Module:template parser: Difference between revisions

Jump to navigation Jump to search

Module:template parser (view source)

Revision as of 12:28, 8 January 2025

18,177 bytes added ,  8 January
no edit summary
(Created page with "--[[ 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...")
 
No edit summary
 
Line 6: Line 6:
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.
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 export = {}
local data_module = "Module:template parser/data"
local load_module = "Module:load"
local magic_words_data_module = "Module:data/magic words"
local pages_module = "Module:pages"
local parser_extension_tags_data_module = "Module:data/parser extension tags"
local parser_module = "Module:parser"
local string_utilities_module = "Module:string utilities"
local table_module = "Module:table"
local require = require
local require = require
local m_parser = require("Module:parser")
 
local m_str_utils = require("Module:string utilities")
local m_parser = require(parser_module)
local mw = mw
local mw = mw
local mw_title = mw.title
local mw_title = mw.title
local mw_uri = mw.uri
local string = string
local string = string
local table = table
local table = table


local anchor_encode = mw_uri.anchorEncode
local build_template -- defined as export.buildTemplate below
local class_else_type = m_parser.class_else_type
local concat = table.concat
local concat = table.concat
local decode_entities = m_str_utils.decode_entities
local encode_uri = mw_uri.encode
local find = string.find
local find = string.find
local format = string.format
local format = string.format
local gsub = string.gsub
local gsub = string.gsub
local html_create = mw.html.create
local insert = table.insert
local insert = table.insert
local is_node = m_parser.is_node
local is_node = m_parser.is_node
local lower = m_str_utils.lower
local lower = string.lower
local make_title = mw_title.makeTitle -- unconditionally adds the specified namespace prefix
local match = string.match
local match = string.match
local new_title = mw_title.new
local new_title = mw_title.new -- specified namespace prefix is only added if the input doesn't contain one
local next = next
local next = next
local pairs = pairs
local parse -- defined as export.parse below
local parse -- defined as export.parse below
local parse_template_name -- defined as export.parseTemplateName below
local parse_template_name -- defined below
local pcall = pcall
local pcall = pcall
local php_trim = m_str_utils.php_trim
local rep = string.rep
local rep = string.rep
local reverse = string.reverse
local reverse = string.reverse
local scribunto_param_key = m_str_utils.scribunto_param_key
local select = select
local select = select
local sub = string.sub
local sub = string.sub
Line 37: Line 53:
local tostring = m_parser.tostring
local tostring = m_parser.tostring
local type = type
local type = type
local type_or_class = m_parser.type_or_class
local umatch = mw.ustring.match
local umatch = mw.ustring.match
local upper = m_str_utils.upper


local data = mw.loadData("Module:template parser/data")
--[==[
local frame = mw.getCurrentFrame()
Loaders for functions in other modules, which overwrite themselves with the target function when called. This ensures modules are only loaded when needed, retains the speed/convenience of locally-declared pre-loaded functions, and has no overhead after the first call, since the target functions are called directly in any subsequent calls.]==]
local invalid_tag_attribute_name_char = data.invalid_tag_attribute_name_char
local function decode_entities(...)
decode_entities = require(string_utilities_module).decode_entities
return decode_entities(...)
end
local function encode_entities(...)
encode_entities = require(string_utilities_module).encode_entities
return encode_entities(...)
end
local function get_namespace_shortcut(...)
get_namespace_shortcut = require(pages_module).get_namespace_shortcut
return get_namespace_shortcut(...)
end
local function is_internal_title(...)
is_internal_title = require(pages_module).is_internal_title
return is_internal_title(...)
end
local function load_data(...)
load_data = require(load_module).load_data
return load_data(...)
end
local function pattern_escape(...)
pattern_escape = require(string_utilities_module).pattern_escape
return pattern_escape(...)
end
local function php_trim(...)
php_trim = require(string_utilities_module).php_trim
return php_trim(...)
end
local function replacement_escape(...)
replacement_escape = require(string_utilities_module).replacement_escape
return replacement_escape(...)
end
local function scribunto_param_key(...)
scribunto_param_key = require(string_utilities_module).scribunto_param_key
return scribunto_param_key(...)
end
local function shallow_copy(...)
shallow_copy = require(table_module).shallowCopy
return shallow_copy(...)
end
local function sorted_pairs(...)
sorted_pairs = require(table_module).sortedPairs
return sorted_pairs(...)
end
local function split(...)
split = require(string_utilities_module).split
return split(...)
end
local function table_len(...)
table_len = require(table_module).length
return table_len(...)
end
local function uupper(...)
uupper = require(string_utilities_module).upper
return uupper(...)
end
 
--[==[
Loaders for objects, which load data (or some other object) into some variable, which can then be accessed as "foo or get_foo()", where the function get_foo sets the object to "foo" and then returns it. This ensures they are only loaded when needed, and avoids the need to check for the existence of the object each time, since once "foo" has been set, "get_foo" will not be called again.]==]
local data
local function get_data()
data, get_data = load_data(data_module), nil
return data
end
local frame
local function get_frame()
frame, get_frame = mw.getCurrentFrame(), nil
return frame
end
local magic_words
local function get_magic_words()
magic_words, get_magic_words = load_data(magic_words_data_module), nil
return magic_words
end
local parser_extension_tags
local function get_parser_extension_tags()
parser_extension_tags, get_parser_extension_tags = load_data(parser_extension_tags_data_module), nil
return parser_extension_tags
end


local Parser, Node = m_parser.new()
local Parser, Node = m_parser.new()
local function preprocess(text, args)
return is_node(text) and text:preprocess(args) or text
end
local export = {}


------------------------------------------------------------------------------------
------------------------------------------------------------------------------------
Line 59: Line 161:
------------------------------------------------------------------------------------
------------------------------------------------------------------------------------


function Node:preprocess(args)
Node.keys_to_remove = {"handler", "head", "pattern", "route", "step"}
 
local function expand(obj, frame_args)
return is_node(obj) and obj:expand(frame_args) or obj
end
export.expand = expand
 
function Node:expand(frame_args)
local output = {}
local output = {}
for i = 1, #self do
for i = 1, #self do
output[i] = preprocess(self[i], args)
output[i] = expand(self[i], frame_args)
end
end
return concat(output)
return concat(output)
Line 84: Line 193:
end
end


