" then
		local before, before_hyphen, after_hyphen, after = rmatch(term, "^(.*)" .. thyph_re .. " " .. thyph_re
			.. "(.*)$")
		if not before or ulen(term) <= 3 then
			-- Unlike with other types of affixes, don't try to add hyphens in the middle of the term to convert it to
			-- a circumfix. Also, if the term is just hyphen + space + hyphen, return it.
			return term
		end
		return before .. get_hyphen(before_hyphen) .. " " .. get_hyphen(after_hyphen) .. after
	elseif affix_type == "infix" or affix_type == "interfix" then
		local before_hyphen, middle, after_hyphen = rmatch(term, "^" .. thyph_re .. "(.*)" .. thyph_re .. "$")
		if before_hyphen and ulen(term) <= 1 then
			-- If the term is just a hyphen, return it.
			return term
		end
		return get_hyphen(before_hyphen) .. (middle or term) .. get_hyphen(after_hyphen)
	elseif affix_type == "prefix" then
		local middle, after_hyphen = rmatch(term, "^(.*)" .. thyph_re .. "$")
		if middle and ulen(term) <= 1 then
			-- If the term is just a hyphen, return it.
			return term
		end
		return (middle or term) .. get_hyphen(after_hyphen)
	elseif affix_type == "suffix" then
		local before_hyphen, middle = rmatch(term, "^" .. thyph_re .. "(.*)$")
		if before_hyphen and ulen(term) <= 1 then
			-- If the term is just a hyphen, return it.
			return term
		end
		return get_hyphen(before_hyphen) .. (middle or term)
	else
		error(("Internal error: Unrecognized affix type '%s'"):format(affix_type))
	end
end


local function lookup_affix_mapping(affix, affix_type, lang, scode, thyph_re, lookup_hyph)
	local function do_lookup(affix)
		local langcode = lang:getCode()
		-- Ensure that the affix uses lookup hyphens regardless of whether it used a different type of hyphens before
		-- or no hyphens.
		local lookup_affix = reconstruct_term_per_hyphens(affix, affix_type, scode, thyph_re, lookup_hyph)
		if export.langs_with_lang_specific_data[langcode] then
			local langdata = mw.loadData(affix_lang_data_module_prefix .. langcode)
			-- If this is a canonical long-form tag, just return it, and don't check for shortcuts. This is an
			-- optimization; see below.
			if langdata.affix_mappings then
				local mapping = langdata.affix_mappings[lookup_affix]
				if mapping then
					return mapping
				end
			end
		end

		return nil
	end

	if affix:find("%[%[") then
		return nil
	end

	-- Double parens because makeEntryName() returns multiple values. Yuck.
	return do_lookup(affix) or do_lookup((lang:makeEntryName(affix))) or nil
end


--[=[
For a given template term in a given language (see the definition of "template affix" near the top of the file),
possibly in an explicitly specified script `sc` (but usually nil), return the term's affix type ("prefix", "infix",
"suffix", "circumfix" or nil for non-affix) along with the corresponding link and display affixes (see definitions
near the top of the file). Three values are returned: `affix_type`, `link_term`, `display_term`. The affix type can
be passed in instead of autodetected (pass in 'false' if the term is not an affix); in this case, the template term
need not have any attached hyphens, and the appropriate hyphens will be added in the appropriate places. If
`do_affix_mapping` is specified, look up the affix in the lang-specific affix mappings, as described in the comment at
the top of the file; otherwise, the link and display terms will always be the same. (They will be the same in any case
if the template term has a bracketed link in it or is not an affix.)
]=]
local function parse_term_for_affixes(term, lang, sc, affix_type, do_affix_mapping)
	if not term then
		return nil, nil, nil
	end

	if term:find("^%^") then
		-- If term begins with ^, it's not an affix no matter what. Strip off the ^ and return "no affix".
		term = usub(term, 2)
		return nil, term, term
	end

