require('strict');
local getArgs = require ('Module:Arguments').getArgs;


--[[--------------------------< A R G S _ D E F A U L T >------------------------------------------------------

a table to specify initial values.

]]

local args_default = {
	bracket_left = '',
	bracket_right = '',
	bracket_year_left = '',
	bracket_year_right = '',
	postscript = '',
	page = '',
	pages = '',
	location = '',
	page_sep = ", p.&nbsp;",
	pages_sep = ", pp.&nbsp;",
	ref = '',
	template = 'harv',															-- if template name not provided in ((#invoke:)) use this
	};


--[[--------------------------< T A R G E T _ C H E C K >------------------------------------------------------

look for anchor_id (CITEREF name-list and year or text from |ref=) in anchor_id_list

the 'no target' error may be suppressed with |ignore-err=yes when target cannot be found because target is inside
a template that wraps another template; 'multiple targets' error may not be suppressed

]]

local function target_check (anchor_id, args)
	local namespace = mw.title.getCurrentTitle().namespace;
	local anchor_id_list_module = mw.loadData ('Module:Footnotes/anchor_id_list');
	local anchor_id_list = anchor_id_list_module.anchor_id_list;
	local article_whitelist = anchor_id_list_module.article_whitelist;
	local template_list = anchor_id_list_module.template_list;
	
	local whitelist_module = mw.loadData ('Module:Footnotes/whitelist');
	local whitelist = whitelist_module.whitelist;
	local special_patterns = whitelist_module.special_patterns;
	local DNB_special_patterns = whitelist_module.DNB_special_patterns;
	local DNB_template_names = whitelist_module.DNB_template_names;

	if 10 == namespace then
		return '';																-- automatic form of |no-tracking=yes; TODO: is this too broad?
	end

	local tally = anchor_id_list[anchor_id];									-- nil when anchor_id not in list; else a tally
	local msg;
	local category;

	if not tally then
		if args.ignore then
			return '';															-- if ignore is true then no message, no category
		end
		
		if article_whitelist and article_whitelist[anchor_id] then				-- if an article-local whitelist and anchor ID is in it
			return '';															-- done
		end
		
		local wl_anchor_id = anchor_id;											-- copy to be modified to index into the whitelist
		
		if args.year then														-- for anchor IDs created by this template (not in |ref=) that have a date
			if args.year:match ('%d%l$') or										-- use the date value to determine if we should remove the disambiguator
				args.year:match ('n%.d%.%l$') or
				args.year:match ('nd%l$') then
					wl_anchor_id = wl_anchor_id:gsub ('%l$', '');				-- remove the disambiguator
			end
		end		

		local t_tbl = whitelist[wl_anchor_id];									-- get list of templates associated with this anchor ID

		if t_tbl then															-- when anchor ID not whitelisted t_tbl is nil
			for _, t in ipairs (t_tbl) do										-- spin through the list of templates associated with this anchor ID
				if template_list[t] then										-- if associated template is found in the list of templates in the article
					return '';													-- anchor ID is whitlisted and article has matching template so no error
				end
			end
		end

		for _, pattern in ipairs (special_patterns) do							-- spin through the spcial patterns and try to match
			if anchor_id:match (pattern) then
				return '';
			end
		end

		for _, dnb_t in ipairs (DNB_template_names or {}) do					-- getting desparate now, are there any DNB templates? DNB_template_names may be nil; empty table prevents script error
			if template_list[dnb_t] then										-- if the article has this DNB template
				for _, pattern in ipairs (DNB_special_patterns) do				-- spin through the DNB-specifiec wildcard patterns
					if anchor_id:match (pattern) then							-- and attempt a match
						return '';												-- found a match
					end
				end
			end
		end

		msg = 'no target: ' .. anchor_id;										-- anchor_id not found
		category = '[[Category:Harv and Sfn no-target errors]]';

	elseif 1 < tally then
		msg = 'multiple targets (' .. tally .. '×): ' .. anchor_id;				-- more than one anchor_id in this article
		category = 0 == namespace and '[[Category:Harv and Sfn multiple-target errors]]' or '';								-- only categorize in article space
		return '<span class="error harv-error" style="display: inline; font-size:100%"> ' .. args.template .. ' error: ' .. msg .. ' ([[:Category:Harv and Sfn template errors|help]])</span>' .. category;
	end

--	category = 0 == namespace and '[[Category:Harv and Sfn template errors]]' or '';	-- only categorize in article space
	category = 0 == namespace and category or '';								-- only categorize in article space

