မော်ဂျူး:template parser
Documentation for this module may be created at မော်ဂျူး:template parser/doc
--[[
NOTE: This module works by using recursive backtracking to build a node tree, which can then be traversed as necessary.
Because it is called by a number of high-use modules, it has been optimised for speed using a profiler, since it is used to scrape data from large numbers of pages very quickly. To that end, it rolls some of its own methods in cases where this is faster than using a function from one of the standard libraries. Please DO NOT "simplify" the code by removing these, since you are almost guaranteed to slow things down, which could seriously impact performance on pages which call this module hundreds or thousands of times.
It has also been designed to emulate the native parser's behaviour as much as possible, which in some cases means replicating bugs or unintuitive behaviours in that code; these should not be "fixed", since it is important that the outputs are the same. Most of these originate from deficient regular expressions, which can't be used here, so the bugs have to be manually reintroduced as special cases (e.g. onlyinclude tags being case-sensitive and whitespace intolerant, unlike all other tags). If any of these are fixed, this module should also be updated accordingly.
]]
local require = require
local require_when_needed = require("Module:require when needed")
local m_parser = require("Module:parser")
local m_str_utils = require("Module:string utilities")
local mw = mw
local mw_title = mw.title
local string = string
local table = table
local build_template -- defined as export.buildTemplate below
local concat = table.concat
local decode_entities = m_str_utils.decode_entities
local encode_entities = m_str_utils.encode_entities
local encode_uri = mw.uri.encode
local find = string.find
local format = string.format
local getmetatable = getmetatable
local gsub = string.gsub
local html_create = mw.html.create
local insert = table.insert
local is_node = m_parser.is_node
local is_valid_title = require_when_needed("Module:pages", "is_valid_title")
local load_data = mw.loadData
local lower = string.lower
local make_title = mw_title.makeTitle -- unconditionally adds the specified namespace prefix
local match = string.match
local new_title = mw_title.new -- specified namespace prefix is only added if the input doesn't contain one
local next = next
local pairs = pairs
local parse -- defined as export.parse below
local parse_template_name -- defined below
local pcall = pcall
local php_trim = m_str_utils.php_trim
local rep = string.rep
local replacement_escape = m_str_utils.replacement_escape
local reverse = string.reverse
local scribunto_param_key = m_str_utils.scribunto_param_key
local select = select
local sorted_pairs = require_when_needed("Module:table", "sortedPairs")
local split = m_str_utils.split
local sub = string.sub
local table_len = require_when_needed("Module:table", "length")
local title_equals = mw_title.equals
local tostring = m_parser.tostring
local type = type
local type_or_class = m_parser.type_or_class
local umatch = mw.ustring.match
local uupper = m_str_utils.upper
local data = load_data("Module:template parser/data")
local frame = mw.getCurrentFrame()
local magic_words = load_data("Module:data/magic words")
local parser_extension_tags = load_data("Module:data/parser extension tags")
local valid_attribute_name = data.valid_attribute_name
local Parser, Node = m_parser.new()
local export = {}
export.type_or_class = type_or_class
local function preprocess(text, frame_args)
return is_node(text) and text:preprocess(frame_args) or text
end
export.preprocess = preprocess
------------------------------------------------------------------------------------
--
-- Nodes
--
------------------------------------------------------------------------------------
Node.keys_to_remove = {"handler", "head", "pattern", "route", "step"}
function Node:preprocess(frame_args)
local output = {}
for i = 1, #self do
output[i] = preprocess(self[i], frame_args)
end
return concat(output)
end
local Wikitext = Node:new_class("wikitext")
-- force_node ensures the output will always be a node.
function Wikitext:new(this, force_node)
if type(this) ~= "table" then
return force_node and Node.new(self, {this}) or this
elseif #this == 1 then
local this1 = this[1]
return force_node and not is_node(this1) and Node.new(self, this) or this1
end
local success, str = pcall(concat, this)
if success then
return force_node and Node.new(self, {str}) or str
end
return Node.new(self, this)
end
-- First value is the parameter name.
-- Second value is the parameter's default value.
-- Any additional values are ignored: e.g. "{{{a|b|c}}}" is parameter "a" with default value "b" (*not* "b|c").
local Parameter = Node:new_class("parameter")
function Parameter:new(this)
local this2 = this[2]
if type_or_class(this2) == "argument" then
insert(this2, 2, "=")
this2 = Wikitext:new(this2)
end
return Node.new(self, {this[1], this2})
end
function Parameter:__tostring()
local output = {}
for i = 1, #self do
output[i] = tostring(self[i])
end
return "{{{" .. concat(output, "|") .. "}}}"
end
function Parameter:next(i)
i = i + 1
if i <= 2 then
return self[i], i
end
end
function Parameter:get_name(frame_args)
return scribunto_param_key(preprocess(self[1], frame_args))
end
function Parameter:get_default(frame_args)
return tostring(self[2]) or "{{{" .. tostring(self[1]) .. "}}}"
end
function Parameter:preprocess(frame_args)
if not frame_args then
return preprocess(self[2], frame_args) or
"{{{" .. preprocess(self[1], frame_args) .. "}}}"
end
local name = preprocess(self[1], frame_args)
return frame_args[php_trim(name)] or
preprocess(self[2], frame_args) or
"{{{" .. name .. "}}}"
end
local Argument = Node:new_class("argument")
function Argument:__tostring()
return tostring(self[1]) .. "=" .. tostring(self[2])
end
local Template = Node:new_class("template")
function Template:__tostring()
local output = {}
for i = 1, #self do
output[i] = tostring(self[i])
end
return "{{" .. concat(output, "|") .. "}}"
end
function Template:get_arguments(frame_args)
local template_args, implicit = {}, 0
for i = 2, #self do
local arg = self[i]
if type_or_class(arg) == "argument" then
template_args[scribunto_param_key(preprocess(arg[1], frame_args))] =
php_trim(tostring(arg[2]))
else
implicit = implicit + 1
template_args[implicit] = tostring(arg) -- Not trimmed.
end
end
return template_args
end
-- Normalize the template name, check it's a valid template, then memoize results (using false for invalid titles).
