Modul:data consistency check

local export = {}

local m_language_data = require("Modul:languages/alldata")
local m_language_codes = require('Modul:languages/code to canonical name')
local m_language_canonical_names = require('Modul:languages/canonical names')
local m_etym_language_data = require("Modul:etymology languages/data")
local m_family_data = require('Modul:families/data')
local m_script_data = require('Modul:scripts/data')

local m_table = require("Modul:table")
local Array = require("Modul:array")

local messages

local function discrepancy(modname, ...)
	messages[modname]:insert(string.format(...))
end

local all_codes = {}

local language_names = {}
local family_names = {}
local script_names = {}

local nonempty_families = {}
local allowed_empty_families = {tbq = true}
local nonempty_scripts = {}
	
local function link(name)
	if not name then
		return "???"
	elseif name:find("[Ll]anguage$") then
		return "[[:Category:" .. name .. "|" .. name .. "]]"
	else
		return "[[:Category:" .. name .. " language|" .. name .. " language]]"
	end
end
	
local function link_script(name)
	if not name then
		return "???"
	elseif name:find("[Cc]ode$") or name:find("[Ss]emaphore$") then
		return "[[:Category:" .. name:gsub("^%l", string.upper) .. "|" .. name .. "]]"
	else
		return "[[:Category:" .. name .. " script|" .. name .. " script]]"
	end
end

local function invalid_keys_message(modname, code, data, invalid_keys, is_script)
	local plural = #invalid_keys ~= 1
	discrepancy(modname, "The data key%s %s for %s (<code>%s</code>) %s invalid.",
		plural and "s" or "",
		invalid_keys
			:map(
				function(key)
					return '<code>' .. key .. '</code>'
				end)
			:concat(", "),
		(is_script and link_script or link)(data.canonicalName or data[1]),
		code,
		plural and "are" or "is")
end

local function check_data_keys(valid_keys, is_script)
	valid_keys = Array(valid_keys):to_set()
	
	return function (modname, code, data)
		local invalid_keys
		for k in pairs(data) do
			if not valid_keys[k] then
				invalid_keys = invalid_keys or Array()
				invalid_keys:insert(k)
			end
		end
		if invalid_keys then
			invalid_keys_message(modname, code, data, invalid_keys, is_script)
		end
	end
end

-- Modification of isArray in [[Module:table]].
local function find_gap(t)
	local i = 0
	for _ in pairs(t) do
		i = i + 1
		if t[i] == nil then
			return i
		end
	end
end

local function check_array(modname, code, data, array_name)
	local gap = find_gap(data[array_name])
	if gap then
		discrepancy(modname, "The %s array in the data table for %s (<code>%s</code>) has a gap at index %d.",
			array_name, data.canonicalName or data[1], code, gap)
	end
end

local function check_other_names_or_aliases(modname, code, canonical_name, data, data_key, allow_nested)
	local array = data[data_key]
	if not array then
		return
	end
	check_array(modname, code, data, data_key)

	local names = {}
	local function check_other_name(other_name)
		if other_name == canonical_name then
			discrepancy(modname,
				"%s, the canonical name for <code>%s</code>, is repeated in the table of <code>%s</code>.",
				canonical_name, code, data_key)
		end
		if names[other_name] then
			discrepancy(modname,
				"The name %s is found twice or more in the list of <code>%s</code> for %s (<code>%s</code>).",
				other_name, data_key, canonical_name, code)
		end
		names[other_name] = true
	end

	for _, other_name in ipairs(array) do
		if type(other_name) == "table" then
			if not allow_nested then
				discrepancy(modname,
					"A nested table is found in the list of <code>%s</code> for %s (<code>%s</code>), but isn't allowed.",
					data_key, canonical_name, code)
			else
				for _, on in ipairs(other_name) do
					check_other_name(on)
				end
			end
		else
			check_other_name(other_name)
		end
	end
end