-- First value is the argument name.
-- First value is the parameter name.
-- Second value is the argument's default value.
-- Second value is the parameter's default value.
-- Any additional values are ignored: "{{{a|b|c}}}" is argument "a" with default value "b" (*not* "b|c").
-- Any additional values are ignored: e.g. "{{{a|b|c}}}" is parameter "a" with default value "b" (*not* "b|c").
local Argument = Node:new_class("argument")
local Parameter = Node:new_class("parameter")


function Argument:new(this)
function Parameter:new(this)
local this2 = this[2]
local this2 = this[2]
if type_or_class(this2) == "parameter" then
if class_else_type(this2) == "argument" then
insert(this2, 2, "=")
insert(this2, 2, "=")
this2 = Wikitext:new(this2)
this2 = Wikitext:new(this2)
end
end
return Node.new(self, {this[1], this2})
if this[3] == nil then
this[2] = this2
else
this = {this[1], this2}
end
return Node.new(self, this)
end
end


function Argument:__tostring()
function Parameter:__tostring()
local output = {}
local output = {}
for i = 1, #self do
for i = 1, #self do
Line 106: Line 220:
end
end


function Argument:next(i)
function Parameter:get_name(frame_args)
i = i + 1
return scribunto_param_key(expand(self[1], frame_args))
if i <= 2 then
return self[i], i
end
end
end


function Argument:get_name(args)
function Parameter:get_default(frame_args)
return scribunto_param_key(preprocess(self[1], args))
local default = self[2]
if default ~= nil then
return expand(default, frame_args)
end
return "{{{" .. expand(self[1], frame_args) .. "}}}"
end
end


function Argument:get_default(args)
function Parameter:expand(frame_args)
return tostring(self[2]) or "{{{" .. tostring(self[1]) .. "}}}"
if frame_args == nil then
end
return self:get_default()
 
end
function Argument:preprocess(args)
local name = expand(self[1], frame_args)
if not args then
local val = frame_args[scribunto_param_key(name)] -- Parameter in use.
return preprocess(self[2], args) or
if val ~= nil then
"{{{" .. preprocess(self[1], args) .. "}}}"
return val
end
val = self[2] -- Default.
if val ~= nil then
return expand(val, frame_args)
end
end
local name = preprocess(self[1], args)
return "{{{" .. name .. "}}}"
return args[php_trim(name)] or
preprocess(self[2], args) or
"{{{" .. name .. "}}}"
end
end


local Parameter = Node:new_class("parameter")
local Argument = Node:new_class("argument")


function Parameter:__tostring()
function Argument:__tostring()
return tostring(self[1]) .. "=" .. tostring(self[2])
return tostring(self[1]) .. "=" .. tostring(self[2])
end
function Argument:expand(frame_args)
return expand(self[1], frame_args) .. "=" .. expand(self[2], frame_args)
end
end


Line 146: Line 266:
end
end
return "{{" .. concat(output, "|") .. "}}"
return "{{" .. concat(output, "|") .. "}}"
end
function Template:get_params(args)
local params, implicit = {}, 0
for i = 2, #self do
local param = self[i]
if type_or_class(param) == "parameter" then
params[scribunto_param_key(preprocess(param[1], args))] = php_trim(tostring(param[2]))
else
implicit = implicit + 1
params[implicit] = tostring(param) -- Not trimmed.
end
end
return params
end
end


Line 166: Line 272:
-- FIXME: Some parser functions have special argument handling (e.g. {{#SWITCH:}}).
-- FIXME: Some parser functions have special argument handling (e.g. {{#SWITCH:}}).
do
do
local memo = {}
local page_title = mw_title.getCurrentTitle()
local namespace_has_subpages = mw.site.namespaces[page_title.namespace].hasSubpages
local raw_pagename = page_title.fullText
local templates = {}
local parser_variables = {}
local parser_functions = {}
local function retrieve_magic_word_data(chunk)
local mgw_data = (magic_words or get_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
local function get_prefixed_invocation_name(title, namespace, shortcut)
return shortcut and get_namespace_shortcut(title) .. ":" .. title.text or
namespace == 0 and ":" .. title.prefixedText or
title.prefixedText
end
local function get_array_params(self)
-- Returns the name required to transclude the title object `title` using
local params = {}
-- template {{ }} syntax. If the `shortcut` flag is set, then any calls
for i = 2, #self do
-- which require a namespace prefix will use the abbreviated form where one
params[i - 1] = tostring(self[i])
-- exists (e.g. "Template:PAGENAME" becomes "T:PAGENAME").
local function get_template_invocation_name(title, shortcut)
if not (title and is_internal_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 get_prefixed_invocation_name(title, namespace, shortcut)
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, ":", nil, true)
if not colon then
local mgw_data = retrieve_magic_word_data(text)
return mgw_data and mgw_data.parser_variable and get_prefixed_invocation_name(title, namespace, shortcut) 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 get_prefixed_invocation_name(title, namespace, shortcut)
end
end
return params
-- Also if "Template:" is necessary for disambiguation (e.g.
-- "Template:Category:Foo" can't be called with "Category:Foo").
local check = new_title(text, namespace)
return check and title_equals(title, check) and text or get_prefixed_invocation_name(title, namespace, shortcut)
end
export.getTemplateInvocationName = get_template_invocation_name
-- Returns whether a title is a redirect or not. Structured like this to
-- allow the use of pcall, since it will throw an error if the expensive
-- parser function limit has been reached.
local function is_redirect(title)
return title.isRedirect
end
end
local function convert_to_parser_function(self, name, arg1)
function parse_template_name(name, has_args, fragment, force_transclusion)
insert(self, 2, arg1)
local chunks, colon, start, n, p = {}, find(name, ":", nil, true), 1, 0, 0
self.get_params = get_array_params
while colon do
return name
-- 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 or nil
if pf then
n = n + 1
chunks[n] = pf .. ":"
return chunks, "parser function", 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 or nil
if pv then
n = n + 1
chunks[n] = pv
return chunks, "parser variable"
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 (title and is_internal_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
-- Resolve any redirects. Note that is_internal_title treats interwiki
-- titles as invalid, which is correct in this case: if the redirect
-- target is an interwiki link, the template won't fail, but the
-- redirect does not get resolved (i.e. the redirect page itself gets
-- transcluded, so the template name should not be normalized to the
-- target). It also treats titles that only have fragments as invalid
-- (e.g. "#foo"), but these can't be used as redirects anyway.
-- 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. force_transclusion forces the
-- use of the fallback.
local redirect = true
if not force_transclusion then
local success, resolved = pcall(is_redirect, title)
if success and not resolved then
redirect = false
end
end
if redirect then
redirect = title.redirectTarget
if redirect and is_internal_title(redirect) then
title = redirect
end
end
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, "template"
end
end
function Template:get_name(args)
-- Note: force_transclusion avoids incrementing the expensive parser
local name = preprocess(self[1], args)
-- function count by forcing transclusion instead. This should only be used
local norm = memo[name]
-- when there is a real risk that the expensive parser function limit of
-- 500 will be hit.
local function process_name(self, frame_args, force_transclusion)
local name = expand(self[1], frame_args)
local has_args, norm = #self > 1
if not has_args then
norm = parser_variables[name]
if norm then
return norm, "parser variable"
end
end
norm = templates[name]
if norm then
if norm then
if type(norm) == "table" then
local pf_arg1 = parser_functions[name]
return convert_to_parser_function(self, norm[1], norm[2])
return norm, pf_arg1 and "parser function" or "template", pf_arg1
end
return norm
elseif norm == false then
elseif norm == false then
return
return nil
end
end
local chunks, pf_arg1 = parse_template_name(name, #self > 1)
local chunks, subclass, pf_arg1 = parse_template_name(name, has_args, nil, force_transclusion)
-- Fail if invalid.
-- Fail if invalid.
if not chunks then
if not chunks then
memo[name] = false
templates[name] = false
return
return nil
end
end
local chunk1 = chunks[1]
local chunk1 = chunks[1]
-- Fail on SUBST:.
-- Fail on SUBST:.
if chunk1 == "SUBST:" then
if chunk1 == "SUBST:" then
memo[name] = false
templates[name] = false
return
return nil
-- If pf_arg1 is returned, it's a parser function with pf_arg1 as the first argument.
-- Any modifiers are ignored.
-- Any modifiers are ignored.
elseif pf_arg1 then
elseif subclass == "parser function" then
local pf = chunks[#chunks]
local pf = chunks[#chunks]
memo[name] = {pf, pf_arg1}
templates[name] = pf
return convert_to_parser_function(self, pf, pf_arg1)
parser_functions[name] = pf_arg1
return pf, "parser function", pf_arg1
end
end
-- Ignore SAFESUBST:, and treat MSGNW: as a parser function with the pagename as its first argument (ignoring any RAW: that comes after).
-- 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
if chunks[chunk1 == "SAFESUBST:" and 2 or 1] == "MSGNW:" then
pf_arg1 = chunks[#chunks]
pf_arg1 = chunks[#chunks]
memo[name] = {"MSGNW:", pf_arg1}
local pf = "MSGNW:"
return convert_to_parser_function(self, "MSGNW:", pf_arg1)
templates[name] = pf
parser_functions[name] = pf_arg1
return pf, "parser function", pf_arg1
end
end
-- Ignore any remaining modifiers, as they've done their job.
-- Ignore any remaining modifiers, as they've done their job.
local output = chunks[#chunks]
local output = chunks[#chunks]
memo[name] = output
if subclass == "parser variable" then
return output
parser_variables[name] = output
else
templates[name] = output
end
return output, subclass
end
function Template:get_name(frame_args, force_transclusion)
-- Only return the first return value.
return (process_name(self, frame_args, force_transclusion))
end
function Template:get_arguments(frame_args)
local name, subclass, pf_arg1 = process_name(self, frame_args)
if name == nil then
return nil
elseif subclass == "parser variable" then
return {}
end
local template_args = {}
if subclass == "parser function" then
template_args[1] = pf_arg1
for i = 2, #self do
template_args[i] = expand(self[i], frame_args) -- Not trimmed.
end
return template_args
end
local implicit = 0
for i = 2, #self do
local arg = self[i]
if class_else_type(arg) == "argument" then
template_args[scribunto_param_key(expand(arg[1], frame_args))] = php_trim(expand(arg[2], frame_args))
else
implicit = implicit + 1
template_args[implicit] = expand(arg, frame_args) -- Not trimmed.
end
end
return template_args
end
end
end
end