-- Parser functions (e.g. {{#IF:a|b|c}}) need to have the first argument extracted from the title, as it comes after the colon. Because of this, the parser function and first argument are memoized as a table.
-- FIXME: Some parser functions have special argument handling (e.g. {{#SWITCH:}}).
do
local page_title = mw.title.getCurrentTitle()
local namespace_has_subpages = mw.site.namespaces[page_title.namespace].hasSubpages
local raw_pagename = page_title.fullText
local memo = {}
local function get_array_args(self)
local template_args = {}
for i = 2, #self do
template_args[i - 1] = tostring(self[i])
end
return template_args
end
local function convert_to_parser_function(self, name, arg1)
insert(self, 2, arg1)
self.get_arguments = get_array_args
return name
end
local function retrieve_magic_word_data(chunk)
local mgw_data = magic_words[chunk]
if mgw_data then
return mgw_data
end
local normalized = uupper(chunk)
mgw_data = magic_words[normalized]
if mgw_data and not mgw_data.case_sensitive then
return mgw_data
end
end
-- Returns the name required to transclude the title object `title` using
-- template {{ }} syntax.
local function get_template_invocation_name(title)
if not is_valid_title(title) then
error("Template invocations require a valid page title, which cannot contain an interwiki prefix.")
end
local namespace = title.namespace
-- If not in the template namespace, include the prefix (or ":" if
-- mainspace).
if namespace ~= 10 then
return namespace == 0 and ":" .. title.text or title.prefixedText
end
-- If in the template namespace and it shares a name with a magic word,
-- it needs the prefix "Template:".
local text = title.text
local colon = find(text, ":", 1, true)
if not colon then
local mgw_data = retrieve_magic_word_data(text)
return mgw_data and mgw_data.parser_variable and title.prefixedText or text
end
local mgw_data = retrieve_magic_word_data(sub(text, 1, colon - 1))
if mgw_data and (mgw_data.parser_function or mgw_data.transclusion_modifier) then
return title.prefixedText
end
-- Also if "Template:" is necessary for disambiguation (e.g.
-- "Template:Category:Foo" can't be abbreviated to "Category:Foo").
local check = new_title(text, 10)
return check and title_equals(title, check) and text or title.prefixedText
end
export.getTemplateInvocationName = get_template_invocation_name
-- Resolve any redirects. If the redirect target is an interwiki link, then
-- the template won't fail, but the redirect page itself gets transcluded
-- (i.e. the template name shouldn't be normalized to the target).
-- title.redirectTarget increments the expensive parser function count, but
-- avoids extraneous transclusions polluting template lists and the
-- performance hit caused by indiscriminately grabbing redirectTarget.
-- However, if the expensive parser function limit has already been hit,
-- redirectTarget is used as a fallback.
local function resolve_redirect(title, force_transclusion)
if not force_transclusion then
local success, is_redirect = pcall(getmetatable(title).__index, title, "isRedirect")
if success and not is_redirect then
return title
end
end
local redirect = title.redirectTarget
return is_valid_title(redirect) and redirect or title
end
function parse_template_name(name, has_args, fragment, force_transclusion)
local chunks, colon, start, n, p = {}, find(name, ":", 1, true), 1, 0, 0
while colon do
-- Pattern applies PHP ltrim.
local mgw_data = retrieve_magic_word_data(match(sub(name, start, colon - 1), "[^%z\t-\v\r ].*") or "")
if not mgw_data then
break
end
local priority = mgw_data.priority
if not (priority and priority > p) then
local pf = mgw_data.parser_function and mgw_data.name
if pf then
n = n + 1
chunks[n] = pf .. ":"
return chunks, sub(name, colon + 1)
end
break
end
n = n + 1
chunks[n] = mgw_data.name .. ":"
start, p = colon + 1, priority
colon = find(name, ":", start, true)
end
if start > 1 then
name = sub(name, start)
end
name = php_trim(name)
-- Parser variables can only take SUBST:/SAFESUBST: as modifiers.
if not has_args and p <= 1 then
local mgw_data = retrieve_magic_word_data(name)
local pv = mgw_data and mgw_data.parser_variable and mgw_data.name
if pv then
n = n + 1
chunks[n] = pv
return chunks
end
end
-- Handle relative template names.
if namespace_has_subpages then
-- If the name starts with "/", it's treated as a subpage of the
-- current page. Final slashes are trimmed, but this can't affect
-- the intervening slash (e.g. {{///}} refers to "{{PAGENAME}}/").
local initial = sub(name, 1, 1)
if initial == "/" then
name = raw_pagename .. (match(name, "^/.*[^/]") or "/")
-- If it starts with "../", trim it and any that follow, and go up
-- that many subpage levels. Then, treat any additional text as
-- a subpage of that page; final slashes are trimmed.
elseif initial == "." and sub(name, 2, 3) == "./" then
local n = 4
while sub(name, n, n + 2) == "../" do
n = n + 3
end
-- Retain an initial "/".
name = sub(name, n - 1)
-- Trim the relevant number of subpages from the pagename.
local pagename, i = reverse(raw_pagename), 0
for _ = 1, (n - 1) / 3 do
i = find(pagename, "/", i + 1, true)
-- Fail if there aren't enough slashes.
if not i then
return nil
end
end
-- Add the subpage text; since the intervening "/" is retained
-- in `name`, it can be trimmed along with any other final
-- slashes (e.g. {{..///}} refers to "{{BASEPAGENAME}}".)
name = reverse(sub(pagename, i + 1)) .. (match(name, "^.*[^/]") or "")
end
end
local title = new_title(name, 10)
if not is_valid_title(title) then
return nil
end
-- If `fragment` is set, save the original title's fragment, since it
-- won't carry through to any redirect targets.