local function check_other_names_aliases_varieties(modname, code, canonical_name, data)
	if data.otherNames then
		check_other_names_or_aliases(modname, code, canonical_name, data, "otherNames")
	end
	if data.aliases then
		check_other_names_or_aliases(modname, code, canonical_name, data, "aliases")
	end
	if data.varieties then
		check_other_names_or_aliases(modname, code, canonical_name, data, "varieties", true)
	end
end

local get_codepoint = mw.ustring.codepoint
local function validate_pattern(pattern, modname, code, data, standardChars)
	if type(pattern) ~= "string" then
		discrepancy(modname, '"%s", the %spattern for %s (<code>%s</code>), is not a string.',
			pattern, standardChars and 'standard character ' or '', code, data.canonicalName)
	end
	local ranges
	for lower, higher in mw.ustring.gmatch(pattern, "(.)%-(.)") do
		if get_codepoint(lower) >= get_codepoint(higher) then
			ranges = ranges or Array()
			table.insert(ranges, { lower, higher })
		end
	end
	if ranges and ranges[1] then
		local plural = #ranges ~= 1 and "s" or ""
		discrepancy(modname, '%s (<code>%s</code>) specifies an invalid pattern ' ..
			'for %scharacter detection: <code>"%s"</code>. The first codepoint%s ' ..
			'in the range%s %s %s must be less than the second.',
			link(data.canonicalName), code, standardChars and 'standard ' or '', pattern, plural, plural,
			ranges
				:map(
					function(range)
						return range[1] .. "-" .. range[2] .. (" (U+%X, U+%X)")
							:format(get_codepoint(range[1]), get_codepoint(range[2]))
					end)
				:concat(", "),
			#ranges ~= 1 and "are" or "is")
	end
	if not pcall(mw.ustring.find, "", "[" .. pattern .. "]") then
		discrepancy(modname, '%s (<code>%s</code>) specifies an invalid pattern for ' ..
			(standardChars and 'standard' or '') .. ' character detection: <code>"%s"</code>',
			link(data.canonical_name), code, pattern)
	end
end

local function check_entry_name_or_sortkey(modname, code, data, replacements_name)
	local replacements = data[replacements_name]
	if type(replacements) == "string" then
		if replacements_name ~= "sort_key" then
			discrepancy(modname, "The %s field in the data table for %s (<code>%s</code>) must be a table.",
				replacements_name, data.canonicalName, code)
		end
		return
	end
	
	if (replacements.from ~= nil) ~= (replacements.to ~= nil) then
		discrepancy(modname,
			"The <code>from</code> and <code>to</code> arrays in the <code>%s</code> table for %s (<code>%s</code>) are not both defined or both undefined.",
			replacements_name, data.canonicalName, code)
	elseif replacements.from then
		for _, key in ipairs { "from", "to" } do
			local gap = find_gap(replacements[key])
			if gap then
				discrepancy(modname,
					"The %s array in the %s table for %s (<code>%s</code>) has a gap at index %d.",
					key, replacements_name, data.canonicalName, code, gap)
			end
		end
	end
	
	if replacements.remove_diacritics and type(replacements.remove_diacritics) ~= "string" then
		discrepancy(modname,
			"The <code>remove_diacritics</code> field in the <code>%s</code> table for %s (<code>%s</code>) table must be a string.",
			replacements_name, data.canonicalName, code)
	end
	
	if replacements.from and replacements.to
			and m_table.length(replacements.to) > m_table.length(replacements.from) then
		discrepancy(modname,
			"The <code>from</code> array in the <code>%s</code> table for %s (<code>%s</code>) must be shorter or the same length as the <code>to</code> array.",
			replacements_name, data.canonicalName, code)
	end
end

local function has_regular_language_child(parent_code)
	for code, data in pairs(m_language_data) do
		local ancestors = data.ancestors
		if ancestors then
			for _, ancestor in pairs(ancestors) do
				if ancestor == parent_code then
					return true
				end
			end
		end
	end
	return false
end

local function check_ancestors(modname, code, data, ancestors, is_etymology_language)
	check_array(modname, code, data, "ancestors")
	