function Template:preprocess()
-- BIG TODO: manual template expansion.
return frame:preprocess(tostring(self))
function Template:expand()
return (frame or get_frame()):preprocess(tostring(self))
end
end


local Tag = Node:new_class("tag")
local Tag = Node:new_class("tag")


function Tag:__tostring()
do
local open_tag, attributes, i = {"<", self.name}, self:get_attributes(), 2
local php_htmlspecialchars_data
for attr, value in next, attributes do
i = i + 1
local function get_php_htmlspecialchars_data()
-- Quote value using "" by default, '' if it contains ", and leave unquoted if it contains both.
php_htmlspecialchars_data, get_php_htmlspecialchars_data = (data or get_data()).php_htmlspecialchars, nil
local quoter = not find(value, "\"", 1, true) and "\"" or
return php_htmlspecialchars_data
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
end
if self.self_closing then
return concat(open_tag) .. "/>"
local function php_htmlspecialchars(str, compat)
return (gsub(str, compat and "[&\"<>]" or "[&\"'<>]", php_htmlspecialchars_data or get_php_htmlspecialchars_data()))
end
end
return concat(open_tag) .. ">" .. concat(self) .. "</" .. self.name .. ">"
end
function Tag:__tostring()
 
local open_tag, attributes, n = {"<", self.name}, self:get_attributes(), 2
function Tag:get_attributes()
for attr, value in next, attributes do
local raw = self.attributes
n = n + 1
if not raw then
open_tag[n] = " " .. php_htmlspecialchars(attr) .. "=\"" .. php_htmlspecialchars(value, true) .. "\""
self.attributes = {}
end
return self.attributes
if self.self_closing then
elseif type(raw) == "table" then
return concat(open_tag) .. "/>"
return raw
end
return concat(open_tag) .. ">" .. concat(self) .. "</" .. self.name .. ">"
end
end
if sub(raw, -1) == "/" then
raw = sub(raw, 1, -2)
local valid_attribute_name
local function get_valid_attribute_name()
valid_attribute_name, get_valid_attribute_name = (data or get_data()).valid_attribute_name, nil
return valid_attribute_name
end
end
local attributes, head = {}, 1
-- Semi-manual implementation of the native regex.
function Tag:get_attributes()
while true do
local raw = self.attributes
local name, loc = match(raw, "([^\t\n\f\r />][^\t\n\f\r /=>]*)()", head)
if not raw then
if not name then
self.attributes = {}
break
return self.attributes
elseif type(raw) == "table" then
return raw
end
if sub(raw, -1) == "/" then
raw = sub(raw, 1, -2)
end
end
head = loc
local attributes, head = {}, 1
local value
-- Semi-manual implementation of the native regex.
loc = match(raw, "^[\t\n\f\r ]*=[\t\n\f\r ]*()", head)
while true do
if loc then
local name, loc = match(raw, "([^\t\n\f\r />][^\t\n\f\r /=>]*)()", head)
if not name then
break
end
head = loc
head = loc
value = match(raw, "^%b\"\"", head) or match(raw, "^%b''", head)
local value
if value then
loc = match(raw, "^[\t\n\f\r ]*=[\t\n\f\r ]*()", head)
head = head + #value
if loc then
value = sub(value, 2, -2)
head = loc
else
-- Either "", '' or the value ends on a space/at the end. Missing
local raw_value
-- end quotes are repaired by closing the value at the end.
raw_value, value = match(raw, "^([\"']?([^\t\n\f\r ]*))", head)
value, loc = match(raw, "^\"([^\"]*)\"?()", head)
head = head + #raw_value
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 or get_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
end
end
if not (
self.attributes = attributes
match(name, "^[%-.]") or
return attributes
umatch(name, invalid_tag_attribute_name_char)
) then
attributes[lower(name)] = value and decode_entities(
php_trim(gsub(value, "[\t\n\r ]+", " "))
) or ""
end
end
end
self.attributes = attributes
return attributes
end
end


function Tag:preprocess()
function Tag:expand()
return frame:preprocess(tostring(self))
return (frame or get_frame()):preprocess(tostring(self))
end
end


Line 308: Line 642:
level = this.level,
level = this.level,
section = this.section,
section = this.section,
pos = this.pos
index = this.index
})
})
end
end
Line 320: Line 654:
end
end


function Heading:get_name(args)
do
return php_trim(Node.preprocess(self, args))
local expand_node = Node.expand
end
 
