X-Git-Url: http://git.nguyen.vg/gitweb/?a=blobdiff_plain;f=src%2Fgrammar2.ml;h=ca21d0a99df0b1df65531b5827117be89dd8c8a5;hb=7270a43a59e6ff4b27ab0a339723ed8a24d0a091;hp=82cf036481133be80fb83adc0c8a37c0f969a69a;hpb=3dc9065cb7e4b38bf25e6fb50017efa5b11de4ff;p=SXSI%2Fxpathcomp.git

diff --git a/src/grammar2.ml b/src/grammar2.ml
index 82cf036..ca21d0a 100644
--- a/src/grammar2.ml
+++ b/src/grammar2.ml
@@ -200,30 +200,40 @@ struct
       loop 1 l
     in
     Hashtbl.iter (fun lhs rhs ->
-      let Node( head, _ ) = lhs in
+      let Node( head, args ) = lhs in
       let Node( tag1, params) = rhs in
       let tag2, pos2 = pos_id2 params in
       let id1 = h_find ~msg:"7" tag_to_id tag1
-      and id2 = h_find ~msg:"8" tag_to_id tag2
+      and id2 = h_find ~msg:"8" tag_to_id tag2 in
+      let conf =
+	if List.length args = 0 then 0
+	else
+	  if List.length args = 1 then
+	  if List.length params = 1 then 1
+	  else if pos2 = 1 then 2
+	  else 3
+	  else (* 2 parameters *)
+	    if List.length params = 1 then 4
+	    else if pos2 = 1 then 5
+	    else 6
       in
       let rule_ = id2 lsl 27 in
-      let rule_ = (rule_ lor id1) lsl 2 in
-      let rule_ = (rule_ lor pos2) lsl 2 in
-      let rule_ = rule_ lor (List.length params) in
+      let rule_ = (rule_ lor id1) lsl 3 in
+      let rule_ = rule_ lor conf in
       r_array.((h_find  ~msg:"9" tag_to_id head) - rules_offset ) <- rule_
     ) rules;
-    let l = Array.length renum_tags in
-    let tag32 = Array32.create l 0 in
+    (*let l = Array.length renum_tags in *)
+    (*let tag32 = Array32.create l 0 in
     for i = 0 to l - 1 do
       Array32.set tag32 i (renum_tags.(i) land 0x7ffffff);
-    done;
+    done; *)
     (* Remove the non-terminal names from the hash tables *)
     let tag_to_id2 = Hashtbl.create 31 in
     Hashtbl.iter (fun s i -> if i < rules_offset then Hashtbl.add tag_to_id2 s i)
       tag_to_id;
     { start = bv;
-      tags = tag32;
-      rules = renum_tags;
+      tags = renum_tags;
+      rules = r_array;
       rules_offset = rules_offset;
       tag_to_id = tag_to_id2;
       tag_of_id = Array.sub old_new_mapping 0 rules_offset
@@ -238,7 +248,7 @@ let parse file =
   g
 
 let _GRAMMAR_MAGIC = 0xaabbcc
-let _GRAMMAR_VERSION = 2
+let _GRAMMAR_VERSION = 3
 
 let save g f =
   let cout = open_out f in
@@ -258,62 +268,66 @@ let save g f =
 
 let load f =
   let cin = open_in f in
-  let pr_pos () =
-    Printf.eprintf "Position: %i kiB\n" (pos_in cin / 1024)
-  in
   let read () = Marshal.from_channel cin in
   if read () != _GRAMMAR_MAGIC then failwith "Invalid grammar file";
   if read () != _GRAMMAR_VERSION then failwith "Deprecated grammar format";
-  pr_pos();
   let tags : int array = read () in
-  pr_pos();
   let rules : int array = read () in
-  pr_pos();
   let rules_offset : int = read () in
-  pr_pos();
   let tag_to_id : (string, int) Hashtbl.t = read () in
-  pr_pos();
   let tag_of_id : string array = read () in
-  pr_pos();
   let fd = Unix.descr_of_in_channel cin in
   let pos = pos_in cin in
   ignore(Unix.lseek fd pos Unix.SEEK_SET);
   let bp = Bp.load fd in
   close_in cin;
-  {
+  let g = {
     start = bp;
     tags = tags;
     rules = rules;
     rules_offset = rules_offset;
     tag_to_id = tag_to_id;
     tag_of_id = tag_of_id;
-  }
+  } in
+  Printf.eprintf "Grammar size:%i kb\n%!"
+    ((Ocaml.size_b g  + Bp.alloc_stats ())/1024);
+  g
 