	-- Remove an asterisk if the morpheme is reconstructed and add it back at the end.
	local reconstructed = ""
	if term:find("^%*") then 
		reconstructed = "*"
		term = term:gsub("^%*", "")
	end

	local scode, thyph, dhyph, lhyph = detect_script_and_hyphens(term, lang, sc)
	thyph = "([" .. thyph .. "])"

	if affix_type == nil then
		if rfind(term, thyph .. " " .. thyph) then
			affix_type = "circumfix"
		else
			local has_beginning_hyphen = rfind(term, "^" .. thyph)
			local has_ending_hyphen = rfind(term, thyph .. "$")
			if has_beginning_hyphen and has_ending_hyphen then
				affix_type = "infix"
			elseif has_ending_hyphen then
				affix_type = "prefix"
			elseif has_beginning_hyphen then
				affix_type = "suffix"
			end
		end
	end

	local link_term, display_term
	if affix_type then
		display_term = reconstruct_term_per_hyphens(term, affix_type, scode, thyph, dhyph)
		if do_affix_mapping then
			link_term = lookup_affix_mapping(term, affix_type, lang, scode, thyph, lhyph)
			-- The return value of lookup_affix_mapping() may be an affix mapping with lookup hyphens if a mapping
			-- was found, otherwise nil if a mapping was not found. We need to convert to display hyphens in
			-- either case, but in the latter case we can reuse the display term, which has already been converted.
			if link_term then
				link_term = reconstruct_term_per_hyphens(link_term, affix_type, scode, thyph, dhyph)
			else
				link_term = display_term
			end
		else
			link_term = display_term
		end
	else
		link_term = term
		display_term = term
	end

	link_term = reconstructed .. link_term
	display_term = reconstructed .. display_term

	return affix_type, link_term, display_term
end

export.get_affix_type = parse_term_for_affixes



-- Add a hyphen to a term in the appropriate place, based on the specified affix type. For example, if `affix_type` ==
-- "prefix", we'll add a hyphen onto the end if it's not already there. In general, if the template and display hyphens
-- are the same and the appropriate hyphen is already present, we leave it, else we strip off the template hyphen if
-- present and add the display hyphen.
function export.make_affix(term, lang, sc, affix_type, do_affix_mapping)
	if not (affix_type == "prefix" or affix_type == "suffix" or affix_type == "circumfix" or affix_type == "infix" or
		affix_type == "interfix") then
		error("Internal error: Invalid affix type " .. (affix_type or "(nil)"))
	end

	local _, link_term, display_term = parse_term_for_affixes(term, lang, sc, affix_type, do_affix_mapping)
	return link_term, display_term
end


-----------------------------------------------------------------------------------------
--                                     Main entry points                               --
-----------------------------------------------------------------------------------------

--[=[
Implementation of {{affix}} and {{surface analysis}}. `data` contains all the information describing the affixes to be
displayed, and contains the following:

`.lang`: Overall language object (REQUIRED). Different from term-specific language objects (see `.parts` below).
`.sc`: Overall script object (usually omitted). Different from term-specific script objects.
`.parts`: List of objects describing the affixes to show (REQUIRED). The general format of each object is as would be
		  passed to full_link(), except that the `.lang` field should be missing unless the term is of a language
		  different from the overall `.lang` value (in such a case, the language name is shown along with the term and
		  an additional "derived from" category is added). WARNING: The data in `.parts` will be destructively modified.
`.pos`: Overall part of speech (used in categories, defaults to "terms"). Different from term-specific part of speech.
`.sort_key`: Overall sort key. Normally omitted except e.g. in Japanese.
`.type`: Type of compound, if the parts in `.parts` describe a compound. Strictly optional, and if supplied, the
		compound type is displayed before the parts (normally capitalized, unless `.nocap` is given).
`.nocap`: Don't capitalize the first letter of text displayed before the parts (relevant only if `.type` or
		  `.surface_analysis` is given).
`.notext`: Don't display any text before the parts (relevant only if `.type` or `.surface_analysis` is given).
`.nocat`: Disable all categorization.
`.lit`: Overall literal definition. Different from term-specific literal definitions.
`.force_cat`: Always display categories, even on userspace pages.
`.surface_analysis`: Implement {{surface analysis}}; adds "By surface analysis, " before the parts.

WARNING: This destructively modifies both `data` and the individual structures within `.parts`.
]=]
function export.show_affix(data)
	data.pos = data.pos or default_pos

	data.pos = pluralize(data.pos)

	local text_sections, categories = process_compound_type(data.type, data.surface_analysis or data.nocap, data.notext,
		#data.parts > 0)

	-- Process each part
	local parts_formatted = {}
	local whole_words = 0

	for i, part in ipairs_with_gaps(data.parts) do
		part = part or {}
		local part_lang = part.lang or data.lang
		local part_sc = part.sc or data.sc

		-- Determine affix type and get link and display terms (see text at top of file).
		local affix_type, link_term, display_term = parse_term_for_affixes(part.term, part_lang, part_sc, nil,
			not part.alt)

