ClaudeBot: Strip wiki markup before splitting: whole-list links were rejoining into one redlink

2026-07-21T11:59:21Z

Strip wiki markup before splitting: whole-list links were rejoining into one redlink

← Older revision		Revision as of 07:59, 21 July 2026
Line 28:		Line 28:

	-- Strip any stored wiki markup, keeping the display text of piped links.		-- Strip any stored wiki markup, keeping the display text of piped links.
			--
			-- This must run over the WHOLE field before it is split on commas. Some legacy
			-- rows wrap the entire list in a single link -- [[Sw Canis Majoris, Binary
			-- stars, Lacy, motion]] -- so splitting first would leave unbalanced fragments
			-- ("[[Sw Canis Majoris" ... "motion]]") that rejoin into one giant redlink.
	local function plainText(s)		local function plainText(s)
	s = s:gsub('%[%[[^%]\|]\|([^%]])%]%]', '%1') -- [[Target\|shown]] -> shown		s = s:gsub('%[%[[^%]\|]\|([^%]])%]%]', '%1') -- [[Target\|shown]] -> shown
	s = s:gsub('%[%[([^%]]*)%]%]', '%1') -- [[Target]] -> Target		s = s:gsub('%[%[([^%]]*)%]%]', '%1') -- [[Target]] -> Target
	s = s:gsub("'''", ''):gsub("''", '') -- stray bold/italic		s = s:gsub("'''", ''):gsub("''", '') -- stray bold/italic
			s = s:gsub('%[%[', ''):gsub('%]%]', '') -- any unbalanced leftovers
	return trim(s)		return trim(s)
	end		end
Line 59:		Line 65:
	local ok, data = pcall(mw.loadData, 'Module:Keywords/data')		local ok, data = pcall(mw.loadData, 'Module:Keywords/data')
	local map = (ok and type(data) == 'table' and data.map) or {}		local map = (ok and type(data) == 'table' and data.map) or {}

			-- Remove markup first, then split -- see plainText above for why the order
			-- matters.
			raw = plainText(raw)

	local out, seen = {}, {}		local out, seen = {}, {}
	for chunk in mw.text.gsplit(raw, '%s[,;]%s') do		for chunk in mw.text.gsplit(raw, '%s[,;]%s') do
	local display = ~~plainText~~(chunk)		local display = trim(chunk)
	if display ~= '' then		if display ~= '' then
	local k = key(display)		local k = key(display)

ClaudeBot: Render infobox keywords: link only curated concepts, no redlinks

2026-07-21T11:57:01Z

Render infobox keywords: link only curated concepts, no redlinks

New page

-- Module:Keywords
--
-- Renders the |keywords= field of the infobox templates as a comma-separated
-- list in which recognised concepts link to their article and everything else
-- stays plain text.
--
-- WHY THIS EXISTS
-- The keyword fields were populated from a legacy database in which the terms
-- were already wrapped in [[...]]. Most of those targets have never existed, so
-- roughly 8,900 keyword links across the paper pages rendered red. Worse, a
-- number of the targets that DO exist are papers rather than concepts -- the
-- wiki has papers titled "Gravity", "Time", "Ether" and "Cosmology" -- so the
-- keyword "gravity" pointed at a single paper instead of the subject.
--
-- The fix is to ignore the markup stored on the page and decide the link here,
-- against the curated allowlist in [[Module:Keywords/data]]. A term links only
-- if it is in that list; otherwise it renders as plain text. Adding a page to
-- the wiki is therefore not enough to create a link -- the mapping has to be
-- added deliberately, which is what keeps bad targets out.
--
-- Usage: {{#invoke:Keywords|render|{{{keywords|}}}}}

local p = {}

local function trim(s)
return (s:gsub('^%s*(.-)%s*$', '%1'))
end

-- Strip any stored wiki markup, keeping the display text of piped links.
local function plainText(s)
s = s:gsub('%[%[[^%]|]*|([^%]]*)%]%]', '%1') -- [[Target|shown]] -> shown
s = s:gsub('%[%[([^%]]*)%]%]', '%1') -- [[Target]] -> Target
s = s:gsub("'''", ''):gsub("''", '') -- stray bold/italic
return trim(s)
end

-- Lookup key: lower case, collapsed whitespace, no trailing punctuation.
local function key(s)
s = trim(s):lower()
s = s:gsub('%s+', ' ')
s = s:gsub('[%.%s]+$', '')
return s
end

function p.render(frame)
local raw = frame.args[1]
if raw == nil or trim(raw) == '' then
local parent = frame:getParent()
raw = parent and parent.args and parent.args.keywords or ''
end
raw = trim(raw or '')
if raw == '' then
return ''
end

-- Load the map defensively: if the data page is missing or broken the
-- keywords must still render, just without links. This template is
-- transcluded on thousands of pages and must not fail loudly.
local ok, data = pcall(mw.loadData, 'Module:Keywords/data')
local map = (ok and type(data) == 'table' and data.map) or {}

local out, seen = {}, {}
for chunk in mw.text.gsplit(raw, '%s*[,;]%s*') do
local display = plainText(chunk)
if display ~= '' then
local k = key(display)
-- Drop duplicates left over from the legacy data (e.g. a term
-- appearing both linked and unlinked in the same field).
if not seen[k] then
seen[k] = true
local target = map[k]
if target and target ~= '' then
if target == display then
table.insert(out, '[[' .. target .. ']]')
else
table.insert(out, '[[' .. target .. '|' .. display .. ']]')
end
else
table.insert(out, display)
end
end
end
end

return table.concat(out, ', ')
end

return p

Module:Keywords - Revision history

ClaudeBot: Strip wiki markup before splitting: whole-list links were rejoining into one redlink

ClaudeBot: Render infobox keywords: link only curated concepts, no redlinks