if fragment then
fragment = title.fragment
end
title = resolve_redirect(title, force_transclusion)
local chunk = get_template_invocation_name(title)
-- Set the fragment (if applicable).
if fragment then
chunk = chunk .. "#" .. fragment
end
chunks[n + 1] = chunk
return chunks
end
-- Note: force_transclusion avoids incrementing the expensive parser
-- function count by forcing transclusion instead. This should only be used
-- when there is a real risk that the expensive parser function limit of
-- 500 will be hit.
function Template:get_name(frame_args, force_transclusion)
local name = preprocess(self[1], frame_args)
local norm = memo[name]
if norm then
if type(norm) == "table" then
return convert_to_parser_function(self, norm[1], norm[2])
end
return norm
elseif norm == false then
return nil
end
local chunks, pf_arg1 = parse_template_name(name, #self > 1, nil, force_transclusion)
-- Fail if invalid.
if not chunks then
memo[name] = false
return nil
end
local chunk1 = chunks[1]
-- Fail on SUBST:.
if chunk1 == "SUBST:" then
memo[name] = false
return nil
-- If pf_arg1 is returned, it's a parser function with pf_arg1 as the first argument.
-- Any modifiers are ignored.
elseif pf_arg1 then
local pf = chunks[#chunks]
memo[name] = {pf, pf_arg1}
return convert_to_parser_function(self, pf, pf_arg1)
end
-- Ignore SAFESUBST:, and treat MSGNW: as a parser function with the pagename as its first argument (ignoring any RAW: that comes after).
if chunks[chunk1 == "SAFESUBST:" and 2 or 1] == "MSGNW:" then
pf_arg1 = chunks[#chunks]
memo[name] = {"MSGNW:", pf_arg1}
return convert_to_parser_function(self, "MSGNW:", pf_arg1)
end
-- Ignore any remaining modifiers, as they've done their job.
local output = chunks[#chunks]
memo[name] = output
return output
end
end
function Template:preprocess()
return frame:preprocess(tostring(self))
end
local Tag = Node:new_class("tag")
function Tag:__tostring()
local open_tag, attributes, i = {"<", self.name}, self:get_attributes(), 2
for attr, value in next, attributes do
i = i + 1
-- Quote value using "" by default, '' if it contains ", and leave unquoted if it contains both.
local quoter = not find(value, "\"", 1, true) and "\"" or
not find(value, "'", 1, true) and "'" or
match(value, "^()[^\t\n\f\r ]*$") and "" or
-- This shouldn't happen, unless the node has been edited manually. Not possible to stringify in a way that can be interpreted by the native parser, since it doesn't recognise escapes.
error("Tag attribute values cannot contain all three of \", ' and whitespace simultaneously.")
open_tag[i] = " " .. attr .. "=" .. quoter .. value .. quoter
end
if self.self_closing then
return concat(open_tag) .. "/>"
end
return concat(open_tag) .. ">" .. concat(self) .. "</" .. self.name .. ">"
end
function Tag:get_attributes()
local raw = self.attributes
if not raw then
self.attributes = {}
return self.attributes
elseif type(raw) == "table" then
return raw
end
if sub(raw, -1) == "/" then
raw = sub(raw, 1, -2)
end
local attributes, head = {}, 1
-- Semi-manual implementation of the native regex.
while true do
local name, loc = match(raw, "([^\t\n\f\r />][^\t\n\f\r /=>]*)()", head)
if not name then
break
end
head = loc
local value
loc = match(raw, "^[\t\n\f\r ]*=[\t\n\f\r ]*()", head)
if loc then
head = loc
-- Either "", '' or the value ends on a space/at the end. Missing
-- end quotes are repaired by closing the value at the end.
value, loc = match(raw, "^\"([^\"]*)\"?()", head)
if not value then
value, loc = match(raw, "^'([^']*)'?()", head)
if not value then
value, loc = match(raw, "^([^\t\n\f\r ]*)()", head)
end
end
head = loc
end
-- valid_attribute_name is a pattern matching a valid attribute name.
-- Defined in the data due to its length - see there for more info.
if umatch(name, valid_attribute_name) then
-- Sanitizer applies PHP strtolower (ASCII-only).
attributes[lower(name)] = value and decode_entities(
php_trim(gsub(value, "[\t\n\r ]+", " "))
) or ""
end
end
self.attributes = attributes
return attributes
end
function Tag:preprocess()
return frame:preprocess(tostring(self))
end
local Heading = Node:new_class("heading")
function Heading:new(this)
if #this > 1 then
local success, str = pcall(concat, this)
if success then
return Node.new(self, {
str,
level = this.level,
section = this.section,
pos = this.pos
})
end
end
return Node.new(self, this)
end
function Heading:__tostring()
local eq = rep("=", self.level)
return eq .. Node.__tostring(self) .. eq
end
function Heading:get_name(frame_args)
return php_trim(Node.preprocess(self, frame_args))
end
function Heading:preprocess(frame_args)
local eq = rep("=", self.level)
return eq .. Node.preprocess(self, frame_args) .. eq
end
------------------------------------------------------------------------------------
--
-- Parser
--
------------------------------------------------------------------------------------
function Parser:read(i, j)
local head, i = self.head, i or 0
return sub(self.text, head + i, head + (j or i))
end
function Parser:advance(n)
self.head = self.head + (n or self[-1].step or 1)
end
function Parser:set_pattern(pattern)
local layer = self[-1]
layer.pattern = pattern
layer.nxt = nil
end
function Parser:consume()
local layer, this = self[-1]
if layer.nxt then
this = layer.nxt
layer.nxt = nil
else
local text, head = self.text, self.head
local loc1, loc2 = find(text, layer.pattern, head)
if loc1 == head or not loc1 then
this = sub(text, head, loc2)
else
this = sub(text, head, loc1 - 1)
layer.nxt = sub(text, loc1, loc2)
end
end
layer.step = #this
return layer.handler(self, this)
end
-- Template or parameter.