	local canonical_name = data[1] or data.canonicalName
	if is_etymology_language then
		if not has_regular_language_child(code) then
			discrepancy(modname,
				"The etymology language %s (<code>%s</code>) has an <code>ancestors</code> field, "
				.. "but no regular languages list it as an ancestor.",
				link(canonical_name), code)
		end
	end
	
	for _, ancestor_code in ipairs(ancestors) do
		if not (m_language_data[ancestor_code] or m_etym_language_data[ancestor_code]) then
			discrepancy(modname,
				"%s (<code>%s</code>) lists an invalid language code <code>%s</code> as ancestor.",
				link(canonical_name), code, ancestor_code)
		end
	end
end

local function check_languages()
	local check_language_data_keys = check_data_keys{
		1, 2, 3, -- canonical name, wikidata item, family
		"entry_name", "sort_key", "otherNames", "aliases", "varieties",
		"type", "scripts", "ancestors",
		"wikimedia_codes", "wikipedia_article", "standardChars",
		"translit_module", "override_translit", "link_tr",
	}
	
	local function check_language(modname, code, data)
		local canonical_name, wikidata_item, lang_type = data[1], data[2], data.type
		
		check_language_data_keys(modname, code, data)
		
		if all_codes[code] then
			discrepancy(modname, "Code <code>%s</code> is not unique; it is also defined in [[Module:%s]].", code, all_codes[code])
		else
			if not m_language_codes[code] then
				discrepancy("languages/code to canonical name", "The code <code>%s</code> (%s) is missing.", code, canonical_name)
			end
			all_codes[code] = modname
		end
		
		if not canonical_name then
			discrepancy(modname, "Code <code>%s</code> has no canonical name specified.", code)
		elseif language_names[canonical_name] then
			discrepancy(modname,
				"%s (<code>%s</code>) has a canonical name that is not unique; it is also used by the code <code>%s</code>.",
				link(canonical_name), code, language_names[canonical_name])
		else
			if not m_language_canonical_names[canonical_name] then
				discrepancy("languages/canonical names", "The canonical name %s (<code>%s</code>) is missing.", canonical_name, code)
			end
			language_names[canonical_name] = code
		end
		
		if wikidata_item then
			if not wikidata_item:match '^Q%d+$' then
				discrepancy(modname,
					"%s (<code>%s</code>) has a Wikidata item with an invalid form: <code>%s</code>.",
					canonical_name, code, wikidata_item)
			end
		end

		check_other_names_aliases_varieties(modname, code, canonical_name, data)
		
		if lang_type and not (lang_type == "regular" or lang_type == "reconstructed" or lang_type == "appendix-constructed") then
			discrepancy(modname, "%s (<code>%s</code>) is of an invalid type <code>%s</code>.", link(canonical_name), code, data.type)
		end
		
		if data.scripts then
			check_array(modname, code, data, "scripts")
			if not data.scripts[1] then
				discrepancy(modname, "%s (<code>%s</code>) has no scripts listed.", link(canonical_name), code)
			else
				for _, sccode in ipairs(data.scripts) do
					if not m_script_data[sccode] then
						discrepancy(modname,
							"%s (<code>%s</code>) lists an invalid script code <code>%s</code>.",
							link(canonical_name), code, sccode)
					end
		
					nonempty_scripts[sccode] = true
				end
			end
		end
		
		if data.ancestors then
			check_ancestors(modname, code, data, data.ancestors, false)
		end
		
		if data[3] then
			local family = data[3]
			if not m_family_data[family] then
				discrepancy(modname,
					"%s (<code>%s</code>) has an invalid family code <code>%s</code>.",
					link(canonical_name), code, family)
			end
			
			nonempty_families[family] = true
		end
		
		if data.sort_key then
			check_entry_name_or_sortkey(modname, code, data, "sort_key")
		end
		
		if data.entry_name then
			check_entry_name_or_sortkey(modname, code, data, "entry_name")
		end

		if data.standardChars then
			validate_pattern(data.standardChars, modname, code, data, true)
		end
		
		if data.override_translit and not data.translit_module then
			discrepancy(modname,
				"%s (<code>%s</code>) has <code>override_translit</code> set, but no transliteration module",
				link(canonical_name), code)
		end
	end
	