--use this version to show error messages
--	return msg and '<span class="error harv-error" style="display: inline; font-size:100%"> ' .. args.template .. ' error: ' .. msg .. ' ([[:Category:Harv and Sfn template errors|help]])</span>' .. category or '';
--use this version to hide error messages
	return msg and '<span class="error harv-error" style="display: none; font-size:100%"> ' .. args.template .. ' error: ' .. msg .. ' ([[:Category:Harv and Sfn template errors|help]])</span>' .. category or '';

end


--[[--------------------------< I S _ Y E A R >----------------------------------------------------------------

evaluates param to see if it is one of these forms with or without lowercase letter disambiguator:
	YYYY
	n.d.
	nd	
	c. YYYY
	YYYY–YYYY	(separator is endash)
	YYYY–YY		(separator is endash)

return true when param has a recognized form; false else

]]

local patterns_date= {
	'^%d%d%d%d?%l?$',
	'^n%.d%.%l?$',
	'^nd%l?$',
	'^c%. %d%d%d%d?%l?$',
	'^%d%d%d%d–%d%d%d%d%l?$',
	'^%d%d%d%d–%d%d%l?$',
	}

local function is_year (param, args)
	args.year = '';																-- used for harv error; 
	
	for _, pattern in ipairs (patterns_date) do
		if mw.ustring.match (param, pattern) then
			args.year = param;													-- used for harv error; 
			return true;
		end
	end
end


--[[--------------------------< C O R E >----------------------------------------------------------------------

returns an anchor link (CITEREF) formed from one to four author names, year, and insource location (|p=, |pp=, loc=)

]]

local function core( args )
	local result;
	local err_msg = ''

	if args.P5 ~= '' then
		if is_year (args.P5, args) then
			result = table.concat ({args.P1, ' et al. ', args.bracket_year_left, args.P5, args.bracket_year_right});
		else
			args.P5 = '';														-- when P5 not a year don't include in anchor
			result = table.concat ({args.P1, ' et al.'});						-- and don't render it
		end

	elseif args.P4 ~= '' then
		if is_year (args.P4, args) then
			result = table.concat ({args.P1, ', ', args.P2, ' &amp; ', args.P3, ' ', args.bracket_year_left, args.P4, args.bracket_year_right});	-- three names and a year
		else
			result = table.concat ({args.P1, ' et al.'});						-- four names
		end

	elseif args.P3 ~= '' then
		if is_year (args.P3, args) then
			result = table.concat ({args.P1, ' &amp; ', args.P2, ' ', args.bracket_year_left, args.P3, args.bracket_year_right});	-- two names and a year
		else
			result = table.concat ({args.P1, ', ', args.P2, ' ', ' &amp; ', args.P3});	-- three names
		end
			
	elseif args.P2 ~= '' then
		if is_year (args.P2, args) then
			result = table.concat ({args.P1, ' ', args.bracket_year_left, args.P2, args.bracket_year_right});	-- one name and year
		else
			result = table.concat ({args.P1, ' &amp; ', args.P2});				-- two names
		end
		
	else
		result = args.P1;														-- one name
	end
																				-- when author-date result ends with a dot (typically when the last positional parameter holds 'n.d.')
																				-- and when no in-source location (no |p=, |pp=, or |loc=)
																				-- and when the first or only character in args.postscript is a dot
																				-- remove the author-date result trailing dot
																				-- the author-date result trailing dot will be replaced later with the content of args.postscript (usually a dot)
	if ('.' == result:sub(-1)) and ('.' == args.postscript:sub(1)) and ('' == args.page) and ('' == args.pages) and ('' == args.location) then
		result = result:gsub ('%.$', '');
	end
	
	if args.ref ~= 'none' then
		local anchor_id;
		if args.ref ~= '' then
			anchor_id = mw.uri.anchorEncode (args.ref);
			err_msg = target_check (anchor_id, args);
			result = table.concat ({'[[#', anchor_id, '|', result, ']]'});
		else
			anchor_id = mw.uri.anchorEncode (table.concat ({'CITEREF', args.P1, args.P2, args.P3, args.P4, args.P5}));
			err_msg = target_check (anchor_id, args);
			result = table.concat ({'[[#', anchor_id, '|', result, ']]'});
		end
	end

	if args.page ~= '' then
		result = table.concat ({result, args.page_sep, args.page});
	elseif args.pages ~= ''then
		result = table.concat ({result, args.pages_sep, args.pages});
	end      

	if args.location ~= '' then
		result = table.concat ({result, ', ', args.location});
	end

	result = table.concat ({args.bracket_left, result, args.bracket_right, args.postscript}):gsub ('%s+', ' ');		-- strip redundant spaces
	return result .. err_msg;
end