-- Parsed by matching the opening braces innermost-to-outermost (ignoring lone closing braces). Parameters {{{ }}} take priority over templates {{ }} where possible, but a double closing brace will always result in a closure, even if there are 3+ opening braces.
-- For example, "{{{{foo}}}}" (4) is parsed as a parameter enclosed by single braces, and "{{{{{foo}}}}}" (5) is a parameter inside a template. However, "{{{{{foo }} }}}" is a template inside a parameter, due to "}}" forcing the closure of the inner node.
do
-- Handlers.
local handle_name
local handle_argument
function handle_name(self, ...)
handle_name = self:switch(handle_name, {
["\n"] = Parser.heading_block,
["<"] = Parser.tag,
["["] = Parser.wikilink_block,
["{"] = Parser.braces,
["|"] = function(self)
self:emit(Wikitext:new(self:pop_sublayer()))
self:push_sublayer(handle_argument)
self:set_pattern("[\n<=[{|}]")
end,
["}"] = function(self)
if self:read(1) == "}" then
self:emit(Wikitext:new(self:pop_sublayer()))
return self:pop()
end
self:emit("}")
end,
[""] = Parser.fail_route,
[false] = Parser.emit
})
return handle_name(self, ...)
end
function handle_argument(self, ...)
local function emit_argument(self)
local arg = Wikitext:new(self:pop_sublayer())
local layer = self[-1]
local key = layer.key
if key then
arg = Argument:new{key, arg}
layer.key = nil
end
self:emit(arg)
end
handle_argument = self:switch(handle_argument, {
["\n"] = function(self)
if self[-1].key then
return self:heading_block()
end
self:newline()
while self:read(0, 2) == "\n==" do
self:advance()
self:emit(select(2, self:get("do_heading_block")))
end
end,
["<"] = Parser.tag,
["="] = function(self)
local key = Wikitext:new(self:pop_sublayer())
self[-1].key = key
self:push_sublayer(handle_argument)
self:set_pattern("[\n<[{|}]")
end,
["["] = Parser.wikilink_block,
["{"] = Parser.braces,
["|"] = function(self)
emit_argument(self)
self:push_sublayer(handle_argument)
self:set_pattern("[\n<=[{|}]")
end,
["}"] = function(self)
if self:read(1) == "}" then
emit_argument(self)
return self:pop()
end
self:emit("}")
end,
[""] = Parser.fail_route,
[false] = Parser.emit
})
return handle_argument(self, ...)
end
function Parser:do_template_or_parameter()
self:push_sublayer(handle_name)
self:set_pattern("[\n<[{|}]")
end
function Parser:template_or_parameter()
local text, head, node_to_emit = self.text, self.head
-- Comments/tags interrupt the brace count.
local braces = match(text, "^{+()", head) - head
self:advance(braces)
repeat
local success, node = self:get("do_template_or_parameter")
if not success then
self:emit(rep("{", braces))
break
elseif node_to_emit then
-- Nest the already-parsed node at the start of the new node.
local node1 = node[1]
node[1] = (
node1 == "" and node_to_emit or
Wikitext:new{node_to_emit, node1}
)
end
if self:read(2) == "}" and braces > 2 then
self:advance(3)
braces = braces - 3
node = Parameter:new(node)
else
self:advance(2)
braces = braces - 2
node = Template:new(node)
end
local pos = head + braces
node.pos = pos
node.raw = sub(text, pos, self.head - 1)
node_to_emit = node
if braces == 1 then
self:emit("{")
break
end
until braces == 0
if node_to_emit then
self:emit(node_to_emit)
end
return braces
end
end
-- Tag.
do
local tags = data.tags
-- Handlers.
local handle_start
local handle_tag
local function is_ignored_tag(self, this)
if self.transcluded then
return this == "includeonly"
end
return this == "noinclude" or this == "onlyinclude"
end
local function ignored_tag(self, text, head)
local loc = find(text, ">", head, true)
if not loc then
return self:fail_route()
end
self.head = loc
self[-1].ignored = true
return self:pop()
end
function handle_start(self, this)
if this == "/" then
local text, head = self.text, self.head + 1
local this = match(text, "^[^%s/>]+", head)
if this and is_ignored_tag(self, lower(this)) then
head = head + #this
if not match(text, "^/[^>]", head) then
return ignored_tag(self, text, head)
end
end
return self:fail_route()
elseif this == "" then
return self:fail_route()
end
-- Tags are only case-insensitive with ASCII characters.
local raw_name = this
this = lower(this)
local end_tag_pattern = tags[this]
if not end_tag_pattern then -- Validity check.
return self:fail_route()
end
local layer = self[-1]
local text, head = self.text, self.head + layer.step
if match(text, "^/[^>]", head) then
return self:fail_route()
elseif is_ignored_tag(self, this) then
return ignored_tag(self, text, head)
elseif this == "onlyinclude" then -- Not ignored and not active: plain text.
return self:fail_route()
elseif this == "noinclude" or this == "includeonly" then
layer.ignored = true -- Ignored block.
layer.raw_name = raw_name
end
layer.name, layer.handler, layer.end_tag_pattern = this, handle_tag, end_tag_pattern
self:set_pattern(">")
end
function handle_tag(self, this)
if this == "" then
return self:fail_route()
elseif this ~= ">" then
self[-1].attributes = this
return
elseif self:read(-1) == "/" then
self[-1].self_closing = true
return self:pop()
end
local text, head, layer = self.text, self.head + 1, self[-1]
local loc1, loc2 = find(text, layer.end_tag_pattern, head)
if loc1 then
if loc1 > head then
self:emit(sub(text, head, loc1 - 1))
end
self.head = loc2
return self:pop()
-- noinclude and includeonly will tolerate having no closing tag, but
-- only if given in lowercase. This is due to a preprocessor bug, as
-- it uses a regex with the /i (case-insensitive) flag to check for
-- end tags, but an array lookup when determining which tags will
-- tolerate no closing tag (case-sensitive).