-- Expanded heading names can contain "\n" (e.g. inside nowiki tags), which
function Heading:preprocess(args)
-- causes any heading containing them to fail. However, in such cases, the
local eq = rep("=", self.level)
-- native parser still treats it as a heading for the purpose of section
return eq .. Node.preprocess(self, args) .. eq
-- numbers.
local function validate_name(self, frame_args)
local name = expand_node(self, frame_args)
if find(name, "\n", nil, true) then
return nil
end
return name
end
function Heading:get_name(frame_args)
local name = validate_name(self, frame_args)
return name ~= nil and php_trim(name) or nil
end
-- FIXME: account for anchor disambiguation.
function Heading:get_anchor(frame_args)
local name = validate_name(self, frame_args)
return name ~= nil and decode_entities(anchor_encode(name)) or nil
end
function Heading:expand(frame_args)
local eq = rep("=", self.level)
return eq .. expand_node(self, frame_args) .. eq
end
end
end


Line 344: Line 701:
end
end


function Parser:consume(this)
function Parser:jump(head)
self.head = head
self[-1].nxt = nil
end
 
function Parser:set_pattern(pattern)
local layer = self[-1]
layer.pattern = pattern
layer.nxt = nil
end
 
function Parser:consume()
local layer = self[-1]
local layer = self[-1]
if not this then
local this = layer.nxt
if this then
layer.nxt = nil
else
local text, head = self.text, self.head
local text, head = self.text, self.head
local loc1, loc2 = find(text, layer.pattern, head)
local loc1, loc2 = find(text, layer.pattern, head)
this = sub(text, head, loc1 and (loc1 == head and loc2 or loc1 - 1) or nil)
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
end
layer.step = #this
layer.step = #this
Line 355: Line 731:
end
end


-- Template or argument.
-- Template or parameter.


-- Parsed by matching the opening braces innermost-to-outermost (ignoring lone closing braces). Arguments {{{ }}} take priority over templates {{ }} where possible, but a double closing brace will always result in a closure, even if there are 3+ opening braces.
-- 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 an argument enclosed by single braces, and "{{{{{foo}}}}}" (5) is an argument inside a template. However, "{{{{{foo }} }}}" is a template inside an argument, due to "}}" forcing the closure of the inner node.
-- 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
do
-- Handlers.
-- Handlers.
local handle_name
local handle_name
local handle_parameter
local handle_argument
local function do_template_or_parameter(self, inner_node)
self:push_sublayer(handle_name)
self:set_pattern("[\n<[{|}]")
-- If a node has already been parsed, nest it at the start of the new
-- outer node (e.g. when parsing"{{{{foo}}bar}}", the template "{{foo}}"
-- is parsed first, since it's the innermost, and becomes the first
-- node of the outer template.
if inner_node then
self:emit(inner_node)
end
end
function handle_name(self, ...)
function handle_name(self, ...)
Line 374: Line 762:
["|"] = function(self)
["|"] = function(self)
self:emit(Wikitext:new(self:pop_sublayer()))
self:emit(Wikitext:new(self:pop_sublayer()))
self:push_sublayer(handle_parameter)
self:push_sublayer(handle_argument)
self[-1].pattern = "[\n<=[{|}]"
self:set_pattern("[\n<=[{|}]")
end,
end,
Line 392: Line 780:
end
end
function handle_parameter(self, ...)
function handle_argument(self, ...)
local function emit_parameter(self)
local function emit_argument(self)
local param = Wikitext:new(self:pop_sublayer())
local arg = Wikitext:new(self:pop_sublayer())
local layer = self[-1]
local layer = self[-1]
local key = layer.key
local key = layer.key
if key then
if key then
param = Parameter:new{key, param}
arg = Argument:new{key, arg}
layer.key = nil
layer.key = nil
end
end
self:emit(param)
self:emit(arg)
end
end
handle_parameter = self:switch(handle_parameter, {
handle_argument = self:switch(handle_argument, {
["\n"] = function(self)
["\n"] = function(self)
if self[-1].key then
return self:heading_block("\n", self[-1].key and "=" or "==")
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,
end,
Line 421: Line 802:
local key = Wikitext:new(self:pop_sublayer())
local key = Wikitext:new(self:pop_sublayer())
self[-1].key = key
self[-1].key = key
self:push_sublayer(handle_parameter)
self:push_sublayer(handle_argument)
self[-1].pattern = "[\n<[{|}]"
self:set_pattern("[\n<[{|}]")
end,
end,
Line 429: Line 810:
["|"] = function(self)
["|"] = function(self)
emit_parameter(self)
emit_argument(self)
self:push_sublayer(handle_parameter)
self:push_sublayer(handle_argument)
self[-1].pattern = "[\n<=[{|}]"
self:set_pattern("[\n<=[{|}]")
end,
end,
["}"] = function(self)
["}"] = function(self)
if self:read(1) == "}" then
if self:read(1) == "}" then
emit_parameter(self)
emit_argument(self)
return self:pop()
return self:pop()
end
end
Line 445: Line 826:
[false] = Parser.emit
[false] = Parser.emit
})
})
return handle_parameter(self, ...)
return handle_argument(self, ...)
end
end
function Parser:do_template_or_argument()
function Parser:template_or_parameter()
self:push_sublayer(handle_name)
local text, head, node_to_emit, failed = self.text, self.head
self[-1].pattern = "[\n<[{|}]"
end
function Parser:template_or_argument()
local text, head, node_to_emit = self.text, self.head
-- Comments/tags interrupt the brace count.
-- Comments/tags interrupt the brace count.
local braces = match(text, "^{+()", head) - head
local braces = match(text, "^{+()", head) - head
self:advance(braces)
self:advance(braces)
repeat
while true do
local success, node = self:get("do_template_or_argument")
local success, node = self:try(do_template_or_parameter, node_to_emit)
-- Fail means no "}}" or "}}}" was found, so emit any remaining
-- unmatched opening braces before any templates/parameters that
-- were found.
if not success then
if not success then
self:emit(rep("{", braces))
self:emit(rep("{", braces))
failed = true
break
break
elseif node_to_emit then
-- If there are 3+ opening and closing braces, it's a parameter.
-- Nest the already-parsed node at the start of the new node.
elseif braces >= 3 and self:read(2) == "}" then
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)
self:advance(3)
braces = braces - 3
braces = braces - 3
node = Argument:new(node)
node = Parameter:new(node)
-- Otherwise, it's a template.
else
else
self:advance(2)
self:advance(2)
Line 480: Line 854:
node = Template:new(node)
node = Template:new(node)
end
end
local pos = head + braces
local index = head + braces
node.pos = pos
node.index = index
node.raw = sub(text, pos, self.head - 1)
node.raw = sub(text, index, self.head - 1)
node_to_emit = node
node_to_emit = node
if braces == 1 then
-- Terminate once not enough braces remain for further matches.
if braces == 0 then
break
-- Emit any stray opening brace before any matched nodes.
elseif braces == 1 then
self:emit("{")
self:emit("{")
break
break
end
end
until braces == 0
end
if node_to_emit then
if node_to_emit then
self:emit(node_to_emit)
self:emit(node_to_emit)
end
end
return braces
return braces, failed
end
end
end
end
Line 498: Line 876:
-- Tag.
-- Tag.
do
do
local tags = data.tags
local end_tags
local function get_end_tags()
end_tags, get_end_tags = (data or get_data()).end_tags, nil
return end_tags
end
-- Handlers.
-- Handlers.
local handle_start
local handle_start
local handle_tag
local handle_tag
local function do_tag(self)
local layer = self[-1]
layer.handler, layer.index = handle_start, self.head
self:set_pattern("[%s/>]")
self:advance()
end
local function is_ignored_tag(self, this)
local function is_ignored_tag(self, this)
Line 516: Line 906:
return self:fail_route()
return self:fail_route()
end
end
self.head = loc
self:jump(loc)
self[-1].ignored = true
local tag = self:pop()
return self:pop()
tag.ignored = true
return tag
end
end
function handle_start(self, this)
function handle_start(self, this)
if this == "/" then
if this == "/" then
Line 535: Line 926:
return self:fail_route()
return self:fail_route()
end
end
-- Tags are only case-insensitive with ASCII characters.
local raw_name = this
this = lower(this)
this = lower(this)
if not tags[this] then
local end_tag_pattern = (end_tags or get_end_tags())[this]
if not end_tag_pattern then -- Validity check.
return self:fail_route()
return self:fail_route()
end
end
Line 545: Line 939:
elseif is_ignored_tag(self, this) then
elseif is_ignored_tag(self, this) then
return ignored_tag(self, text, head)
return ignored_tag(self, text, head)
-- If an onlyinclude tag is not ignored (and cannot be active since it
-- would have triggered special handling earlier), it must be plaintext.
elseif this == "onlyinclude" then
return self:fail_route()
elseif this == "noinclude" or this == "includeonly" then
elseif this == "noinclude" or this == "includeonly" then
layer.ignored = true -- Ignored block.
layer.ignored = true -- Ignored block.
layer.raw_name = raw_name
end
end
layer.name, layer.handler, layer.pattern = this, handle_tag, ">"
layer.name, layer.handler, layer.end_tag_pattern = this, handle_tag, end_tag_pattern
self:set_pattern(">")
end
end
Line 562: Line 962:
end
end
local text, head, layer = self.text, self.head + 1, self[-1]
local text, head, layer = self.text, self.head + 1, self[-1]
local loc1, loc2 = find(text, tags[layer.name], head)
local loc1, loc2 = find(text, layer.end_tag_pattern, head)
if loc1 then
if loc1 then
if loc1 > head then
if loc1 > head then
self:emit(sub(text, head, loc1 - 1))
self:emit(sub(text, head, loc1 - 1))
end
end
self.head = loc2
self:jump(loc2)
return self:pop()
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 a simple array lookup with lowercase tag names when
-- looking up which tags should tolerate no closing tag (exact match
-- only, so case-sensitive).
elseif layer.ignored then
elseif layer.ignored then
self.head = #self.text
local raw_name = layer.raw_name
return self:pop()
if raw_name == "noinclude" or raw_name == "includeonly" then
self:jump(#text)
return self:pop()
end
end
end
return self:fail_route()
return self:fail_route()
end
function Parser:do_tag()
local layer = self[-1]
layer.handler, layer.pattern = handle_start, "[%s/>]"
self:advance()
end
local function find_next_chunk(text, pattern, head)
return select(2, find(text, pattern, head, true)) or #text
end
end
Line 589: Line 988:
-- HTML comment.
-- HTML comment.
if self:read(1, 3) == "!--" then
if self:read(1, 3) == "!--" then
self.head = find_next_chunk(self.text, "-->", self.head + 4)
local text = self.text
-- onlyinclude closing tag (whitespace intolerant).
self:jump(select(2, find(text, "-->", self.head + 4, true)) or #text)
-- onlyinclude tags (which must be lowercase with no whitespace).
elseif self.onlyinclude and self:read(1, 13) == "/onlyinclude>" then
elseif self.onlyinclude and self:read(1, 13) == "/onlyinclude>" then
self.head = find_next_chunk(self.text, "<onlyinclude>", self.head + 14)
local text = self.text
self:jump(select(2, find(text, "<onlyinclude>", self.head + 14, true)) or #text)
else
else
local success, tag = self:get("do_tag")
local success, tag = self:try(do_tag)
if not success then
if not success then
self:emit("<")
self:emit("<")
elseif not tag.ignored then
elseif not tag.ignored then
tag.end_tag_pattern = nil
self:emit(Tag:new(tag))
self:emit(Tag:new(tag))
end
end
Line 605: Line 1,007:


-- Heading.
-- 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, arguments 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.
-- 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
do
-- Handlers.
-- Handlers.
Line 611: Line 1,013:
local handle_body
local handle_body
local handle_possible_end
local handle_possible_end
local function do_heading(self)
local layer, head = self[-1], self.head
layer.handler, layer.index = 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
local function do_heading_possible_end(self)
local layer = self[-1]
layer.handler = handle_possible_end
self:set_pattern("[\n<]")
end
function handle_start(self, ...)
function handle_start(self, ...)
Line 630: Line 1,048:
local function whitespace(self)
local function whitespace(self)
local success, possible_end = self:get("do_heading_possible_end")
local success, possible_end = self:try(do_heading_possible_end)
if success then
if success then
self:emit(Wikitext:new(possible_end))
self:emit(Wikitext:new(possible_end))
local layer = self[-1]
local layer = self[-1]
layer.handler, layer.pattern = handle_body, "[\n<={]"
layer.handler = handle_body
self:set_pattern("[\n<=[{]")
return self:consume()
return self:consume()
end
end
Line 654: Line 1,073:
layer.level = 6
layer.level = 6
end
end
layer.handler, layer.pattern = handle_body, "[\n<=[{]"
layer.handler = handle_body
self:set_pattern("[\n<=[{]")
return self:consume()
return self:consume()
end
end
Line 671: Line 1,091:
local eq_len = #eq
local eq_len = #eq
self:advance(eq_len)
self:advance(eq_len)
local success, possible_end = self:get("do_heading_possible_end")
local success, possible_end = self:try(do_heading_possible_end)
if success then
if success then
self:emit(eq)
self:emit(eq)
Line 689: Line 1,109:
["["] = Parser.wikilink_block,
["["] = Parser.wikilink_block,
["{"] = Parser.braces,
["{"] = function(self, this)
return self:braces("{", true)
end,
[""] = Parser.fail_route,
[""] = Parser.fail_route,
[false] = Parser.emit
[false] = Parser.emit
Line 701: Line 1,125:
["<"] = function(self)
["<"] = function(self)
local head = (
if self:read(1, 3) ~= "!--" then
self:read(1, 3) == "!--" and
return self:pop()
select(2, find(self.text, "-->", self.head + 4, true))
end
)
local head = select(2, find(self.text, "-->", self.head + 4, true))
if not head then
if not head then
return self:pop()
return self:pop()
end
end
self.head = head
self:jump(head)
end,
end,
Line 714: Line 1,138:
[false] = function(self, this)
[false] = function(self, this)
if not match(this, "^[\t ]+$") then
if not match(this, "^[\t ]+()$") then
return self:pop()
return self:pop()
end
end
Line 721: Line 1,145:
})
})
return handle_possible_end(self, ...)
return handle_possible_end(self, ...)
end
function Parser:do_heading()
local layer, head = self[-1], self.head
layer.handler, layer.pattern, layer.pos = handle_start, "[\t\n ]", head
-- 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, layer.pattern = handle_possible_end, "[\n<]"
end
end
function Parser:heading()
function Parser:heading()
local success, heading = self:get("do_heading")
local success, heading = self:try(do_heading)
if success then
if success then
local section = self.section + 1
local section = self.section + 1
Line 759: Line 1,169:
-- Block handlers.
-- Block handlers.