		-- If link_term is an empty string, either a bare ^ was specified or an empty term was used along with inline
		-- modifiers. The intention in either case is not to link the term.
		part.term = link_term ~= "" and link_term or nil
		part.alt = part.alt or display_term

		-- Make a link for the part
		table.insert(parts_formatted, export.link_term(part, data.lang, data.sc, data.sort_key, data.force_cat,
			data.nocat))

		if affix_type then
			-- We cannot distinguish interfixes from infixes by appearance. Prefer interfixes; infixes will need to
			-- use {{infix}}.
			if affix_type == "infix" then affix_type = "interfix" end

			-- Make a sort key. For the first part, use the second part as the sort key; the intention is that if the
			-- term has a prefix, sorting by the prefix won't be very useful so we sort by what follows, which is
			-- presumably the root.
			local part_sort_base = nil
			local part_sort = part.sort or data.sort_key

			if i == 1 and data.parts[2] and data.parts[2].term then
				local part2_lang = data.parts[2].lang or data.lang
				local part2_sc = data.parts[2].sc or data.sc
				local part2_affix_type, part2_link_term, part2_display_term = parse_term_for_affixes(data.parts[2].term,
					part2_lang, part2_sc, nil, not data.parts[2].alt)
				part_sort_base = make_entry_name_no_links(part2_lang, part2_link_term)
			end

			if part.pos and rfind(part.pos, "patronym") then
				table.insert(categories, {cat="patronymics", sort_key=part_sort, sort_base=part_sort_base})
			end

			if data.pos ~= "terms" and part.pos and rfind(part.pos, "diminutive") then
				table.insert(categories, {cat="diminutive " .. data.pos, sort_key=part_sort, sort_base=part_sort_base})
			end

			if link_term and link_term ~= "" then
				table.insert(categories, {cat=data.pos .. " " .. affix_type .. "ed with " ..
					make_entry_name_no_links(part_lang, link_term) .. (part.id and " (" .. part.id .. ")" or ""),
					sort_key=part_sort, sort_base=part_sort_base})
			end
		else
			whole_words = whole_words + 1

			if whole_words == 2 then
				table.insert(categories, "compound " .. data.pos)
			end
		end
	end

	-- If there are no categories, then there were no actual affixes, only a single regular term.
	if #categories == 0 then
		error("The parameters did not include any affixes, and the term is not a compound. Please provide at least one affix.")
	end

	if data.surface_analysis then
		local text = "by " .. glossary_link("surface analysis") .. ", "
		if not data.nocap then
			text = require("Module:string utilities").ucfirst(text)
		end

		table.insert(text_sections, 1, text)
	end

	table.insert(text_sections, export.concat_parts(data.lang, parts_formatted, categories, data.nocat, data.sort_key,
		data.lit, data.force_cat))
	return table.concat(text_sections)
end


function export.show_surface_analysis(data)
	data.surface_analysis = true
	return export.show_affix(data)
end


--[=[
Implementation of {{compound}}.

WARNING: This destructively modifies both `data` and the individual structures within `.parts`.
]=]
function export.show_compound(data)
	data.pos = data.pos or default_pos

	data.pos = pluralize(data.pos)

	local text_sections, categories = process_compound_type(data.type, data.nocap, data.notext, #data.parts > 0)

	local parts_formatted = {}
	table.insert(categories, "compound " .. data.pos)

	-- Make links out of all the parts
	local whole_words = 0
	for i, part in ipairs(data.parts) do
		local part_lang = part.lang or data.lang
		local part_sc = part.sc or data.sc
		-- Determine affix type and get link and display terms (see text at top of file).
		local affix_type, link_term, display_term = parse_term_for_affixes(part.term, part_lang, part_sc, nil,
			not part.alt)