--[[--------------------------< H Y P H E N _ T O _ D A S H >--------------------------------------------------

Converts a hyphen to a dash under certain conditions.  The hyphen must separate
like items; unlike items are returned unmodified.  These forms are modified:
	letter - letter (A - B)
	digit - digit (4-5)
	digit separator digit - digit separator digit (4.1-4.5 or 4-1-4-5)
	letterdigit - letterdigit (A1-A5) (an optional separator between letter and
		digit is supported – a.1-a.5 or a-1-a-5)
	digitletter - digitletter (5a - 5d) (an optional separator between letter and
		digit is supported – 5.a-5.d or 5-a-5-d)

any other forms are returned unmodified.

str may be a comma- or semicolon-separated list

This code copied from Module:Citation/CS1.  The only modification is to require Module:Citation/CS1/Utilities
so that it has access to the functions is_set() and has_accept_as_written()

]]

local function hyphen_to_dash( str )
	local utilities = require ('Module:Citation/CS1/Utilities');				-- only modification so that this function has access to is_set() and has_accept_as_written()

	if not utilities.is_set (str) then
		return str;
	end

	local accept; -- Boolean

	str = str:gsub ('&[nm]dash;', {['&ndash;'] = '–', ['&mdash;'] = '—'});		-- replace &mdash; and &ndash; entities with their characters; semicolon mucks up the text.split
	str = str:gsub ('&#45;', '-'); -- replace HTML numeric entity with hyphen character

	str = str:gsub ('&nbsp;', ' '); -- replace &nbsp; entity with generic keyboard space character
	
	local out = {};
	local list = mw.text.split (str, '%s*[,;]%s*');								-- split str at comma or semicolon separators if there are any

	for _, item in ipairs (list) do												-- for each item in the list
		item, accept = utilities.has_accept_as_written (item);					-- remove accept-this-as-written markup when it wraps all of item
		if not accept and mw.ustring.match (item, '^%w*[%.%-]?%w+%s*[%-–—]%s*%w*[%.%-]?%w+$') then	-- if a hyphenated range or has endash or emdash separators
			if item:match ('^%a+[%.%-]?%d+%s*%-%s*%a+[%.%-]?%d+$') or			-- letterdigit hyphen letterdigit (optional separator between letter and digit)
				item:match ('^%d+[%.%-]?%a+%s*%-%s*%d+[%.%-]?%a+$') or			-- digitletter hyphen digitletter (optional separator between digit and letter)
				item:match ('^%d+[%.%-]%d+%s*%-%s*%d+[%.%-]%d+$') or			-- digit separator digit hyphen digit separator digit
				item:match ('^%d+%s*%-%s*%d+$') or								-- digit hyphen digit
				item:match ('^%a+%s*%-%s*%a+$') then							-- letter hyphen letter
					item = item:gsub ('(%w*[%.%-]?%w+)%s*%-%s*(%w*[%.%-]?%w+)', '%1–%2');	-- replace hyphen, remove extraneous space characters
			else
				item = mw.ustring.gsub (item, '%s*[–—]%s*', '–');				-- for endash or emdash separated ranges, replace em with en, remove extraneous whitespace
			end
		end
		table.insert (out, item);												-- add the (possibly modified) item to the output table
	end

	local temp_str = '';														-- concatenate the output table into a comma separated string
	temp_str, accept = utilities.has_accept_as_written (table.concat (out, ', ')); -- remove accept-this-as-written markup when it wraps all of concatenated out
	if accept then
		temp_str = utilities.has_accept_as_written (str);						-- when global markup removed, return original str; do it this way to suppress boolean second return value
		return temp_str;
	else
		return temp_str;														-- else, return assembled temp_str
	end
end


--[[--------------------------< A R G S  _ F E T C H >---------------------------------------------------------

Because all of the templates share a common set of parameters, a single common function to fetch those parameters
from frame and parent frame.

]]

local function args_fetch (frame, ps)
	local args = args_default;													-- create a copy of the default table
	local pframe = frame:getParent();											-- point to the template's parameter table

	for k, v in pairs (frame.args) do											-- override defaults with values provided in the #invoke: if any
		args[k] = v;	   
	end
	
	args.postscript = pframe.args.postscript or pframe.args.ps or ps;
	if 'none' == args.postscript then
		args.postscript = '';
	end
	args.page = pframe.args.p or pframe.args.page or '';
	args.pages = pframe.args.pp or pframe.args.pages or '';
	args.pages = ('' ~= args.pages) and hyphen_to_dash (args.pages) or '';
	args.location = pframe.args.loc or '';
	args.ref = pframe.args.ref or pframe.args.Ref or '';
	args.ignore = ('yes' == pframe.args['ignore-false-positive']) or ('yes' == pframe.args['ignore-err']);