 
-type node = [ `Grammar ] Node.t
+type node = [ `Start ] Node.t
 
-type p_type  = [ `Parameter ]
 type n_type = [ `NonTerminal ]
 type t_type = [ `Terminal ]
-type any_type = [ p_type | n_type | t_type ]
-type symbol = [ any_type ] Node.t
+type r_type = [ `Rule ]
+type any_type = [ n_type | t_type ]
+type rhs = [ r_type ] Node.t
 
-type p_symbol = p_type Node.t
 type n_symbol = n_type Node.t
 type t_symbol = t_type Node.t
-type tn_symbol = [ n_type | t_type ] Node.t
+type tn_symbol = [ any_type ] Node.t
+
+
+type 'a partial =
+  | Cache of 'a
+  | Leaf of int*int * StateSet.t array * node
+  | Node0 of tn_symbol (* No parameters *)
+  | Node1 of tn_symbol * 'a partial
+  | Node2 of tn_symbol * 'a partial * 'a partial
 
 
-let is_nil : (t:t_symbol) =
+let is_nil  (t : t_symbol) =
   (Node.to_int t) == 4
 
 let nil_symbol : t_symbol =
   (Node.of_int 4)
 
 let translate_tag _ t  = if t == 4 then ~-1 else t
-let to_string t tag = tag_of_id.(Tag.to_int tag)
-let register_tag t tag =
-  try Hashtbl.find t.tag_to_id (Tag.to_int tag) with
+let to_string t tag =
+  if tag < Array.length t.tag_of_id then t.tag_of_id.(Tag.to_int tag)
+  else "<!INVALID TAG!>"
+
+let register_tag t s =
+  try Hashtbl.find t.tag_to_id s with
     Not_found -> 4
 
 let tag_operations t = {
@@ -322,21 +336,171 @@ let tag_operations t = {
   Tag.translate = (fun s -> translate_tag t s);
 }
 
+let start_root : node = Node.of_int 0
+let start_tag g (idx : node) : [<any_type] Node.t =
+  Node.of_int (g.tags.(Bp.preorder_rank g.start (Node.to_int idx)))
 
-let rhs_tag t idx =
-  t.tags.(Bp.preorder_rank t.start idx)
+;;
 
-let rhs_first_child t idx =
-  Bp.first_child t.start idx
+let start_first_child t (idx : node) =
+  Node.of_int (Bp.first_child t.start (Node.to_int idx))
 
-let rhs_next_sibling t idx =
-  Bp.next_sibling t.start idx
+let start_next_sibling t (idx : node) =
+  Node.of_int (Bp.next_sibling t.start (Node.to_int idx))
 
 let is_non_terminal t (n : [< any_type ] Node.t) =
   let n = Node.to_int n in
   n >= t.rules_offset
 
-let is_terminal t (n : [< any_type ] Node.t) = not(is_terminal t n)
+let is_terminal t (n : [< any_type ] Node.t) = not(is_non_terminal t n)
+
+external terminal : [< any_type ] Node.t -> t_symbol = "%identity"
+external non_terminal : [< any_type ] Node.t -> n_symbol = "%identity"
+
 