		-- If the term is an infix, recognize it as such (which means e.g. that we will display the term without
		-- hyphens for East Asian languages). Otherwise, ignore the fact that it looks like an affix and display as
		-- specified in the template (but pay attention to the detected affix type for certain tracking purposes).
		if affix_type == "infix" then
			-- If link_term is an empty string, either a bare ^ was specified or an empty term was used along with
			-- inline modifiers. The intention in either case is not to link the term.
			if link_term and link_term ~= "" then
				table.insert(categories, {cat=data.pos .. " interfixed with " .. make_entry_name_no_links(part_lang,
					link_term), sort_key=part.sort or data.sort_key})
			end
			part.term = link_term ~= "" and link_term or nil
			part.alt = part.alt or display_term
		else
			if affix_type then
				track { affix_type, affix_type .. "/lang/" .. data.lang:getNonEtymologicalCode() }
			else
				whole_words = whole_words + 1
			end
		end
		table.insert(parts_formatted, export.link_term(part, data.lang, data.sc, data.sort_key, data.force_cat,
			data.nocat))
	end

	if whole_words == 1 then
		track("one whole word")
	elseif whole_words == 0 then
		track("looks like confix")
	end

	table.insert(text_sections, export.concat_parts(data.lang, parts_formatted, categories, data.nocat, data.sort_key,
		data.lit, data.force_cat))
	return table.concat(text_sections)
end


--[=[
Implementation of {{blend}}, {{univerbation}} and similar "compound-like" templates.

WARNING: This destructively modifies both `data` and the individual structures within `.parts`.
]=]
function export.show_compound_like(data)
	local parts_formatted = {}
	local categories = {}

	if data.cat then
		table.insert(categories, data.cat)
	end

	-- Make links out of all the parts
	for i, part in ipairs(data.parts) do
		table.insert(parts_formatted, export.link_term(part, data.lang, data.sc, data.sort_key, data.force_cat,
			data.nocat))
	end

	local text_sections = {}
	if data.text then
		table.insert(text_sections, data.text)
	end
	if #data.parts > 0 and data.oftext then
		table.insert(text_sections, " ")
		table.insert(text_sections, data.oftext)
		table.insert(text_sections, " ")
	end
	table.insert(text_sections, export.concat_parts(data.lang, parts_formatted, categories, data.nocat, data.sort_key,
		data.lit, data.force_cat))
	return table.concat(text_sections)
end


--[=[
Make a given part into an affix of a specific type, and apply any relevant affix mappings. For example, if the desired
affix type is "suffix", this will (in general) add a hyphen onto the beginning of the term, alt, tr and ts components
of the part if not already present. The hyphen that's added is the "display hyphen" (see above) and may be
script-specific. In the case of East Asian scripts, the display hyphen is an empty string whereas the template hyphen
is the regular hyphen, meaning that any regular hyphen at the beginning of the part will be effectively removed. Note
that this also applies any language-specific affix mappings, so that e.g. if the language is Finnish and the user
specified [[-käs]] in the affix and didn't specify an .alt value, `part.term` will contain [[-kas]] and `part.alt` will
contain [[-käs]].

WARNING: This destructively modifies `part`.
]=]
local function make_part_affix(part, lang, sc, affix_type)
	local part_lang = part.lang or lang
	local part_sc = part.sc or sc

	local link_term, display_term = export.make_affix(part.term, part_lang, part_sc, affix_type, not part.alt)
	part.term = link_term
	-- When we don't specify do_affix_mapping, link and display terms (first and second retvals) are the same.
	part.alt = part.alt and export.make_affix(part.alt, part_lang, part_sc, affix_type) or display_term
	local Latn = require("Module:scripts").getByCode("Latn")
	part.tr = export.make_affix(part.tr, part_lang, Latn, affix_type)
	part.ts = export.make_affix(part.ts, part_lang, Latn, affix_type)
end


local function track_wrong_affix_type(template, part, lang, sc, expected_affix_type, part_name)
	if part then
		local affix_type = parse_term_for_affixes(part.term, part.lang or lang, part.sc or sc)
		if affix_type ~= expected_affix_type then
			part_name = part_name or expected_affix_type or "base"
			require("Module:debug/track") {
				template,
				template .. "/" .. part_name,
				template .. "/" .. part_name .. "/" .. (affix_type or "none"),
				template .. "/" .. part_name .. "/" .. (affix_type or "none") .. "/lang/" ..
					lang:getNonEtymologicalCode()
			}
		end
	end
end


local function insert_affix_category(categories, lang, pos, affix_type, part, sort_key, sort_base)
	if part.term then
		local part_lang = part.lang or lang
		local cat = pos .. " " .. affix_type .. "ed with " .. make_entry_name_no_links(part_lang, part.term) ..
			(part.id and " (" .. part.id .. ")" or "")
		if sort_key or sort_base then
			table.insert(categories, {cat=cat, sort_key=sort_key, sort_base=sort_base})
		else
			table.insert(categories, cat)
		end
	end
end


--[=[
Implementation of {{circumfix}}.

