X-Git-Url: http://nitlanguage.org

diff --git a/src/parser/xss/parser.xss b/src/parser/xss/parser.xss
index 49c742f..56c1dfd 100644
--- a/src/parser/xss/parser.xss
+++ b/src/parser/xss/parser.xss
@@ -33,7 +33,7 @@ private class State
 end
 
 class Parser
-special ParserTable
+	super TablesCapable
 	# Associated lexer
 	var _lexer: Lexer
 
@@ -49,8 +49,6 @@ special ParserTable
 		_lexer = lexer
 		_stack = new Array[State]
 		_stack_pos = -1
-		build_goto_table
-		build_action_table
 		build_reduce_table
 	end
 
@@ -58,24 +56,24 @@ special ParserTable
 	private fun go_to(index: Int): Int
 	do
 		var state = state
-		var table = _goto_table[index]
 		var low = 1
-		var high = table.length/2 - 1
+		var high = parser_goto(index, 0) - 1
 
 		while low <= high do
 			var middle = (low + high) / 2
-			var subindex = middle * 2
+			var subindex = middle * 2 + 1 # +1 because parser_goto(index, 0) is the length
 
-			if state < table[subindex] then
+			var goal = parser_goto(index, subindex)
+			if state < goal then
 				high = middle - 1
-			else if state > table[subindex] then
+			else if state > goal then
 				low = middle + 1
 			else
-				return table[subindex + 1]
+				return parser_goto(index, subindex+1)
 			end
 		end
 
-		return table[1] # Default value
+		return parser_goto(index, 2) # Default value
 	end
 
 	# Push someting in the state stack
@@ -112,35 +110,32 @@ special ParserTable
 		push(0, null)
 
 		var lexer = _lexer
-		while true do
+		loop
 			var token = lexer.peek
-			var last_pos = token.location.column_start
-			var last_line = token.location.line_start
-
 			if token isa PError then
 				return new Start(null, token)
 			end
 
 			var index = token.parser_index
-			var table = _action_table[state]
-			var action_type = table[1]
-			var action_value = table[2]
+			var action_type = parser_action(state, 2)
+			var action_value = parser_action(state, 3)
 
 			var low = 1
-			var high = table.length/3 - 1
+			var high = parser_action(state, 0) - 1
 
 			while low <= high do
 				var middle = (low + high) / 2
-				var subindex = middle * 3
+				var subindex = middle * 3 + 1 # +1 because parser_action(state, 0) is the length
 
-				if index < table[subindex] then
+				var goal = parser_action(state, subindex)
+				if index < goal then
 					high = middle - 1
-				else if index > table[subindex] then
+				else if index > goal then
 					low = middle + 1
 				else
-					action_type = table[subindex + 1]
-					action_value = table[subindex + 2]
-					high = low -1 # break
+					action_type = parser_action(state, subindex+1)
+					action_value = parser_action(state, subindex+2)
+					break
 				end
 			end
 
@@ -154,16 +149,14 @@ special ParserTable
 				var node1 = pop
 				assert node1 isa ${/parser/prods/prod/@ename}
 				var node = new Start(node1, node2)
-				(new SearchTokensVisitor).enter_visit(node)
+				(new ComputeProdLocationVisitor).enter_visit(node)
 				return node
 			else if action_type == 3 then # ERROR
-				var location = new Location(lexer.filename, last_line, last_line, last_pos, last_pos)
-				var node2 = new PError.init_error(error_messages[errors[action_value]],location)
+				var node2 = new PError.init_error("Syntax error: unexpected token.", token.location)
 				var node = new Start(null, node2)
 				return node
 			end
 		end
-		abort
 	end
 
 	var _reduce_table: Array[ReduceAction]
@@ -171,62 +164,123 @@ special ParserTable
 	do
 		_reduce_table = new Array[ReduceAction].with_items(
 $ foreach {rules/rule}
-			new ReduceAction@index[-sep ','-]
+			new ReduceAction@index(@leftside)[-sep ','-]
 $ end foreach
 		)
 	end
 end
 