elseif layer.ignored then
local raw_name = layer.raw_name
if raw_name == "noinclude" or raw_name == "includeonly" then
self.head = #self.text
return self:pop()
end
end
return self:fail_route()
end
function Parser:do_tag()
local layer = self[-1]
layer.handler = handle_start
self:set_pattern("[%s/>]")
self:advance()
end
local function find_next_chunk(text, pattern, head)
return select(2, find(text, pattern, head, true)) or #text
end
function Parser:tag()
-- HTML comment.
if self:read(1, 3) == "!--" then
self.head = find_next_chunk(self.text, "-->", self.head + 4)
-- onlyinclude closing tag (whitespace intolerant).
elseif self.onlyinclude and self:read(1, 13) == "/onlyinclude>" then
self.head = find_next_chunk(self.text, "<onlyinclude>", self.head + 14)
else
local success, tag = self:get("do_tag")
if not success then
self:emit("<")
elseif not tag.ignored then
tag.end_tag_pattern = nil
self:emit(Tag:new(tag))
end
end
end
end
-- Heading.
-- The preparser assigns each heading a number, which is used for things like section edit links. The preparser will only do this for heading blocks which aren't nested inside templates, parameters and parser tags. In some cases (e.g. when template blocks contain untrimmed newlines), a preparsed heading may not be treated as a heading in the final output. That does not affect the preparser, however, which will always count sections based on the preparser heading count, since it can't know what a template's final output will be.
do
-- Handlers.
local handle_start
local handle_body
local handle_possible_end
function handle_start(self, ...)
-- ===== is "=" as an L2; ======== is "==" as an L3 etc.
local function newline(self)
local layer = self[-1]
local eq = layer.level
if eq <= 2 then
return self:fail_route()
end
-- Calculate which equals signs determine the heading level.
local level_eq = eq - (2 - eq % 2)
level_eq = level_eq > 12 and 12 or level_eq
-- Emit the excess.
self:emit(rep("=", eq - level_eq))
layer.level = level_eq / 2
return self:pop()
end
local function whitespace(self)
local success, possible_end = self:get("do_heading_possible_end")
if success then
self:emit(Wikitext:new(possible_end))
local layer = self[-1]
layer.handler = handle_body
self:set_pattern("[\n<={]")
return self:consume()
end
return newline(self)
end
handle_start = self:switch(handle_start, {
["\t"] = whitespace,
["\n"] = newline,
[" "] = whitespace,
[""] = newline,
[false] = function(self)
-- Emit any excess = signs once we know it's a conventional heading. Up till now, we couldn't know if the heading is just a string of = signs (e.g. ========), so it wasn't guaranteed that the heading text starts after the 6th.
local layer = self[-1]
local eq = layer.level
if eq > 6 then
self:emit(1, rep("=", eq - 6))
layer.level = 6
end
layer.handler = handle_body
self:set_pattern("[\n<=[{]")
return self:consume()
end
})
return handle_start(self, ...)
end
function handle_body(self, ...)
handle_body = self:switch(handle_body, {
["\n"] = Parser.fail_route,
["<"] = Parser.tag,
["="] = function(self)
-- Comments/tags interrupt the equals count.
local eq = match(self.text, "^=+", self.head)
local eq_len = #eq
self:advance(eq_len)
local success, possible_end = self:get("do_heading_possible_end")
if success then
self:emit(eq)
self:emit(Wikitext:new(possible_end))
return self:consume()
end
local layer = self[-1]
local level = layer.level
if eq_len > level then
self:emit(rep("=", eq_len - level))
elseif level > eq_len then
layer.level = eq_len
self:emit(1, rep("=", level - eq_len))
end
return self:pop()
end,
["["] = Parser.wikilink_block,
["{"] = Parser.braces,
[""] = Parser.fail_route,
[false] = Parser.emit
})
return handle_body(self, ...)
end
function handle_possible_end(self, ...)
handle_possible_end = self:switch(handle_possible_end, {
["\n"] = Parser.fail_route,
["<"] = function(self)
local head = (
self:read(1, 3) == "!--" and
select(2, find(self.text, "-->", self.head + 4, true))
)
if not head then
return self:pop()
end
self.head = head
end,
[""] = Parser.fail_route,
[false] = function(self, this)
if not match(this, "^[\t ]+$") then
return self:pop()
end
self:emit(this)
end
})
return handle_possible_end(self, ...)
end
function Parser:do_heading()
local layer, head = self[-1], self.head
layer.handler, layer.pos = handle_start, head
self:set_pattern("[\t\n ]")
-- Comments/tags interrupt the equals count.
local eq = match(self.text, "^=+()", head) - head
layer.level = eq
self:advance(eq)
end
function Parser:do_heading_possible_end()
local layer = self[-1]
layer.handler = handle_possible_end
self:set_pattern("[\n<]")
end
function Parser:heading()
local success, heading = self:get("do_heading")
if success then
local section = self.section + 1
heading.section = section
self.section = section
self:emit(Heading:new(heading))
return self:consume()
else
self:emit("=")
end
end
end
------------------------------------------------------------------------------------
--
-- Block handlers
--
------------------------------------------------------------------------------------
-- Block handlers.
-- These are blocks which can affect template/parameter parsing, since they're also parsed by Parsoid at the same time (even though they aren't processed until later).
-- All blocks (including templates/parameters) can nest inside each other, but an inner block must be closed before the outer block which contains it. This is why, for example, the wikitext "{{template| [[ }}" will result in an unprocessed template, since the inner "[[" is treated as the opening of a wikilink block, which prevents "}}" from being treated as the closure of the template block. On the other hand, "{{template| [[ ]] }}" will process correctly, since the wikilink block is closed before the template closure. It makes no difference whether the block will be treated as valid or not when it's processed later on, so "{{template| [[ }} ]] }}" would also work, even though "[[ }} ]]" is not a valid wikilink.
