Update parsing of tree.

diff --git a/src/tree.ml b/src/tree.ml
index 587e649..0b2070b 100644 (file)
--- a/src/tree.ml
+++ b/src/tree.ml
@@ -59,12 +59,32 @@ struct
   external close_tag : t -> string -> unit = "caml_xml_tree_builder_new_closing_tag"
   external text : t -> string -> unit = "caml_xml_tree_builder_new_text"
 
+  let is_whitespace s =
+    let rec loop len i =
+      if i < len then
+       let c = s.[i] in
+       (c == '\n' || c == '\t' || c == ' ') && loop len (i+1)
+      else
+       true
+    in
+    loop (String.length s) 0
+
+
+  let display_count =
+    let event_counter = ref 0 in
+    (fun parser_ ->
+      incr event_counter;
+    if !event_counter land 0xffffff == 0 then
+      Printf.eprintf "Current position: %i\n%!" (Expat.get_current_byte_index parser_))
 
   let do_text b t =
     if Buffer.length t > 0 then begin
-      open_tag b "<$>";
-      text b (Buffer.contents t);
-      close_tag b "<$>";
+      let s = Buffer.contents t in
+      begin
+       open_tag b "<$>";
+       text b s;
+       close_tag b "<$>";
+      end;
       Buffer.clear t
     end
 
@@ -72,11 +92,12 @@ struct
     let atname = "<@>" ^ name in
     open_tag b atname;
     open_tag b "<@$>";
-      text b value;
-      close_tag b "<@$>";
-      close_tag b atname
+    text b value;
+    close_tag b "<@$>";
+    close_tag b atname
 
-  let start_element_handler b t tag attr_list =
+  let start_element_handler parser_ b t tag attr_list =
+    display_count parser_;
     do_text b t;
     open_tag b tag;
     match attr_list with
@@ -84,14 +105,16 @@ struct
       | l ->
        open_tag b "<@>";
        List.iter (fun (name, value) -> output_attr b name value) l;
-         close_tag b "<@>"
+       close_tag b "<@>"
 
 
-  let end_element_handler b t tag =
+  let end_element_handler parser_ b t tag =
+    display_count parser_;
     do_text b t;
     close_tag b tag
 
-  let character_data_handler _b t text =
+  let character_data_handler parser_ _ t text =
+    display_count parser_;
     Buffer.add_string t text
 
   let create_parser () =
@@ -101,11 +124,16 @@ struct
     let finalize () =
       do_text build buf;
       close_tag build "";
-      close_document build
+      Printf.eprintf "Finished Parsing\n%!";
+      Printf.eprintf "Started Index construction\n%!";
+      let r = close_document build in
+      Printf.eprintf "Finished Index construction\n%!";
+      r
     in
-    Expat.set_start_element_handler parser_ (start_element_handler build buf);
-    Expat.set_end_element_handler parser_ (end_element_handler build buf);
-    Expat.set_character_data_handler parser_ (character_data_handler build buf);
+    Expat.set_start_element_handler parser_ (start_element_handler parser_ build buf);
+    Expat.set_end_element_handler parser_ (end_element_handler parser_ build buf);
+    Expat.set_character_data_handler parser_ (character_data_handler parser_ build buf);
+    Printf.eprintf "Started Parsing\n%!";
     open_document build !Options.index_empty_texts !Options.sample_factor
       !Options.disable_text_collection !Options.text_index_type;
     open_tag build "";
@@ -134,8 +162,6 @@ struct
     in
       finalizer ()
 
-
-
 end
 
 
@@ -521,7 +547,9 @@ let is_node t = t != nil
 let is_root t = t == root
 
 let node_of_t t  =
+  eprintf "Initializing tag structure\n%!";
   let _ = Tag.init (mk_tag_ops t) in
+  eprintf "Starting tag table construction\n%!";
   let f, n, c, d = time collect_labels t ~msg:"Building tag relationship table" in
   let c = Array.map TagS.to_ptset c in
   let n = Array.map TagS.to_ptset n in
@@ -585,6 +613,7 @@ let save_tag_table channel t =
 let save t str =
   let fd = Unix.openfile str [ Unix.O_WRONLY;Unix.O_TRUNC;Unix.O_CREAT] 0o644 in
   let out_c = Unix.out_channel_of_descr fd in
+  let index_prefix = Filename.chop_suffix str ".srx" in
   let _ = set_binary_mode_out out_c true in
     output_string out_c magic_string;
     output_char out_c '\n';
@@ -597,7 +626,7 @@ let save t str =
     (* we need to move the fd to the correct position *)
     flush out_c;
     ignore (Unix.lseek fd (pos_out out_c) Unix.SEEK_SET);
-    tree_save t.doc fd str;
+    tree_save t.doc fd index_prefix;
     close_out out_c
 ;;
 let load_tag_table channel =
@@ -607,6 +636,7 @@ let load_tag_table channel =
 let load ?(sample=64) ?(load_text=true) str =
   let fd = Unix.openfile str [ Unix.O_RDONLY ] 0o644 in
   let in_c = Unix.in_channel_of_descr fd in
+  let index_prefix = Filename.chop_suffix str ".srx" in
   let _ = set_binary_mode_in in_c true in
   let load_table () =
     (let ms = input_line in_c in if ms <> magic_string then failwith "Invalid index file");
@@ -619,7 +649,7 @@ let load ?(sample=64) ?(load_text=true) str =
   in
   let c, s, d, f = time ~msg:"Loading tag table"(load_table) () in
   ignore(Unix.lseek fd (pos_in in_c) Unix.SEEK_SET);
-    let xml_tree = tree_load fd str load_text sample in
+    let xml_tree = tree_load fd index_prefix load_text sample in
     let () = Tag.init (Obj.magic xml_tree) in
     let attributes = Ptset.Int.add Tag.attribute d.(Tag.attribute) in
     let elements = Ptset.Int.add Tag.document_node