 let tag (n : t_symbol) : Tag.t = Obj.magic n
 
+let get_rule g (r : n_symbol) : rhs =
+  Node.of_int (g.rules.((Node.to_int r) - g.rules_offset))
+
+let get_id1 (r : rhs) : tn_symbol =
+  Node.of_int(((Node.to_int r) lsr 3) land 0x7ffffff)
+
+let get_id2 (r : rhs) : tn_symbol =
+  Node.of_int((Node.to_int r) lsr 30)
+
+type conf = | C0 (* B(C) *)
+	    | C1 (* B(C(y0)) *)
+	    | C2 (* B(C, y0) *)
+	    | C3 (* B(y0, C) *)
+	    | C4 (* B(C(y0, y1)) *)
+	    | C5 (* B(C(y0), y1) *)
+	    | C6 (* B(y0, C(y1)) *)
+
+let get_conf (r : rhs) : conf =
+  (Obj.magic ((Node.to_int r) land 0b111))
+
+
+let get_rank (r : rhs) : int =
+  match get_conf r with
+  | C0 -> 0
+  | C1 | C2 | C3 -> 1
+  | C4 | C5 | C6 -> 2
+
+let get_id1_rank (r : rhs) : int =
+  match get_conf r with
+  | C0 | C1 | C4 -> 1
+  | _ -> 2
+
+let get_id2_pos (r : rhs) : int =
+  match get_conf r with
+  | C0 | C1 |C2 | C4 | C5 -> 1
+  | _ -> 2
+
+let get_id2_rank (r : rhs) : int =
+  match get_conf r with
+  | C0 | C2 | C3 -> 0
+  | C1 | C5 | C6 -> 1
+  | C4 -> 2
+
+let is_attribute g tag =
+  tag > 4 && (to_string g tag).[0] == '2'
+
+let dummy_param : 'a partial = Leaf (~-1,~-1, [||], Node.nil)
+
+(*
+let rec start_skip g idx count =
+  if idx < Node.null then count else
+    let symbol = start_tag g idx in
+    if is_terminal g symbol then
+      let symbol = terminal symbol in
+      if symbol == nil_symbol then count else
+	let count = count+1 in
+	let fs = start_first_child g idx in
+	let countl = start_skip g fs count in
+	start_skip g fs countl
+    else
+      let nt = non_terminal symbol in
+      let rhs = get_rule g nt in
+      let nparam = get_rank rhs in
+      match nparam with
+      | 0 -> rule_skip g nt dummy_param dummy_param count
+      | 1 -> rule_skip g nt (Leaf(0,StateSet.empty, Node.nil,start_first_child g idx)) dummy_param count
+      | 2 ->
+	let fc = start_first_child g idx in
+	let ns = start_next_sibling g fc in
+	rule_skip g nt (Leaf (0,[||],fc)) (Leaf (1,[||],ns)) count
+      | _ -> assert false
+
+and rule_skip g t y0 y1 count =
+  let rhs = get_rule g t in
+  let id1 = get_id1 rhs in
+  let id2 = get_id2 rhs in
+  let conf = get_conf rhs in
+  if is_non_terminal g id1 then
+    let id1 = non_terminal id1 in
+    match conf with
+    | C0 ->rule_skip g id1 (Node0 id2) dummy_param count
+    | C1 -> rule_skip g id1 (Node1(id2,y0)) dummy_param count
+    | C2 -> rule_skip g id1 (Node0 id2) y0 count
+    | C3 -> rule_skip g id1 y0 (Node0 id2) count
+    | C4 -> rule_skip g id1 (Node2(id2, y0, y1)) dummy_param count
+    | C5 -> rule_skip g id1 (Node1(id2, y0)) y1 count
+    | C6 -> rule_skip g id1 y0 (Node1(id2, y1)) count
+  else
+    let id1 = terminal id1 in
+    match conf with
+    | C0 | C1 -> assert false
+    | C2  -> terminal_skip g id1 (Node0 id2) y0 count
+    | C3  -> terminal_skip g id1 y0 (Node0 id2) count
+    | C4  -> assert false
+    | C5  -> terminal_skip g id1 (Node1(id2, y0)) y1 count
+    | C6  -> terminal_skip g id1 y0 (Node1(id2, y1)) count
+
+and terminal_skip g (symbol : t_symbol) y0 y1 count =
+  if symbol == nil_symbol then count else
+    let count = count + 1 in
+    let countl = partial_skip g y0 count in
+    partial_skip g y1 countl
+
+and partial_skip g l count =
+  match l with
+  | Cache _ -> assert false
+  | Leaf (_,_,_, id) -> start_skip g id count
+  | Node0 id ->
+    if (terminal id) == nil_symbol then count
+    else
+      rule_skip g (non_terminal id) dummy_param dummy_param count
+
+  | Node1 (id, y0) ->
+    rule_skip g (non_terminal id) y0 dummy_param count
+
+  | Node2 (id, y0, y1) ->
+
+    if is_terminal g id then
+      terminal_skip g (terminal id) y0 y1 count
+    else
+      rule_skip g (non_terminal id) y0 y1 count
+
+
+let dispatch_param0 conf id2 y0 y1 =
+  match conf with
+  | C0 -> Node0 id2
+  | C1 -> Node1(id2,y0)
+  | C2 -> Node0 id2
+  | C3 -> Node0 id2
+  | C4 -> Node2(id2, y0, y1)
+  | C5 -> Node1(id2, y0)
+  | C6 -> y0
+
+let dispatch_param1 conf id2 y0 y1 =
+  match conf with
+  | C0 -> dummy_param
+  | C1 -> dummy_param
+  | C2 -> y0
+  | C3 -> Node0 id2
+  | C4 -> dummy_param
+  | C5 -> y1
+  | C6 -> Node1(id2, y1)
+
+*)