WARNING: This destructively modifies both `data` and `.prefix`, `.base` and `.suffix`.
]=]
function export.show_circumfix(data)
	local categories = {}
	data.pos = data.pos or default_pos

	data.pos = pluralize(data.pos)

	-- Hyphenate the affixes and apply any affix mappings.
	make_part_affix(data.prefix, data.lang, data.sc, "prefix")
	make_part_affix(data.suffix, data.lang, data.sc, "suffix")

	track_wrong_affix_type("circumfix", data.prefix, data.lang, data.sc, "prefix")
	track_wrong_affix_type("circumfix", data.base, data.lang, data.sc, nil)
	track_wrong_affix_type("circumfix", data.suffix, data.lang, data.sc, "suffix")

	-- Create circumfix term.
	local circumfix = nil

	if data.prefix.term and data.suffix.term then
		circumfix = data.prefix.term .. " " .. data.suffix.term
		data.prefix.alt = data.prefix.alt or data.prefix.term
		data.suffix.alt = data.suffix.alt or data.suffix.term
		data.prefix.term = circumfix
		data.suffix.term = circumfix
	end

	-- Make links out of all the parts.
	local parts_formatted = {}
	local sort_base
	if data.base.term then
		sort_base = make_entry_name_no_links(data.base.lang or data.lang, data.base.term)
	end

	table.insert(parts_formatted, export.link_term(data.prefix, data.lang, data.sc, data.sort_key, data.force_cat,
		data.nocat))
	table.insert(parts_formatted, export.link_term(data.base, data.lang, data.sc, data.sort_key, data.force_cat,
		data.nocat))
	table.insert(parts_formatted, export.link_term(data.suffix, data.lang, data.sc, data.sort_key, data.force_cat,
		data.nocat))

	-- Insert the categories.
	table.insert(categories, {cat=data.pos .. " circumfixed with " .. make_entry_name_no_links(data.prefix.lang or
		data.lang, circumfix), sort_key=data.sort_key, sort_base=sort_base})

	return export.concat_parts(data.lang, parts_formatted, categories, data.nocat, data.sort_key, data.lit,
		data.force_cat)
end


--[=[
Implementation of {{confix}}.

WARNING: This destructively modifies both `data` and `.prefix`, `.base` and `.suffix`.
]=]
function export.show_confix(data)
	data.pos = data.pos or default_pos

	data.pos = pluralize(data.pos)

	-- Hyphenate the affixes and apply any affix mappings.
	make_part_affix(data.prefix, data.lang, data.sc, "prefix")
	make_part_affix(data.suffix, data.lang, data.sc, "suffix")

	track_wrong_affix_type("confix", data.prefix, data.lang, data.sc, "prefix")
	track_wrong_affix_type("confix", data.base, data.lang, data.sc, nil)
	track_wrong_affix_type("confix", data.suffix, data.lang, data.sc, "suffix")

	-- Make links out of all the parts.
	local parts_formatted = {}
	local prefix_sort_base
	if data.base and data.base.term then
		prefix_sort_base = make_entry_name_no_links(data.base.lang or data.lang, data.base.term)
	elseif data.suffix.term then
		prefix_sort_base = make_entry_name_no_links(data.suffix.lang or data.lang, data.suffix.term)
	end

	-- Insert the categories and parts.
	local categories = {}

	table.insert(parts_formatted, export.link_term(data.prefix, data.lang, data.sc, data.sort_key, data.force_cat,
		data.nocat))
	insert_affix_category(categories, data.lang, data.pos, "prefix", data.prefix, data.sort_key, prefix_sort_base)

	if data.base then
		table.insert(parts_formatted, export.link_term(data.base, data.lang, data.sc, data.sort_key, data.force_cat,
			data.nocat))
	end

	table.insert(parts_formatted, export.link_term(data.suffix, data.lang, data.sc, data.sort_key, data.force_cat,
		data.nocat))
	-- FIXME, should we be specifying a sort base here?
	insert_affix_category(categories, data.lang, data.pos, "suffix", data.suffix)

	return export.concat_parts(data.lang, parts_formatted, categories, data.nocat, data.sort_key, data.lit,
		data.force_cat)
end


--[=[
Implementation of {{infix}}.

