New run function based on rank2 grammars:

[SXSI/xpathcomp.git] / src / grammar2.ml

diff --git a/src/grammar2.ml b/src/grammar2.ml
index d2d497e..a7ccb09 100644 (file)
--- a/src/grammar2.ml
+++ b/src/grammar2.ml
@@ -205,11 +205,21 @@ struct
       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 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)) lsl 2 in
-      let rule_ = rule_ lor (List.length args) 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 *)
@@ -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
@@ -271,14 +281,17 @@ let load f =
   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 = [ `Start ] Node.t
@@ -296,7 +309,9 @@ type tn_symbol = [ any_type ] Node.t
 
 type partial =
     Leaf of node
-  | Node of tn_symbol * partial array
+  | Node0 of tn_symbol (* No parameters *)
+  | Node1 of tn_symbol * partial
+  | Node2 of tn_symbol * partial * partial
 
 
 let is_nil  (t : t_symbol) =
@@ -348,20 +363,42 @@ 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 6) land 0x7ffffff)
+  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 33)
+  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 =
-  (Node.to_int r) land 0b11
+  match get_conf r with
+  | C0 -> 0
+  | C1 | C2 | C3 -> 1
+  | C4 | C5 | C6 -> 2
 
 let get_id1_rank (r : rhs) : int =
-  ((Node.to_int r) lsr 2) land 0b11
+  match get_conf r with
+  | C0 | C1 | C4 -> 1
+  | _ -> 2
 
 let get_id2_pos (r : rhs) : int =
-  ((Node.to_int r) lsr 4) land 0b11
+  match get_conf r with
+  | C0 | C1 |C2 | C4 | C5 -> 1
+  | _ -> 2
 
 let get_id2_rank (r : rhs) : int =
-  get_rank r  + 1 - get_id1_rank r
+  match get_conf r with
+  | C0 | C2 | C3 -> 0
+  | C1 | C5 | C6 -> 1
+  | C4 -> 2
+