-- These are blocks which can affect template/argument parsing, since they're also parsed by Parsoid at the same time (even though they aren't processed until later).
-- 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/arguments) 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.
-- 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.
-- Note that nesting also affects pipes and equals signs, in addition to block closures.
Line 770: Line 1,180:


-- Language conversion block.
-- Language conversion block.
-- Opens with "-{" and closes with "}-". However, templates/arguments take priority, so "-{{" is parsed as "-" followed by the opening of a template/argument block (depending on what comes after).
-- 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.
-- 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
do
local function handle_language_conversion_block(self, ...)
--Handler.
local handle_language_conversion_block
local function do_language_conversion_block(self)
local layer = self[-1]
layer.handler = handle_language_conversion_block
self:set_pattern("[\n<[{}]")
end
function handle_language_conversion_block(self, ...)
handle_language_conversion_block = self:switch(handle_language_conversion_block, {
handle_language_conversion_block = self:switch(handle_language_conversion_block, {
["\n"] = Parser.heading_block,
["\n"] = Parser.heading_block,
Line 795: Line 1,214:
end
end
function Parser:do_language_conversion_block()
function Parser:braces(this, fail_on_unclosed_braces)
local layer = self[-1]
layer.handler, layer.pattern = handle_language_conversion_block, "[\n<[{}]"
end
function Parser:braces()
local language_conversion_block = self:read(-1) == "-"
local language_conversion_block = self:read(-1) == "-"
if self:read(1) == "{" then
if self:read(1) == "{" then
local braces = self:template_or_argument()
local braces, failed = self:template_or_parameter()
if not (braces == 1 and language_conversion_block) then
-- Headings will fail if they contain an unclosed brace block.
if failed and fail_on_unclosed_braces then
return self:fail_route()
-- Language conversion blocks cannot begin "-{{", but can begin
-- "-{{{" iff parsed as "-{" + "{{".
elseif not (language_conversion_block and braces == 1) then
return self:consume()
return self:consume()
end
end
else
else
self:emit("{")
self:emit(this)
if not language_conversion_block then
if not language_conversion_block then
return
return
Line 814: Line 1,233:
self:advance()
self:advance()
end
end
self:emit(Wikitext:new(select(2, self:get("do_language_conversion_block"))))
self:emit(Wikitext:new(self:get(do_language_conversion_block)))
end
end
end
end
Line 825: Line 1,244:
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: 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 parameter with no previous equals signs, a newline followed by a single equals sign is parsed as a parameter equals sign, not the opening of a new L1 heading block. This does not apply to any other heading levels. As such, {{template|parameter\n=}}, {{template|key\n=value}} or even {{template|\n=}} will successfully close, but {{template|parameter\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.
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
do
local function handle_heading_block(self, ...)
--Handler.
local handle_heading_block
local function do_heading_block(self)
local layer = self[-1]
layer.handler = handle_heading_block
self:set_pattern("[\n<[{]")
end
function handle_heading_block(self, ...)
handle_heading_block = self:switch(handle_heading_block, {
handle_heading_block = self:switch(handle_heading_block, {
["\n"] = function(self)
["\n"] = function(self)
Line 844: Line 1,272:
end
end
function Parser:do_heading_block()
function Parser:heading_block(this, nxt)
local layer = self[-1]
layer.handler, layer.pattern = handle_heading_block, "[\n<[{]"
end
function Parser:heading_block()
self:newline()
self:newline()
while self:read(0, 1) == "\n=" do
this = this .. (nxt or "=")
local loc = #this - 1
while self:read(0, loc) == this do
self:advance()
self:advance()
self:emit(Wikitext:new(select(2, self:get("do_heading_block"))))
self:emit(Wikitext:new(self:get(do_heading_block)))
end
end
end
end
Line 861: Line 1,286:
-- Opens with "[[" and closes with "]]".
-- Opens with "[[" and closes with "]]".
do
do
local function handle_wikilink_block(self, ...)
-- Handler.
local handle_wikilink_block
local function do_wikilink_block(self)
local layer = self[-1]
layer.handler = handle_wikilink_block
self:set_pattern("[\n<[%]{]")
end
function handle_wikilink_block(self, ...)
handle_wikilink_block = self:switch(handle_wikilink_block, {
handle_wikilink_block = self:switch(handle_wikilink_block, {
["\n"] = Parser.heading_block,
["\n"] = Parser.heading_block,
Line 881: Line 1,315:
})
})
return handle_wikilink_block(self, ...)
return handle_wikilink_block(self, ...)
end
function Parser:do_wikilink_block()
local layer = self[-1]
layer.handler, layer.pattern = handle_wikilink_block, "[\n<[%]{]"
end
end
Line 892: Line 1,321:
self:emit("[[")
self:emit("[[")
self:advance(2)
self:advance(2)
self:emit(Wikitext:new(select(2, self:get("do_wikilink_block"))))
self:emit(Wikitext:new(self:get(do_wikilink_block)))
else
else
self:emit("[")
self:emit("[")
Line 920: Line 1,349:
break
break
end
end
self.head = head
self:jump(head)
end
end
self:emit("\n")
self:emit("\n")
Line 929: Line 1,358:
local handle_start
local handle_start
local main_handler
local main_handler
-- 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.
local function do_parse(self, 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>", nil, true) then
local head = find(text, "<onlyinclude>", nil, true)
if head then
self.onlyinclude = true
self:jump(head + 13)
end
end
end
-- If the first character is "=", try parsing it as a heading.
-- If the first character is "=", try parsing it as a heading.
function handle_start(self, this)
function handle_start(self, this)
local layer = self[-1]
local layer = self[-1]
layer.handler, layer.pattern = main_handler, "[\n<{]"
layer.handler = main_handler
self:set_pattern("[\n<{]")
if this == "=" then
if this == "=" then
return self:heading()
return self:heading()
Line 954: Line 1,413:
["{"] = function(self)
["{"] = function(self)
if self:read(1) == "{" then
if self:read(1) == "{" then
self:template_or_argument()
self:template_or_parameter()
return self:consume()
return self:consume()
end
end
Line 964: Line 1,423:
})
})
return main_handler(self, ...)
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, layer.pattern = handle_start, "."
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
end
Line 998: Line 1,429:
return (select(2, Parser:parse{
return (select(2, Parser:parse{
text = text_type == "string" and text or
text = text_type == "string" and text or
text_type == "number" and format("%.14g", text) or
text_type == "number" and tostring(text) or
error("bad argument #1 (string expected, got " .. text_type .. ")"),
error("bad argument #1 (string expected, got " .. text_type .. ")"),
node = {Wikitext, true},
node = {Wikitext, true},
route = {"do_parse", transcluded}
route = {do_parse, transcluded}
}))
}))
end
end
Line 1,008: Line 1,439:


do
do
local page_title = mw.title.getCurrentTitle()
local function is_template(v)
local namespace_has_subpages = mw.site.namespaces[page_title.namespace].hasSubpages
return class_else_type(v) == "template"
local raw_pagename = page_title.fullText
end
function export.find_templates(text, not_transcluded)
return parse(text, not not_transcluded):iterate(is_template)
end
end
 
do
local link_parameter_1, link_parameter_2
local parser_functions_s = data.parser_functions_case_sensitive
local function get_link_parameter_1()
local parser_functions_i = data.parser_functions_case_insensitive
link_parameter_1, get_link_parameter_1 = (data or get_data()).template_link_param_1, nil
local parser_variables_s = data.parser_variables_case_sensitive
return link_parameter_1
local parser_variables_i = data.parser_variables_case_insensitive
end
local transclusion_modifiers = data.transclusion_modifiers
-- Mainspace titles starting with "#" should be invalid, but a bug in
local function get_link_parameter_2()
-- mw.title.new means a title object is returned that has the empty string
link_parameter_2, get_link_parameter_2 = (data or get_data()).template_link_param_2, nil
-- for prefixedText, so we need to filter them out. Interwiki links aren't
return link_parameter_2
-- valid as templates, either.
local function is_valid_title(title)
return title and
#title.prefixedText > 0 and
#title.interwiki == 0
end
end
function export.parseTemplateName(name, has_args, fragment)
-- Generate a link. If the target title doesn't have a fragment, use "#top"
local chunks, colon, start, n, p = {}, find(name, ":", 1, true), 1, 0, 0
-- (which is an implicit anchor at the top of every page), as this ensures
while colon do
-- self-links still display as links, since bold display is distracting and
-- Pattern is a PHP left-trim.
-- unintuitive for template links.
local chunk = match(sub(name, start, colon), "[^%z\t-\v\r ].*") or ""
local function link_page(title, display)
local normalized = upper(chunk)
local fragment = title.fragment
local priority = transclusion_modifiers[normalized]
if fragment == "" then
if not (priority and priority > p) then
fragment = "top"
local pf = parser_functions_s[chunk] or parser_functions_i[normalized]
if pf then
n = n + 1
chunks[n] = normalized
return chunks, sub(name, colon + 1)
end
break
end
n = n + 1
chunks[n] = normalized
start, p = colon + 1, priority
colon = find(name, ":", start, true)
end
end
if start > 1 then
return format(
name = sub(name, start)
"[[:%s|%s]]",
encode_uri(title.prefixedText .. "#" .. fragment, "WIKI"),
display
)
end
-- pf_arg1 or pf_arg2 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 namespace == nil then
return pagename
end
end
name = php_trim(name)
local title = make_title(namespace, pagename)
-- Parser variables can only take SUBST:/SAFESUBST: as modifiers.
return title and is_internal_title(title) and link_page(title, pagename) or pagename
if not has_args and p <= 1 then
end
local pv = parser_variables_s[name] or parser_variables_i[upper(name)]
if pv then
local function render_title(name, args)
n = n + 1
-- parse_template_name returns a table of transclusion modifiers plus
chunks[n] = pv
-- the normalized template/magic word name, which will be used as link
return chunks
-- targets. The third return value pf_arg1 is the first argument of a
end
-- a parser function, which comes after the colon (e.g. "foo" in
end
-- "{{#IF:foo|bar|baz}}"). This means args[1] (i.e. the first argument
-- Handle relative template names.
-- that comes after a pipe is actually argument 2, and so on. Note: the
if namespace_has_subpages then
-- second parameter of parse_template_name checks if there are any
-- If the name starts with "/", it's treated as a subpage of the
-- arguments, since parser variables cannot take arguments (e.g.
-- current page. Final slashes are trimmed, but this can't affect
-- {{CURRENTYEAR}} is a parser variable, but {{CURRENTYEAR|foo}}
-- the intervening slash (e.g. {{///}} refers to "{{PAGENAME}}/").
-- transcludes "Template:CURRENTYEAR"). In such cases, the returned
if sub(name, 1, 1) == "/" then
-- table explicitly includes the "Template:" prefix in the template
name = raw_pagename .. (match(name, "^/.*[^/]") or "/")
-- name. The third parameter instructs it to retain any fragment in the
-- If it starts with "../", trim it and any that follow, and go up
-- template name in the returned table, if present.
-- that many subpage levels. Then, treat any additional text as
local chunks, subclass, pf_arg1 = parse_template_name(
-- a subpage of that page; final slashes are trimmed.
name,
elseif match(name, "^()%.%./") then
args and pairs(args)(args) ~= nil,
local n = 4
true
while sub(name, n, n + 2) == "../" do
)
n = n + 3
if chunks == nil then
end
return name, args
-- 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
end
local title = new_title(name, 10)
local chunks_len = #chunks
if not is_valid_title(title) then
-- Additionally, generate the corresponding table `rawchunks`, which
return nil
-- 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 or get_magic_words())[sub(chunks[i], 1, -2)].transclusion_modifier, "WIKI"),
rawchunks[i]
)
end
end
-- If `fragment` is set, save the original title's fragment, since it
local chunk = chunks[chunks_len]
-- won't carry through to any redirect targets.
-- If it's a template, return a link to it with link_page, concatenating
if fragment then
-- the remaining chunks in `rawchunks` to form the display text.
fragment = title.fragment
-- 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 subclass == "template" then
chunks[chunks_len] = link_page(
new_title(chunk, 10),
concat(rawchunks, "&#58;", chunks_len) -- :
)
return concat(chunks, "&#58;"), args -- :
elseif subclass == "parser variable" then
chunks[chunks_len] = format(
"[[%s|%s]]",
encode_uri((magic_words or get_magic_words())[chunk].parser_variable, "WIKI"),
rawchunks[chunks_len]
)
return concat(chunks, "&#58;"), args -- :
end
end
-- Resolve any redirects. Unlike links, double redirects seem to work
-- Otherwise, it must be a parser function.
-- (but not higher). If the redirect target is an interwiki link, then
local mgw_data = (magic_words or get_magic_words())[sub(chunk, 1, -2)]
-- the template won't fail, but the redirect page itself gets
local link = mgw_data.parser_function or mgw_data.transclusion_modifier
-- transcluded (i.e. the template name shouldn't be normalized to the
local pf_arg2 = args and args[1] or nil
-- target).
-- Some magic words have different links, depending on whether argument
for _ = 1, 2 do
-- 2 is specified (e.g. "baz" in {{foo:bar|baz}}).
local redirect = title.redirectTarget
if type(link) == "table" then
if not is_valid_title(redirect) then
link = pf_arg2 and link[2] or link[1]
break
end
title = redirect
end
end
local namespace, chunk = title.namespace
chunks[chunks_len] = format(
-- Normalize name.
"[[%s|%s]]",
if namespace == 10 then
encode_uri(link, "WIKI"),
-- If in the template namespace and it shares a name with a magic
rawchunks[chunks_len]
-- word, it needs the prefix "Template:".
)
local text = title.text
-- #TAG: has special handling, because documentation links for parser
if parser_functions_s[text] or parser_variables_s[text] then
-- extension tags come from [[Module:data/parser extension tags]].
chunk = title.prefixedText
if chunk == "#TAG:" then
else
-- Tags are only case-insensitive with ASCII characters.
local normalized = upper(text)
local tag = (parser_extension_tags or get_parser_extension_tags())[lower(php_trim(pf_arg1))]
if (
if tag then
parser_functions_i[normalized] or
pf_arg1 = format(
parser_variables_i[normalized] or
"[[%s|%s]]",
transclusion_modifiers[normalized]
encode_uri(tag, "WIKI"),
) then
pf_arg1
chunk = title.prefixedText
)
elseif not find(text, ":", 1, true) then
chunk = text
-- Also if the prefix is necessary for disambiguation (e.g.
-- "Template:Category:Foo" can't be abbreviated to
-- "Category:Foo").
else
local check = new_title(text, 10)
chunk = check and title_equals(title, check) and text or title.prefixedText
end
end
end
-- If not in the template namespace, include the prefix (or ":" if
-- Otherwise, finalize pf_arg1 and add it to `chunks`.
-- mainspace).
else
else
chunk = namespace == 0 and ":" .. title.text or title.prefixedText
pf_arg1 = finalize_arg(pf_arg1, (link_parameter_1 or get_link_parameter_1())[chunk])
end
end
-- Set the fragment (if applicable).
chunks[chunks_len + 1] = pf_arg1
if fragment then
-- Finalize pf_arg2 (if applicable), then return.
chunk = chunk .. "#" .. fragment
if pf_arg2 then
args = shallow_copy(args) -- Avoid destructively modifying args.
args[1] = finalize_arg(pf_arg2, (link_parameter_2 or get_link_parameter_2())[chunk])
end
end
chunks[n + 1] = chunk
return concat(chunks, "&#58;"), args -- :
return chunks
end
end
parse_template_name = export.parseTemplateName
end
function export.buildTemplate(title, args)
 