-- Note that nesting also affects pipes and equals signs, in addition to block closures.
-- These blocks can be nested to any degree, so "{{template| [[ [[ [[ ]] }}" will not work, since only one of the three wikilink blocks has been closed. On the other hand, "{{template| [[ [[ [[ ]] ]] ]] }}" will work.
-- All blocks are implicitly closed by the end of the text, since their validity is irrelevant at this stage.
-- Language conversion block.
-- Opens with "-{" and closes with "}-". However, templates/parameters take priority, so "-{{" is parsed as "-" followed by the opening of a template/parameter block (depending on what comes after).
-- Note: Language conversion blocks aren't actually enabled on the English Wiktionary, but Parsoid still parses them at this stage, so they can affect the closure of outer blocks: e.g. "[[ -{ ]]" is not a valid wikilink block, since the "]]" falls inside the new language conversion block.
do
local function handle_language_conversion_block(self, ...)
handle_language_conversion_block = self:switch(handle_language_conversion_block, {
["\n"] = Parser.heading_block,
["<"] = Parser.tag,
["["] = Parser.wikilink_block,
["{"] = Parser.braces,
["}"] = function(self)
if self:read(1) == "-" then
self:emit("}-")
self:advance()
return self:pop()
end
self:emit("}")
end,
[""] = Parser.pop,
[false] = Parser.emit
})
return handle_language_conversion_block(self, ...)
end
function Parser:do_language_conversion_block()
local layer = self[-1]
layer.handler = handle_language_conversion_block
self:set_pattern("[\n<[{}]")
end
function Parser:braces()
local language_conversion_block = self:read(-1) == "-"
if self:read(1) == "{" then
local braces = self:template_or_parameter()
if not (braces == 1 and language_conversion_block) then
return self:consume()
end
else
self:emit("{")
if not language_conversion_block then
return
end
self:advance()
end
self:emit(Wikitext:new(select(2, self:get("do_language_conversion_block"))))
end
end
--[==[
Headings
Opens with "\n=" (or "=" at the start of the text), and closes with "\n" or the end of the text. Note that it doesn't matter whether the heading will fail to process due to a premature newline (e.g. if there are no closing signs), so at this stage the only thing that matters for closure is the newline or end of text.
Note: Heading blocks are only parsed like this if they occur inside a template, since they do not iterate the preparser's heading count (i.e. they aren't proper headings).
Note 2: if directly inside a template argument with no previous equals signs, a newline followed by a single equals sign is parsed as an argument equals sign, not the opening of a new L1 heading block. This does not apply to any other heading levels. As such, {{template|key\n=}}, {{template|key\n=value}} or even {{template|\n=}} will successfully close, but {{template|key\n==}}, {{template|key=value\n=more value}}, {{template\n=}} etc. will not, since in the latter cases the "}}" would fall inside the new heading block.
]==]
do
local function handle_heading_block(self, ...)
handle_heading_block = self:switch(handle_heading_block, {
["\n"] = function(self)
self:newline()
return self:pop()
end,
["<"] = Parser.tag,
["["] = Parser.wikilink_block,
["{"] = Parser.braces,
[""] = Parser.pop,
[false] = Parser.emit
})
return handle_heading_block(self, ...)
end
function Parser:do_heading_block()
local layer = self[-1]
layer.handler = handle_heading_block
self:set_pattern("[\n<[{]")
end
function Parser:heading_block()
self:newline()
while self:read(0, 1) == "\n=" do
self:advance()
self:emit(Wikitext:new(select(2, self:get("do_heading_block"))))
end
end
end
-- Wikilink block.
-- Opens with "[[" and closes with "]]".
do
local function handle_wikilink_block(self, ...)
handle_wikilink_block = self:switch(handle_wikilink_block, {
["\n"] = Parser.heading_block,
["<"] = Parser.tag,
["["] = Parser.wikilink_block,
["]"] = function(self)
if self:read(1) == "]" then
self:emit("]]")
self:advance()
return self:pop()
end
self:emit("]")
end,
["{"] = Parser.braces,
[""] = Parser.pop,
[false] = Parser.emit
})
return handle_wikilink_block(self, ...)
end
function Parser:do_wikilink_block()
local layer = self[-1]
layer.handler = handle_wikilink_block
self:set_pattern("[\n<[%]{]")
end
function Parser:wikilink_block()
if self:read(1) == "[" then
self:emit("[[")
self:advance(2)
self:emit(Wikitext:new(select(2, self:get("do_wikilink_block"))))
else
self:emit("[")
end
end
end
-- Lines which only contain comments, " " and "\t" are eaten, so long as
-- they're bookended by "\n" (i.e. not the first or last line).
function Parser:newline()
local text, head = self.text, self.head
while true do
repeat
local loc = match(text, "^[\t ]*<!%-%-()", head + 1)
if not loc then
break
end
loc = select(2, find(text, "-->", loc, true))
head = loc or head
until not loc
-- Fail if no comments found.
if head == self.head then
break
end
head = match(text, "^[\t ]*()\n", head + 1)
if not head then
break
end
self.head = head
end
self:emit("\n")
end
do
-- Handlers.
local handle_start
local main_handler
-- If the first character is "=", try parsing it as a heading.
function handle_start(self, this)
local layer = self[-1]
layer.handler = main_handler
self:set_pattern("[\n<{]")
if this == "=" then
return self:heading()
end
return self:consume()
end
function main_handler(self, ...)
main_handler = self:switch(main_handler, {
["\n"] = function(self)
self:newline()
if self:read(1) == "=" then
self:advance()
return self:heading()
end
end,
["<"] = Parser.tag,
["{"] = function(self)
if self:read(1) == "{" then
self:template_or_parameter()
return self:consume()
end
self:emit("{")
end,
[""] = Parser.pop,
[false] = Parser.emit
})
return main_handler(self, ...)
end
-- If `transcluded` is true, then the text is checked for a pair of
-- onlyinclude tags. If these are found (even if they're in the wrong
-- order), then the start of the page is treated as though it is preceded
-- by a closing onlyinclude tag.