	-- Check two-letter codes
	local modname = "languages/data2"
	local data2 = require("Module:" .. modname)
	
	for code, data in pairs(data2) do
		if not code:find("^[a-z][a-z]$") then
			discrepancy(modname, '%s (<code>%s</code>) does not have a two-letter code.', link(data.canonicalName), code)
		end
		
		check_language(modname, code, data)
	end
	
	-- Check three-letter codes
	for i = string.byte('a'), string.byte('z') do
		local letter = string.char(i)
		local modname = "languages/data3/" .. letter
		local data3 = require("Module:" .. modname)
		local code_pattern = "^" .. letter .. "[a-z][a-z]$"
		
		for code, data in pairs(data3) do
			if not code:find(code_pattern) then
				discrepancy(modname,
					'%s (<code>%s</code>) does not have a three-letter code starting with "<code>%s</code>".',
					link(data.canonicalName), code, letter)
			end
			
			check_language(modname, code, data)
		end
	end
	
	-- Check exceptional codes
	modname = "languages/datax"
	local datax = require("Module:" .. modname)
	
	for code, data in pairs(datax) do
		if code:find("^[a-z][a-z][a-z]?$") then
			discrepancy(modname, '%s (<code>%s</code>) has a two- or three-letter code.', link(data.canonicalName), code)
		end
		
		check_language(modname, code, data)
	end
	
	-- These checks must be done while all_codes only contains language codes:
	-- that is, after language data modules have been processed, but before
	-- etymology languages, families, and scripts have.
	local function check_code_and_name(modname, code, canonical_name)
		if not all_codes[code] then
			if not language_names[canonical_name] then
				discrepancy(modname,
					"The code <code>%s</code> and the canonical name %s should be removed; they are not found in a submodule of [[Module:languages]].",
					code, canonical_name)
			else
				discrepancy(modname,
					"<code>%s</code>, the code for the canonical name %s, is wrong; it should be <code>%s</code>.",
					code, canonical_name, language_names[canonical_name])
			end
		elseif not language_names[canonical_name] then
			local data_table = require("Module:" .. all_codes[code])[code]
			discrepancy(modname,
				"%s, the canonical name for the code <code>%s</code>, is wrong; it should be %s.",
				canonical_name, code, data_table[1] or data_table.canonicalName)
		end
	end
	
	for code, canonical_name in pairs(m_language_codes) do
		check_code_and_name("languages/code to canonical name", code, canonical_name)
	end
	
	for canonical_name, code in pairs(m_language_canonical_names) do
		check_code_and_name("languages/canonical names", code, canonical_name)
	end		
end

local function check_etym_languages()
	local modname = "etymology languages/data"
	
	local check_etymology_language_data_keys = check_data_keys{
		"canonicalName", "otherNames", "aliases", "varieties", "parent",
		"wikipedia_article", "wikidata_item", "ancestors"
	}
	
	local function link(name)
		if not name then
			return "???"
		elseif name:find("[Ll]anguage$") then
			return name
		else
			return name .. " language"
		end
	end
	
	for code, data in pairs(m_etym_language_data) do
		local canonical_name, parent, ancestors =
			data.canonicalName, data.parent, data.ancestors
		check_etymology_language_data_keys(modname, code, data)
		
		if all_codes[code] then
			discrepancy(modname, "Code <code>%s</code> is not unique; it is also defined in [[Module:%s]].", code, all_codes[code])
		else
			all_codes[code] = modname
		end
		
		if not canonical_name then
			discrepancy(modname, "Code <code>%s</code> has no canonical name specified.", code)
		elseif language_names[canonical_name] then
			--[=[
			discrepancy(modname,
				"%s (<code>%s</code>) has a canonical name that is not unique; it is also used by the code <code>%s</code>.",
				link(data.names[1]), code, language_names[data.names[1]])
			--]=]
		else
			language_names[canonical_name] = code
		end
		
		check_other_names_aliases_varieties(modname, code, canonical_name, data)
		
