$ template make_parser()
-# State of the parser automata as stored in the parser stack.
-private class State
- # The internal state number
- readable writable var _state: Int
-
- # The node stored with the state in the stack
- readable writable var _nodes: nullable Object
-
- init(state: Int, nodes: nullable Object)
- do
- _state = state
- _nodes = nodes
- end
-end
-
-class Parser
-special ParserTable
- # Associated lexer
- var _lexer: Lexer
-
- # Stack of pushed states and productions
- var _stack: Array[State]
-
- # Position in the stack
- var _stack_pos: Int
-
- # Create a new parser based on a given lexer
- init(lexer: Lexer)
- do
- _lexer = lexer
- _stack = new Array[State]
- _stack_pos = -1
- build_goto_table
- build_action_table
- build_reduce_table
- end
-
- # Do a transition in the automata
- 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
-
- while low <= high do
- var middle = (low + high) / 2
- var subindex = middle * 2
-
- if state < table[subindex] then
- high = middle - 1
- else if state > table[subindex] then
- low = middle + 1
- else
- return table[subindex + 1]
- end
- end
-
- return table[1] # Default value
- end
-
- # Push someting in the state stack
- private fun push(numstate: Int, list_node: nullable Object)
- do
- var pos = _stack_pos + 1
- _stack_pos = pos
- if pos < _stack.length then
- var state = _stack[pos]
- state.state = numstate
- state.nodes = list_node
- else
- _stack.push(new State(numstate, list_node))
- end
- end
-
- # The current state
- private fun state: Int
- do
- return _stack[_stack_pos].state
- end
-
- # Pop something from the stack state
- private fun pop: nullable Object
- do
- var res = _stack[_stack_pos].nodes
- _stack_pos = _stack_pos -1
- return res
- end
-
- # Build and return a full AST.
- fun parse: Start
- do
- push(0, null)
-
- var lexer = _lexer
- loop
- var token = lexer.peek
- 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 low = 1
- var high = table.length/3 - 1
-
- while low <= high do
- var middle = (low + high) / 2
- var subindex = middle * 3
-
- if index < table[subindex] then
- high = middle - 1
- else if index > table[subindex] then
- low = middle + 1
- else
- action_type = table[subindex + 1]
- action_value = table[subindex + 2]
- high = low -1 # break
- end
- end
-
- if action_type == 0 then # SHIFT
- push(action_value, lexer.next)
- else if action_type == 1 then # REDUCE
- _reduce_table[action_value].action(self)
- else if action_type == 2 then # ACCEPT
- var node2 = lexer.next
- assert node2 isa EOF
- var node1 = pop
- assert node1 isa ${/parser/prods/prod/@ename}
- var node = new Start(node1, node2)
- (new ComputeProdLocationVisitor).enter_visit(node)
- return node
- else if action_type == 3 then # ERROR
- var node2 = new PError.init_error("Syntax error: unexpected token.", token.location)
- var node = new Start(null, node2)
- return node
- end
- if false then break # FIXME remove once unreach loop exits are in c_src
- end
- abort # FIXME remove once unreach loop exits are in c_src
- end
-
- var _reduce_table: Array[ReduceAction]
- private fun build_reduce_table
+redef class Parser
+ redef fun build_reduce_table
do
- _reduce_table = new Array[ReduceAction].with_items(
+ var reduce_table = new Array[ReduceAction].with_capacity(${count(rules/rule)})
$ foreach {rules/rule}
- new ReduceAction@index[-sep ','-]
+ reduce_table.add new ReduceAction@index(@leftside)
$ end foreach
- )
- end
-end
-
-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
-special 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
- 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
- _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
- _need_first_prods.add(n)
-
- var old_last = _last_location
- _last_location = null
- n.visit_all(self)
- 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 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
+ return reduce_table
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
var ${translate(@result,"ABCDEFGHIJKLMNOPQRSTUVWXYZ","abcdefghijklmnopqrstuvwxyz")} = new Array[Object]
$ end
$ when {@cmd='MAKENODE'}
+$ if {count(arg)!=0}
var ${translate(@result,"ABCDEFGHIJKLMNOPQRSTUVWXYZ","abcdefghijklmnopqrstuvwxyz")}: nullable @etype = new @etype.init_${translate(@etype,"ABCDEFGHIJKLMNOPQRSTUVWXYZ","abcdefghijklmnopqrstuvwxyz")}(
$ foreach {arg}
$ if @null
$ end
$ end foreach
)
+$ else
+ var ${translate(@result,"ABCDEFGHIJKLMNOPQRSTUVWXYZ","abcdefghijklmnopqrstuvwxyz")}: nullable @etype = new @etype.init_${translate(@etype,"ABCDEFGHIJKLMNOPQRSTUVWXYZ","abcdefghijklmnopqrstuvwxyz")}
+$ end
$ end
$ when {@cmd='RETURNNODE'}
$ if @null
$ 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
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
- 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
nitlanguage / nit.git / blobdiff