-- Note 1: unlike other parser extension tags, onlyinclude tags are case-
-- sensitive and cannot contain whitespace.
-- Note 2: onlyinclude tags *can* be implicitly closed by the end of the
-- text, but the hard requirement above means this can only happen if
-- either the tags are in the wrong order or there are multiple onlyinclude
-- blocks.
function Parser:do_parse(transcluded)
local layer = self[-1]
layer.handler = handle_start
self:set_pattern(".")
self.section = 0
if not transcluded then
return
end
self.transcluded = true
local text = self.text
if find(text, "</onlyinclude>", 1, true) then
local head = find(text, "<onlyinclude>", 1, true)
if head then
self.onlyinclude = true
self.head = head + 13
end
end
end
function export.parse(text, transcluded)
local text_type = type(text)
return (select(2, Parser:parse{
text = text_type == "string" and text or
text_type == "number" and format("%.14g", text) or
error("bad argument #1 (string expected, got " .. text_type .. ")"),
node = {Wikitext, true},
route = {"do_parse", transcluded}
}))
end
parse = export.parse
end
function export.parseTemplate(text, not_transcluded)
text = parse(text, not not_transcluded)
if type_or_class(text) == "template" then
local name = text:get_name()
if name then
return name, text:get_arguments()
end
end
return nil, nil
end
do
local function next_template(iter)
while true do repeat -- break acts like continue
local node = iter()
if not node then
return nil, nil, nil, nil
elseif type_or_class(node) ~= "template" then
break
end
local name = node:get_name()
if name then
return name, node:get_arguments(), node.raw, node.pos
end
until true end
end
function export.findTemplates(text, not_transcluded)
return next_template, parse(text, not not_transcluded):__pairs("next_node")
end
end
do
local link_parameter_0 = data.template_link_param_0
local link_parameter_1 = data.template_link_param_1
-- Generate a link. If the target title doesn't have a fragment, use "#top"
-- (which is an implicit anchor at the top of every page), as this ensures
-- self-links still display as links, since bold display is distracting and
-- unintuitive for template links.
local function link_page(title, display)
local fragment = title.fragment
fragment = fragment == "" and "top" or fragment
return format(
"[[:%s|%s]]",
encode_uri(title.prefixedText .. "#" .. fragment, "WIKI"),
display
)
end
-- pf_arg0 or arg1 may need to be linked if a given parser function treats
-- them as a pagename. If a key exists in `namespace`, the value is the
-- namespace for the page: if not 0, then the namespace prefix will always
-- be added to the input (e.g. {{#invoke}} can only target the Module:
-- namespace, so inputting "Template:foo" gives "Module:Template:foo", and
-- "Module:foo" gives "Module:Module:foo"). However, this isn't possible
-- with mainspace (namespace 0), so prefixes are respected. make_title
-- handles all of this automatically.
local function finalize_arg(pagename, namespace)
if not namespace then
return pagename
end
local title = make_title(namespace, pagename)
if not (title and is_valid_title(title)) then
return pagename
end
return link_page(title, pagename)
end
local function render_title(name, args)
-- parse_template_name returns a table of transclusion modifiers plus
-- the normalized template/magic word name, which will be used as link
-- targets. The second return value pf_arg0 is argument 0, which is
-- returned if the target is a parser function (e.g. "foo" in
-- "{{#IF:foo|bar|baz}}"). Note: the second parameter checks if there
-- are any arguments, since parser variables cannot take arguments
-- (e.g. {{CURRENTYEAR}} is a parser variable, but {{CURRENTYEAR|foo}}
-- transcludes "Template:CURRENTYEAR"). In such cases, the returned
-- table explicitly includes the "Template:" prefix in the template
-- name. The third parameter instructs it to retain any fragment in the
-- template name in the returned table, if present.
local chunks, pf_arg0 = parse_template_name(
name,
args and pairs(args)(args) ~= nil,
true
)
if chunks == nil then
return name
end
local chunks_len = #chunks
-- Additionally, generate the corresponding table `rawchunks`, which
-- is a list of colon-separated chunks in the raw input. This is used
-- to retrieve the display forms for each chunk.
local rawchunks = split(name, ":")
for i = 1, chunks_len - 1 do
chunks[i] = format(
"[[%s|%s]]",
encode_uri(magic_words[sub(chunks[i], 1, -2)].transclusion_modifier, "WIKI"),
rawchunks[i]
)
end
local chunk = chunks[chunks_len]
local colon = sub(chunk, -1) == ":"
local mgw_data = magic_words[colon and sub(chunk, 1, -2) or chunk]
local link = mgw_data and (
colon and (mgw_data.parser_function or mgw_data.transclusion_modifier) or
not colon and mgw_data.parser_variable
)
-- If the name is not listed in magic_words, it must be a template,
-- so return a link to it with link_page, concatenating the remaining
-- chunks in `rawchunks` to form the display text.