WARNING: This destructively modifies both `data` and `.base` and `.infix`.
]=]
function export.show_infix(data)
	local categories = {}
	data.pos = data.pos or default_pos

	data.pos = pluralize(data.pos)

	-- Hyphenate the affixes and apply any affix mappings.
	make_part_affix(data.infix, data.lang, data.sc, "infix")

	track_wrong_affix_type("infix", data.base, data.lang, data.sc, nil)
	track_wrong_affix_type("infix", data.infix, data.lang, data.sc, "infix")

	-- Make links out of all the parts.
	local parts_formatted = {}

	table.insert(parts_formatted, export.link_term(data.base, data.lang, data.sc, data.sort_key, data.force_cat,
		data.nocat))
	table.insert(parts_formatted, export.link_term(data.infix, data.lang, data.sc, data.sort_key, data.force_cat,
		data.nocat))

	-- Insert the categories.
	-- FIXME, should we be specifying a sort base here?
	insert_affix_category(categories, data.lang, data.pos, "infix", data.infix)

	return export.concat_parts(data.lang, parts_formatted, categories, data.nocat, data.sort_key, data.lit,
		data.force_cat)
end


--[=[
Implementation of {{prefix}}.

WARNING: This destructively modifies both `data` and the structures within `.prefixes`, as well as `.base`.
]=]
function export.show_prefix(data)
	data.pos = data.pos or default_pos

	data.pos = pluralize(data.pos)

	-- Hyphenate the affixes and apply any affix mappings.
	for i, prefix in ipairs(data.prefixes) do
		make_part_affix(prefix, data.lang, data.sc, "prefix")
	end

	for i, prefix in ipairs(data.prefixes) do
		track_wrong_affix_type("prefix", prefix, data.lang, data.sc, "prefix")
	end

	track_wrong_affix_type("prefix", data.base, data.lang, data.sc, nil)

	-- Make links out of all the parts.
	local parts_formatted = {}
	local first_sort_base = nil
	local categories = {}

	if data.prefixes[2] then
		first_sort_base = make_entry_name_no_links(data.prefixes[2].lang or data.lang, data.prefixes[2].term)
	elseif data.base then
		first_sort_base = make_entry_name_no_links(data.base.lang or data.lang, data.base.term)
	end

	for i, prefix in ipairs(data.prefixes) do
		table.insert(parts_formatted, export.link_term(prefix, data.lang, data.sc, data.sort_key, data.force_cat,
			data.nocat))
		insert_affix_category(categories, data.lang, data.pos, "prefix", prefix, data.sort_key, i == 1 and
			first_sort_base or nil)
	end

	if data.base then
		table.insert(parts_formatted, export.link_term(data.base, data.lang, data.sc, data.sort_key, data.force_cat,
			data.nocat))
	else
		table.insert(parts_formatted, "")
	end

	return export.concat_parts(data.lang, parts_formatted, categories, data.nocat, data.sort_key, data.lit,
		data.force_cat)
end


--[=[
Implementation of {{suffix}}.

WARNING: This destructively modifies both `data` and the structures within `.suffixes`, as well as `.base`.
]=]
function export.show_suffix(data)
	local categories = {}
	data.pos = data.pos or default_pos

	data.pos = pluralize(data.pos)

	-- Hyphenate the affixes and apply any affix mappings.
	for i, suffix in ipairs(data.suffixes) do
		make_part_affix(suffix, data.lang, data.sc, "suffix")
	end

	track_wrong_affix_type("suffix", data.base, data.lang, data.sc, nil)

	for i, suffix in ipairs(data.suffixes) do
		track_wrong_affix_type("suffix", suffix, data.lang, data.sc, "suffix")
	end

	-- Make links out of all the parts.
	local parts_formatted = {}

	if data.base then
		table.insert(parts_formatted, export.link_term(data.base, data.lang, data.sc, data.sort_key, data.force_cat,
			data.nocat))
	else
		table.insert(parts_formatted, "")
	end

	for i, suffix in ipairs(data.suffixes) do
		table.insert(parts_formatted, export.link_term(suffix, data.lang, data.sc, data.sort_key, data.force_cat,
			data.nocat))
	end

	-- Insert the categories.
	for i, suffix in ipairs(data.suffixes) do
		if suffix.term then
			-- FIXME, should we be specifying a sort base here?
			insert_affix_category(categories, data.lang, data.pos, "suffix", suffix)
		end

		if suffix.pos and rfind(suffix.pos, "patronym") then
			table.insert(categories, "patronymics")
		end
	end

	return export.concat_parts(data.lang, parts_formatted, categories, data.nocat, data.sort_key, data.lit,
		data.force_cat)
end

return export