-# Find first and last tokens of production nodes
-private class SearchTokensVisitor
-special Visitor
-	var _untokenned_nodes: Array[Prod]
-	var _last_token: nullable Token = null
+redef class Prod
+	# Location on the first token after the start of a production
+	# So outside the production for epilon production
+	var _first_location: nullable Location
+
+	# Location of the last token before the end of a production
+	# So outside the production for epilon production
+	var _last_location: nullable Location
+end
+
+# Find location of production nodes
+# Uses existing token locations to infer location of productions.
+private class ComputeProdLocationVisitor
+	super Visitor
+	# Currenlty visited productions that need a first token
+	var _need_first_prods: Array[Prod] = new Array[Prod]
+
+	# Already visited epsilon productions that waits something after them
+	var _need_after_epsilons: Array[Prod] = new Array[Prod]
+
+	# Already visited epsilon production that waits something before them
+	var _need_before_epsilons: Array[Prod] = new Array[Prod]
+
+	# Location of the last visited token in the current production
+	var _last_location: nullable Location = null
+
 	redef fun visit(n: nullable PNode)
 	do
 		if n == null then
 			return
 		else if n isa Token then
-			_last_token = n
-			for no in _untokenned_nodes do
-				no.first_token = n
+			var loc = n.location
+			_last_location = loc
+
+			# Add a first token to productions that need one
+			for no in _need_first_prods do
+				no._first_location = loc
 			end
-			_untokenned_nodes.clear
+			_need_first_prods.clear
+
+			# Find location for already visited epsilon production that need one
+			for no in _need_after_epsilons do
+				# Epsilon production that is in the middle of a non-epsilon production
+				# The epsilon production has both a token before and after it
+				var endl = loc
+				var startl = no._last_location
+				no.location = new Location(endl.file, startl.line_end, endl.line_start, startl.column_end, endl.column_start)
+			end
+			_need_after_epsilons.clear
 		else
 			assert n isa Prod
-			_untokenned_nodes.add(n)
+			_need_first_prods.add(n)
+
+			var old_last = _last_location
+			_last_location = null
 			n.visit_all(self)
-			n.last_token = _last_token
-
-			if n.first_token != null then
-				var start_location = n.first_token.location
-				var end_location = _last_token.location
-
-				if start_location != null and end_location != null then
-					var file = end_location.file
-					var line_start = start_location.line_start
-					var line_end = end_location.line_end
-					var column_start = start_location.column_start
-					var column_end = end_location.column_end
-					n.location = new Location(file, line_start, line_end, column_start, column_end)
+			var endl = _last_location
+			if endl == null then _last_location = old_last
+
+			n._last_location = endl
+			var startl = n._first_location
+			if startl != null then
+				# Non-epsilon production
+				assert endl != null
+
+				n.location = new Location(startl.file, startl.line_start, endl.line_end, startl.column_start, endl.column_end)
+
+				for no in _need_before_epsilons do
+					# Epsilon production that starts the current non-epsilon production
+					#var startl = n.location
+					no.location = new Location(startl.file, startl.line_start, startl.line_start, startl.column_start, startl.column_start)
+				end
+				_need_before_epsilons.clear
+
+				for no in _need_after_epsilons do
+					# Epsilon production that finishes the current non-epsilon production
+					#var endl = n.location
+					no.location = new Location(endl.file, endl.line_end, endl.line_end, endl.column_end, endl.column_end)
+				end
+				_need_after_epsilons.clear
+			else
+				# No first token means epsilon production (or "throw all my tokens" production)
+				# So, it must be located it later
+				if endl == null then
+					# Epsilon production that starts a parent non-epsilon production
+					_need_before_epsilons.add(n)
+				else
+					# Epsilon production in the middle or that finishes a parent non-epsilon production
+					_need_after_epsilons.add(n)
 				end
 			end
 		end
 	end
-	init
-	do
-		_untokenned_nodes = new Array[Prod]
-	end
+
+	init do end
 end
 
 # Each reduca action has its own class, this one is the root of the hierarchy.
 private abstract class ReduceAction
 	fun action(p: Parser) is abstract
+	fun concat(l1, l2 : Array[Object]): Array[Object]
+	do
+		if l1.is_empty then return l2
+		l1.append(l2)
+		return l1
+	end
 end
 