-- Use new_title with the default namespace 10 (Template:) to generate
-- a target title, which is the same setting used for retrieving
-- templates (including those in other namespaces, as prefixes override
-- the default).
if not link then
chunks[chunks_len] = link_page(
new_title(chunk, 10),
concat(rawchunks, ":", chunks_len) -- :
)
return concat(chunks, ":") -- :
end
local arg1 = args and args[1] or nil
-- Otherwise, it's a magic word. Some magic words have different links,
-- depending on whether argument 1 is specified (e.g. "baz" in
-- {{foo:bar|baz}}).
if type(link) == "table" then
link = arg1 and link[1] or link[0]
end
chunks[chunks_len] = format(
"[[%s|%s]]",
encode_uri(link, "WIKI"),
rawchunks[chunks_len]
)
-- If we don't have pf_arg0, it must be a parser variable, so return.
if not pf_arg0 then
return concat(chunks, ":") -- :
-- #TAG: has special handling, because documentation links for parser
-- extension tags come from [[Module:data/parser extension tags]].
elseif chunk == "#TAG:" then
-- Tags are only case-insensitive with ASCII characters.
local tag = parser_extension_tags[lower(php_trim(pf_arg0))]
if tag then
pf_arg0 = format(
"[[%s|%s]]",
encode_uri(tag, "WIKI"),
pf_arg0
)
end
-- Otherwise, finalize pf_arg0 and add it to `chunks`.
else
pf_arg0 = finalize_arg(pf_arg0, link_parameter_0[chunk])
end
chunks[chunks_len + 1] = pf_arg0
-- Finalize arg1 (if applicable) then return.
if arg1 then
args[1] = finalize_arg(arg1, link_parameter_1[chunk])
end
return concat(chunks, ":") -- :
end
function export.buildTemplate(title, args)
local output = {title}
-- Iterate over all numbered parameters in order, followed by any
-- remaining parameters in codepoint order. Implicit parameters are
-- used wherever possible, even if explicit numbers are interpolated
-- between them (e.g. 0 would go before any implicit parameters, and
-- 2.5 between 2 and 3).
-- TODO: handle "=" and "|" in params/values.
if args then
local iter, implicit = sorted_pairs(args), table_len(args)
local k, v = iter()
while k ~= nil do
if type(k) == "number" and k >= 1 and k <= implicit and k % 1 == 0 then
insert(output, v)
else
insert(output, k .. "=" .. v)
end
k, v = iter()
end
end
return output
end
build_template = export.buildTemplate
function export.templateLink(title, args, no_link)
local output = build_template(no_link and title or render_title(title, args), args)
for i = 1, #output do
output[i] = encode_entities(output[i], "={}", true, true)
end
return tostring(html_create("code")
:css("white-space", "pre-wrap")
:css("background-color", "inherit")
:wikitext("{{" .. concat(output, "|") .. "}}") -- {{ | }}
)
end
end
do
local function next_parameter(iter)
while true do
local node = iter()
if not node then
return nil, nil, nil, nil
elseif type_or_class(node) == "parameter" then
local frame_args = iter.frame_args
return node:get_name(frame_args), node:get_default(frame_args), node.raw, node.pos
end
end
end
function export.findParameters(text, frame_args, not_transcluded)
local iter = parse(text, not not_transcluded):__pairs("next_node")
iter.frame_args = frame_args
return next_parameter, iter
end
function export.displayParameter(name, default)
return tostring(html_create("code")
:css("white-space", "pre-wrap")
:css("background-color", "inherit")
:wikitext("{{{" .. concat({name, default}, "|") .. "}}}") -- {{{ | }}}
)
end
end
do
local function check_level(level)
if type(level) ~= "number" then
error("Heading levels must be numbers.")
elseif level < 1 or level > 6 or level % 1 ~= 0 then
error("Heading levels must be integers between 1 and 6.")
end
return level
end
local function next_heading(iter)
while true do repeat -- break acts like continue
local node = iter()
if not node then
return nil, nil, nil, nil
elseif type_or_class(node) ~= "heading" then
break
end
local level = node.level
if level < iter.i or level > iter.j then
break
end
local name = node:get_name()
if not find(name, "\n", 1, true) then
return name, level, node.section, node.pos
end
until true end
end
-- Note: heading names can contain "\n" (e.g. inside nowiki tags), which
-- causes any heading containing them to fail. When that happens, the
-- heading is not returned by this function, but the heading count is still
-- iterated, since Parsoid's preprocessor still counts it as a heading for
-- the purpose of heading strip markers (i.e. the section number).
-- TODO: section numbers for edit links seem to also include headings
-- nested inside templates and parameters (but apparently not those in
-- parser extension tags - need to test this more). If we ever want to add
-- section edit links manually, this will need to be accounted for.
function export.findHeadings(text, i, j)
local iter = parse(text):__pairs("next_node")
iter.i, iter.j = i and check_level(i) or 1, j and check_level(j) or 6
return next_heading, iter
end
end
do
local transclusion_tag_link = "mw:Transclusion#Partial transclusion"
local function make_tag(tag)
return tostring(html_create("code")
:css("white-space", "pre-wrap")
:css("background-color", "inherit")
:wikitext("<" .. tag .. ">")
)
end
-- Note: invalid tags are returned without links.
function export.wikitagLink(tag)
-- ">" can't appear in tags (including attributes) since the parser
-- unconditionally treats ">" as the end of a tag.
if find(tag, ">") then
return make_tag(tag)
end
-- Tags must start "<tagname..." or "</tagname...", with no whitespace
-- after "<" or "</".
local slash, tagname, remainder = match(tag, "^(/?)([^/%s]+)(.*)$")
if not tagname then
return make_tag(tag)
end
-- Tags are only case-insensitive with ASCII characters.
local link = lower(tagname)
if (
-- onlyinclude tags must be lowercase and are whitespace intolerant.
link == "onlyinclude" and (link ~= tagname or remainder ~= "") or
-- Closing wikitags (except onlyinclude) can only have whitespace
-- after the tag name.
slash == "/" and not match(remainder, "^%s*$") or
-- Tagnames cannot be followed immediately by "/", unless it comes
-- at the end.
remainder ~= "/" and sub(remainder, 1, 1) == "/"
) then
return make_tag(tag)
end
if link == "noinclude" or link == "includeonly" or link == "onlyinclude" then
link = transclusion_tag_link
else
link = parser_extension_tags[link]
end
if link then
tag = gsub(tag, tagname, "[[" .. replacement_escape(encode_uri(link, "WIKI")) .. "|%0]]", 1)
end
return make_tag(tag)
end
end
return export