	for i, v in ipairs ({'P1', 'P2', 'P3', 'P4', 'P5'}) do						-- loop through the five positional parameters and trim if set else empty string
		args[v] = (pframe.args[i] and mw.text.trim (pframe.args[i])) or '';
	end

	if args.P5 and not is_year (args.P5, args) then
		local i = 6;															-- initialize the indexer to the sixth positional parameter
		while pframe.args[i] do													-- in case there are too many authors loop through the authors looking for a year
			local v = mw.text.trim (pframe.args[i]);							-- trim
			if is_year (v, args) then											-- if a year
				args.P5 = v;													-- overwrite whatever was in args.P5 with year
				break;															-- and abandon the search
			end
			i = i + 1;															-- bump the indexer
		end
	end
	return args;
end


--[[--------------------------< H A R V A R D _ C I T A T I O N >----------------------------------------------

common entry point for:
	((harvard citation)) aka ((harv))
	((Harvard citation no brackets)) aka ((harvnb))
	((harvcol))
	((harvcolnb))
	((harvcoltxt))
	((Harvard citation text)) aka ((harvtxt))
	((Harvp))

Distinguishing features (brackets and page separators) are specified in this module's ((#invoke)) in the respective templates.

]]

local function harvard_citation (frame)
	local args = args_fetch (frame, '');										-- get the template and invoke parameters; default postscript is empty string

	return core (args);
end


--[[--------------------------< S T R I P _ U R L >------------------------------------------------------------

used by sfn() and sfnm().  This function fixes an issue with reference tooltip gadget where the tooltip is not displayed
when an insource locator (|p=, |pp=, |loc=) has an external wikilink that contains a # character

strip uri-reserved characters from urls in |p=, |pp-, and |loc= parameters  The researved characters are:
	!#$&'()*+,/:;=?@[]
	
]]

local function strip_url (pages)
	local escaped_uri;
	if not pages or ('' == pages) then
		return pages;
	end
	
	for uri in pages:gmatch ('%[(%a[%w%+%.%-]*://%S+)') do						-- for each external link get the uri
		escaped_uri = uri:gsub ("([%(%)%.%%%+%-%*%?%[%^%$%]])", "%%%1" );		-- save a copy with lua pattern characters escaped
		uri = uri:gsub ("[!#%$&'%(%)%*%+,/:;=%?@%[%]%.%%]", '');				-- remove reserved characters and '%' because '%20' (space character) is a lua 'invalid capture index'
		pages = pages:gsub (escaped_uri, uri, 1);								-- replace original uri with the stripped version
	end
	
	return pages;
end


--[[--------------------------< S F N >------------------------------------------------------------------------

entry point for ((sfn)) and ((sfnp))

]]

local function sfn (frame)
	local args = args_fetch (frame, '.');										-- get the template and invoke parameters; default postscript is a dot

	local result = core (args);													-- go make a CITEREF anchor
																				-- put it all together and then strip redundant spaces
	local name = table.concat ({'FOOTNOTE', args.P1, args.P2, args.P3, args.P4, args.P5, strip_url (args.page), strip_url (args.pages), strip_url (args.location)}):gsub ('%s+', ' ');

	return frame:extensionTag ({name='ref', args={name=name}, content=result});	

	
end


--[[--------------------------< S F N M >----------------------------------------------------------------------

common entry point for ((sfnm)) and ((sfnmp))

Distinguishing features (brackets) are specified in this module's ((#invoke)) in the respective templates.

]]

local function sfnm (frame)
	local args = args_default;													-- create a copy of the default table
	local pframe = frame:getParent();											-- point to the template's parameter table
	
	local n = 1;																-- index of source; this is the 'n' in na1, ny, etc
	local first_pnum = 1;														-- first of a pair of positional parameters
	local second_pnum = 2;														-- second of a pair of positional parameters

	local last_ps = 0;															-- index of the last source with |nps= set
	local last_index = 0;														-- index of the last source; these used to determine which of |ps= or |nps= will terminate the whole rendering

	local out = {};																-- table to hold rendered sources
	local footnote = {'FOOTNOTE'};												-- all author, date, insource location stuff becomes part of the reference's footnote id; added as we go

	for k, v in pairs (frame.args) do											-- override defaults with values provided in the #invoke: if any
		args[k] = v;	   
	end
	
	while true do
		if not pframe.args[table.concat ({n, 'a1'})] and not pframe.args[first_pnum] then
			break;																-- no na1 or matching positional parameter so done
		end
		