 $ foreach {rules/rule}
 private class ReduceAction@index
-special ReduceAction
+	super ReduceAction
 	redef fun action(p: Parser)
 	do
 					var node_list: nullable Object = null
@@ -249,13 +303,7 @@ $     when {@cmd='ADDNODE'}
 					end
 $     end
 $     when {@cmd='ADDLIST'}
-#					if ${translate(@fromlist,"ABCDEFGHIJKLMNOPQRSTUVWXYZ","abcdefghijklmnopqrstuvwxyz")} != null then
-						if ${translate(@tolist,"ABCDEFGHIJKLMNOPQRSTUVWXYZ","abcdefghijklmnopqrstuvwxyz")}.is_empty then
-							${translate(@tolist,"ABCDEFGHIJKLMNOPQRSTUVWXYZ","abcdefghijklmnopqrstuvwxyz")} = ${translate(@fromlist,"ABCDEFGHIJKLMNOPQRSTUVWXYZ","abcdefghijklmnopqrstuvwxyz")}
-						else
-							${translate(@tolist,"ABCDEFGHIJKLMNOPQRSTUVWXYZ","abcdefghijklmnopqrstuvwxyz")}.append(${translate(@fromlist,"ABCDEFGHIJKLMNOPQRSTUVWXYZ","abcdefghijklmnopqrstuvwxyz")})
-						end
-#					end
+					${translate(@tolist,"ABCDEFGHIJKLMNOPQRSTUVWXYZ","abcdefghijklmnopqrstuvwxyz")} = concat(${translate(@tolist,"ABCDEFGHIJKLMNOPQRSTUVWXYZ","abcdefghijklmnopqrstuvwxyz")}, ${translate(@fromlist,"ABCDEFGHIJKLMNOPQRSTUVWXYZ","abcdefghijklmnopqrstuvwxyz")})
 $     end
 $     when {@cmd='MAKELIST'}
 					var ${translate(@result,"ABCDEFGHIJKLMNOPQRSTUVWXYZ","abcdefghijklmnopqrstuvwxyz")} = new Array[Object]
@@ -283,67 +331,42 @@ $     when {@cmd='RETURNLIST'}
 $     end
 $   end choose
 $   end foreach
-					p.push(p.go_to(@leftside), node_list)
+					p.push(p.go_to(_goto), node_list)
 	end
-init do end
+	var _goto: Int
+	init(g: Int) do _goto = g
 end
 $ end foreach
 $ end template
 
-$ template make_parser_tables()
-# Parser that build a full AST
-abstract class ParserTable
-	var _action_table: Array[Array[Int]]
-	private fun build_action_table
-	do
-		_action_table = once [
+$ template make_parser_table()
 $ foreach {parser_data/action_table/row}
-			action_table_row${position()}[-sep ','-]
+static int parser_action_row${position()}[] = {
+	${count(action)},
+$   foreach {action}
+	@from, @action, @to[-sep ','-]
+$   end foreach
+};
 $ end foreach
-	        ]
-	end
 
+const int* const parser_action_table[] = {
 $ foreach {parser_data/action_table/row}
-	private fun action_table_row${position()}: Array[Int]
-	do
-		return [
-$   foreach {action}
-				@from, @action, @to[-sep ','-]
-$   end foreach
-			]
-	end
+	parser_action_row${position()}[-sep ','-]
 $ end foreach
+};
 
-	var _goto_table: Array[Array[Int]]
-	private fun build_goto_table
-	do
-		_goto_table = once [
 $ foreach {parser_data/goto_table/row}
-			[
+static int parser_goto_row${position()}[] = {
+	${count(goto)},
 $   foreach {goto}
-				@from, @to[-sep ','-]
+	@from, @to[-sep ','-]
 $   end foreach
-			][-sep ','-]
+};
 $ end foreach
-		]
-	end
-
-	private fun error_messages: Array[String]
-	do
-		return once [
-$ foreach {parser_data/error_messages/msg}
-			"${sablecc:string2escaped_unicode(.)}"[-sep ','-]
-$ end
-		]
-	end
 
-	private fun errors: Array[Int]
-	do
-		return once [
-			[-foreach {parser_data/errors/i}-]${.}[-sep ','-][-end-]
-		]
-	end
-
-	init do end
-end
+const int* const parser_goto_table[] = {
+$ foreach {parser_data/goto_table/row}
+	parser_goto_row${position()}[-sep ','-]
+$ end foreach
+};
 $ end template