		if parent then
			if type(parent) ~= "string" then
				discrepancy(modname,
					"Etymology-only %s (<code>%s</code>) has a parent language or family code that is %s rather than a string.",
					link(canonical_name), code, parent == nil and "nil" or "a " .. type(parent))
			elseif not (m_language_data[parent] or m_family_data[parent] or m_etym_language_data[parent]) then
				discrepancy(modname,
					"Etymology-only %s (<code>%s</code>) has invalid parent language or family code <code>%s</code>.",
					link(canonical_name), code, parent)
			end
			
			nonempty_families[parent] = true
		else
			discrepancy(modname,
				"Etymology-only %s (<code>%s</code>) has no parent language or family code.",
				link(canonical_name), code)
		end
		
		if ancestors then
			check_ancestors(modname, code, data, ancestors, true)
		end
	end

	local checked = {}
	for code, data in pairs(m_etym_language_data) do
		local stack = {}

		while data do
			if checked[data] then
				break	
			end
			if stack[data] then
				discrepancy(modname, "%s (<code>%s</code>) has a cyclic parental relationship to %s (<code>%s</code>)",
					link(data[1] or data.canonicalName), code,
					link(m_etym_language_data[data.parent].canonicalName), data.parent
				)
				break
			end
			stack[data] = true
			code, data = data.parent, data.parent and m_etym_language_data[data.parent]
		end
		
		for data in pairs(stack) do
			checked[data] = true	
		end
	end
end

local function check_families()
	local modname = "families/data"
	
	local check_family_data_keys = check_data_keys{
		"canonicalName", "otherNames", "aliases", "varieties", "family",
		"protoLanguage", "wikidata_item"
	}

	local function link(name)
		if not name then
			return "???"
		elseif name:find("[Ll]anguages$") then
			return "[[:Category:" .. name .. "|" .. name .. " family]]"
		else
			return "[[:Category:" .. name .. " languages|" .. name .. " family]]"
		end
	end
	
	for code, data in pairs(m_family_data) do
		check_family_data_keys(modname, code, data)
		
		if all_codes[code] then
			discrepancy(modname, "Code <code>%s</code> is not unique; it is also defined in [[Module:%s]].", code, all_codes[code])
		else
			all_codes[code] = modname
		end
		
		if not data.canonicalName then
			discrepancy(modname, "<code>%s</code> has no canonical name specified.", code)
		elseif family_names[data.canonicalName] then
			discrepancy(modname,
				"%s (<code>%s</code>) has a canonical name that is not unique; it is also used by the code <code>%s</code>.",
				link(data.canonicalName), code, family_names[data.canonicalName])
		else
			family_names[data.canonicalName] = code
		end
		
		check_other_names_aliases_varieties(modname, code, data.canonicalName, data)
		
		if data.family then
			if data.family == code and code ~= "qfa-not" then
				discrepancy(modname,
					"%s (<code>%s</code>) has itself as its family.",
					link(data.canonicalName), code)
			elseif not m_family_data[data.family] then
				discrepancy(modname,
					"%s (<code>%s</code>) has an invalid parent family code <code>%s</code>.",
					link(data.canonicalName), code, data.family)
			end
			
			nonempty_families[data.family] = true
		end
	end
	
	for code, data in pairs(m_family_data) do
		if not (nonempty_families[code] or allowed_empty_families[code]) then
			discrepancy(modname, "%s (<code>%s</code>) has no child families or languages.", link(data.canonicalName), code)
		end
	end

	local checked = { ['qfa-not'] = true }
	for code, data in pairs(m_family_data) do
		local stack = {}

		while data do
			if checked[code] then
				break	
			end
			if stack[code] then
				discrepancy(modname, "%s (<code>%s</code>) has a cyclic parental relationship to %s (<code>%s</code>)",
					link(data[1] or data.canonicalName), code,
					link(m_family_data[data[3]].canonicalName), data[3]
				)
				break
			end
			stack[code] = true
			code, data = data.family, m_family_data[data[3]]
		end
		
		for code in pairs(stack) do
			checked[code] = true	
		end
	end
end

local function check_scripts()
	local modname = "scripts/data"
	