		if pframe.args[table.concat ({n, 'a1'})] then							-- does this source use named parameters?
			for _, v in ipairs ({'P1', 'P2', 'P3', 'P4', 'P5'}) do				-- initialize for this source
				args[v] = '';
			end

			for i, v in ipairs ({'P1', 'P2', 'P3', 'P4', 'P5'}) do				-- extract author and year parameters for this source
				args[v] = pframe.args[table.concat ({n, 'a', i})] or '';		-- attempt to assign author name
				if '' == args[v] then											-- when there wasn't an author name
					args[v] = pframe.args[table.concat ({n, 'y'})] or '';		-- attempt to assign year
					break;														-- done with author/date for this source
				end
			end

		else																	-- this source uses positional parameters
			args.P1 = mw.text.trim (pframe.args[first_pnum]);					-- yes, only one author supported
			args.P2 = (pframe.args[second_pnum] and mw.text.trim (pframe.args[second_pnum])) or '';	-- when positional author, year must also be positional

			for _, v in ipairs ({'P3', 'P4', 'P5'}) do							-- blank the rest of these for this source
				args[v] = '';
			end

			first_pnum = first_pnum + 2;										-- source must use positional author and positional year
			second_pnum = first_pnum + 1;										-- bump these for possible next positional source
		end
		
		args.postscript = pframe.args[table.concat ({n, 'ps'})] or '';
		if 'none' == args.postscript then										-- this for compatibility with other footnote templates; does nothing
			args.postscript = '';
		end

		args.ref = pframe.args[table.concat ({n, 'ref'})] or '';				-- alternate reference for this source

		args.page = pframe.args[table.concat ({n, 'p'})] or '';					-- insource locations for this source
		args.pages = pframe.args[table.concat ({n, 'pp'})] or '';
		args.pages = ('' ~= args.pages) and hyphen_to_dash (args.pages) or '';
		args.location = pframe.args[table.concat ({n, 'loc'})] or '';
		args.ignore = ('yes' == pframe.args[table.concat ({n, 'ignore-false-positive'})]) or ('yes' == pframe.args[table.concat ({n, 'ignore-err'})]);

		table.insert (out, core (args));										-- save the rendering of this source
		
		for k, v in ipairs ({'P1', 'P2', 'P3', 'P4', 'P5'}) do					-- create the FOOTNOTE id
			if '' ~= args[v] then
				table.insert (footnote, args[v]);
			end
		end
		for k, v in ipairs ({'page', 'pages', 'location'}) do					-- these done separately so that we can strip uri-reserved characters from extlinked page numbers 
			if '' ~= args[v] then
				table.insert (footnote, strip_url (args[v]))
			end
		end
		
		last_index = n;															-- flags used to select terminal postscript from nps or from end_ps
		if '' ~= args.postscript then							
			last_ps = n;
		end
		
		n = n+1;																-- bump for the next one
	end
	
	local name = table.concat (footnote):gsub ('%s+', ' ');						-- put the footnote together and strip redundant space
	
	args.end_ps = pframe.args.postscript or pframe.args.ps or '.';				-- this is the postscript for the whole not for the individual sources
	if 'none' == args.end_ps then												-- not an original sfnm parameter value; added for compatibility with other footnote templates
		args.end_ps = '';
	end

	local result = table.concat ({table.concat (out, '; '), (last_index == last_ps) and '' or  args.end_ps});
	return frame:extensionTag ({name='ref', args={name=name}, content=result});
end


--[[--------------------------< S F N R E F >------------------------------------------------------------------

implements ((sfnref))

]]

local function sfnref (frame)
	local args = getArgs (frame);
	local out = {};
	
	for i=1, 5 do																-- get the first five args if there are five args
		if args[i] then
			out[i] = args[i];
		else
			break;																-- less than 5 args break out
		end
	end
	
	if 5 == #out then															-- when we have seen five args there may bemore
		local i = 6;															-- initialize the indexer to the sixth positional parameter
		while args[i] do														-- in case there are too many authors loop through the authors looking for a year
			if is_year (args[i], args) then										-- if a year
				out[5] = args[i];												-- overwrite whatever was in args[5] with year
				break;															-- and abandon the search
			end
			i = i + 1;															-- bump the indexer
		end
	end
	
	return mw.uri.anchorEncode ('CITEREF' .. table.concat (out));
end


--[[--------------------------< E X P O R T E D   F U N C T I O N S >------------------------------------------
]]

return {
	harvard_citation = harvard_citation,
	sfn = sfn,
	sfnm = sfnm,
	sfnref = sfnref,
	};