local output = {title}
function export.parseTemplate(text, not_transcluded)
if not args then
text = parse(text, not not_transcluded)
return output
if type_or_class(text) == "template" then
local name = text:get_name()
if name then
return name, text:get_params()
end
end
end
-- Iterate over all numbered parameters in order, followed by any
return nil, nil
-- remaining parameters in codepoint order. Implicit parameters are
end
-- used wherever possible, even if explicit numbers are interpolated
 
-- between them (e.g. 0 would go before any implicit parameters, and
do
-- 2.5 between 2 and 3).
local function next_template(iter)
-- TODO: handle "=" and "|" in params/values.
while true do repeat -- break acts like continue
local implicit
local node = iter()
for k, v in sorted_pairs(args) do
if not node then
if type(k) == "number" and k >= 1 and k % 1 == 0 then
return nil, nil, nil, nil
if implicit == nil then
elseif type_or_class(node) ~= "template" then
implicit = table_len(args)
break
end
insert(output, k <= implicit and v or k .. "=" .. v)
else
insert(output, k .. "=" .. v)
end
end
local name = node:get_name()
end
if name then
return output
return name, node:get_params(), node.raw, node.pos
end
until true end
end
end
build_template = export.buildTemplate
function export.findTemplates(text, not_transcluded)
function export.templateLink(title, args, no_link)
return next_template, parse(text, not not_transcluded):__pairs("next_node")
if not no_link then
title, args = render_title(title, args)
end
local output = build_template(title, 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")
:wikitext("&#123;&#123;" .. concat(output, "&#124;") .. "&#125;&#125;") -- {{ | }}
)
end
end
end
end


do
do
local function next_argument(iter)
local function is_parameter(v)
while true do
return class_else_type(v) == "parameter"
local node = iter()
end
if not node then
return nil, nil, nil, nil
function export.find_parameters(text, not_transcluded)
elseif type_or_class(node) == "argument" then
return parse(text, not not_transcluded):iterate(is_parameter)
local args = iter.args
return node:get_name(args), node:get_default(args), node.raw, node.pos
end
end
end
end
function export.findArguments(text, args, not_transcluded)
function export.displayParameter(name, default)
local iter = parse(text, not not_transcluded):__pairs("next_node")
return tostring(html_create("code")
iter.args = args
:css("white-space", "pre-wrap")
return next_argument, iter
:wikitext("&#123;&#123;&#123;" .. concat({name, default}, "&#124;") .. "&#125;&#125;&#125;") -- {{{ | }}}
)
end
end
end
end
Line 1,218: Line 1,655:
end
end
local function next_heading(iter)
local function is_heading(v)
while true do repeat -- break acts like continue
return class_else_type(v) == "heading"
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
end
-- Note: heading names can contain "\n" (e.g. inside nowiki tags), which
-- FIXME: should headings which contain "\n" be returned? This may depend
-- causes any heading containing them to fail. When that happens, the
-- on variable factors, like template expansion. They iterate the heading
-- heading is not returned by this function, but the heading count is still
-- count number, but fail on rendering. However, in some cases a different
-- iterated, since Parsoid's preprocessor still counts it as a heading for
-- heading might still be rendered due to intermediate equals signs; it
-- the purpose of heading strip markers (i.e. the section number).
-- may even be of a different heading level: e.g., this is parsed as an
-- L2 heading with a newline (due to the wikilink block), but renders as the
-- L1 heading "=foo[[". Section edit links are sometimes (but not always)
-- present in such cases.
-- ==[[=
-- ]]==
-- TODO: section numbers for edit links seem to also include headings
-- TODO: section numbers for edit links seem to also include headings
-- nested inside templates and arguments (but apparently not those in
-- nested inside templates and parameters (but apparently not those in
-- parser extension tags - need to test this more). If we ever want to add
-- 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.
-- section edit links manually, this will need to be accounted for.
function export.findHeadings(text, i, j)
function export.find_headings(text, i, j)
local iter = parse(text):__pairs("next_node")
local parsed = parse(text)
iter.i, iter.j = i and check_level(i) or 1, j and check_level(j) or 6
if i == nil and j == nil then
return next_heading, iter
return parsed:iterate(is_heading)
end
i = i and check_level(i) or 1
j = j and check_level(j) or 6
return parsed:iterate(function(v)
if class_else_type(v) == "heading" then
local level = v.level
return level >= i and level <= j
end
end)
end
end
end
end
do
local function make_tag(tag)
return tostring(html_create("code")
:css("white-space", "pre-wrap")
:wikitext("&lt;" .. tag .. "&gt;")
)
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, ">", nil, true) 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 (e.g. "<nowiki/>", but not "<nowiki/ >").
remainder ~= "/" and sub(remainder, 1, 1) == "/"
) then
-- Output with no link.
return make_tag(tag)
end
-- Partial transclusion tags aren't in the table of parser extension
-- tags.
if link == "noinclude" or link == "includeonly" or link == "onlyinclude" then
link = "mw:Transclusion#Partial transclusion"
else
link = (parser_extension_tags or get_parser_extension_tags())[link]
end
if link then
tag = gsub(tag, pattern_escape(tagname), "[[" .. replacement_escape(encode_uri(link, "WIKI")) .. "|%0]]", 1)
end
return make_tag(tag)
end
end
-- For convenience.
export.class_else_type = class_else_type


return export
return export
Retrieved from "https://linguifex.com/wiki/Special:MobileDiff/410280"

Navigation menu