	local check_script_data_keys = check_data_keys({
		"canonicalName", "otherNames", "aliases", "varieties", "parent",
		"systems", "wikipedia_article", "characters", "direction",
		"character_category",
	}, true)
	
	local m_script_codes = require('Module:scripts/code to canonical name')
	local m_script_canonical_names = require('Module:scripts/by name')
	
	for code, data in pairs(m_script_data) do
		local canonical_name = data.canonicalName
		if not m_script_codes[code] and #code == 4 then
			discrepancy('scripts/code to canonical name', '<code>%s</code> (%s) is missing', code, canonical_name)
		end
		
		check_script_data_keys(modname, code, data)
		
		if not canonical_name then
			discrepancy(modname, "Code <code>%s</code> has no canonical name specified.", code)
		elseif script_names[canonical_name] then
			--[=[
			discrepancy(modname,
				"%s (<code>%s</code>) has a canonical name that is not unique; it is also used by the code <code>%s</code>.",
				link_script(data.names[1]), code, script_names[data.names[1]])
			--]=]
		else
			if not m_script_canonical_names[canonical_name] and #code == 4 then
				discrepancy('scripts/by name', '%s (<code>%s</code>) is missing', canonical_name, code)
			end
			script_names[canonical_name] = code
		end
		
		check_other_names_aliases_varieties(modname, code, canonical_name, data)
		
		if not nonempty_scripts[code] then
			discrepancy(modname,
				"%s (<code>%s</code>) is not used by any language%s.",
				link_script(canonical_name), code, data.characters and ""
					or " and has no characters listed for auto-detection")
		--[[
		elseif not data.characters then
			discrepancy(modname, "%s (<code>%s</code>) has no characters listed for auto-detection.", link_script(canonical_name), code)
		--]]
		end

		if data.characters then
			validate_pattern(data.characters, modname, code, data, false)
		end
	end
end

-- Warning: cannot be called twice in the same module invocation because
-- some module-global variables are not reset between calls.
function export.do_checks()
	messages = setmetatable({}, {
		__index = function (self, k)
			local val = Array()
			self[k] = val
			return val
		end
	})
	
	check_languages()
	check_etym_languages()

	-- families and scripts must be checked AFTER languages; languages checks fill out
	-- the nonempty_families and nonempty_scripts tables, used for testing if a family/script
	-- is ever used in the data
	check_families()
	check_scripts()
	
	setmetatable(messages, nil)
	
	local function find_code(message)
		return string.match(message, "<code>([^<]+)</code>")
	end
	
	find_code = require("Module:fun").memoize(find_code)
	
	local function comp(message1, message2)
		local code1, code2 = find_code(message1), find_code(message2)
		if code1 and code2 then
			return code1 < code2
		else
			return message1 < message2
		end
	end
	
	for modname, msglist in pairs(messages) do
		msglist:sort(comp)
	end
	
	local ret = messages
	messages = nil
	return ret
end

function export.format_message(modname, msglist)
	return '===[[Module:' .. modname .. ']]==='
		.. msglist
			:map(
				function(msg)
					return "\n* " .. msg
				end)
			:concat()
end

function export.check_modules(...)
	local ret = Array()
	local messages = export.do_checks()
	for _, module in ipairs {...} do
		local msglist = messages[module]
		if msglist then
			ret:insert(export.format_message(module, msglist))
		end
	end
	return ret:concat("\n")
end

function export.check_modules_t(frame)
	local args = m_table.shallowcopy(frame.args)
	return export.check_modules(unpack(args))
end

function export.perform(frame)
	local messages = export.do_checks()
	
	-- Format the messages
	local ret = Array()
	for modname, msglist in m_table.sortedPairs(messages) do
		ret:insert(export.format_message(modname, msglist))
	end
	
	-- Are there any messages?
	if i == 1 then
		return '<b class="success">Glory to Arstotzka.</b>'
	else
		ret:insert(1, '<b class="warning">Discrepancies detected:</b>')
		
		return ret:concat('\n')
	end
end

return export