INCLUDE "utils.ml"
-external init_lib : unit -> unit = "caml_init_lib"
+external init_lib : unit -> unit = "sxsi_cpp_init"
exception CPlusPlusError of string
type tree
+
+external register_tag : tree -> string -> Tag.t = "caml_xml_tree_register_tag"
+
+external tag_name : tree -> Tag.t -> string = "caml_xml_tree_get_tag_name"
+
+let tag t = (); fun s ->
+ match s with
+ | "<$>" -> Tag.pcdata
+ | "<@>" -> Tag.attribute
+ | "" -> Tag.document_node
+ | "<@$>" -> Tag.attribute_data
+ | _ -> register_tag t s
+
+let to_string d = ();
+ fun t ->
+ if t == Tag.pcdata then "<$>"
+ else if t == Tag.attribute_data then "<@$>"
+ else if t == Tag.attribute then "<@>"
+ else if t == Tag.nullt then "<!NIL!>"
+ else tag_name d t
+
+let translate x = x
+
+let mk_tag_ops t = {
+ Tag.tag = tag t;
+ Tag.to_string = to_string t;
+ Tag.translate = translate
+}
+
+module TreeBuilder =
+struct
+ type t
+ external create : unit -> t = "caml_xml_tree_builder_create"
+ external open_document : t -> bool -> int -> bool -> int -> unit = "caml_xml_tree_builder_open_document"
+ external close_document : t -> tree = "caml_xml_tree_builder_close_document"
+ external open_tag : t -> string -> unit = "caml_xml_tree_builder_new_open_tag"
+ 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
+ let s = Buffer.contents t in
+ begin
+ open_tag b "<$>";
+ text b s;
+ close_tag b "<$>";
+ end;
+ Buffer.clear t
+ end
+
+ let output_attr b name value =
+ let atname = "<@>" ^ name in
+ open_tag b atname;
+ open_tag b "<@$>";
+ text b value;
+ close_tag b "<@$>";
+ close_tag b atname
+
+ 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
+ [] -> ()
+ | l ->
+ open_tag b "<@>";
+ List.iter (fun (name, value) -> output_attr b name value) l;
+ close_tag b "<@>"
+
+
+ let end_element_handler parser_ b t tag =
+ display_count parser_;
+ do_text b t;
+ close_tag b tag
+
+ let character_data_handler parser_ _ t text =
+ display_count parser_;
+ Buffer.add_string t text
+
+ let create_parser () =
+ let buf = Buffer.create 512 in
+ let build = create () in
+ let parser_ = Expat.parser_create ~encoding:None in
+ let finalize () =
+ do_text build buf;
+ close_tag 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 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 "";
+ parser_, finalize
+
+ let parse_string s =
+ let parser_, finalizer = create_parser () in
+ Expat.parse parser_ s;
+ finalizer ()
+
+ let parse_file file =
+ let in_chan = open_in file in
+ let buffer = String.create 4096 in
+ let parser_, finalizer = create_parser () in
+ let () =
+ try
+ while true do
+ let read = input in_chan buffer 0 4096 in
+ if read == 0 then raise End_of_file else
+ Expat.parse_sub parser_ buffer 0 read;
+ done
+
+ with
+ | End_of_file -> close_in in_chan
+ | e -> raise e
+ in
+ finalizer ()
+
+end
+
+
+
+
type bit_vector = string
external bool_of_int : int -> bool = "%identity"
}
-
+let tag_operations t = mk_tag_ops t.doc
+(*
external parse_xml_uri : string -> int -> bool -> bool -> int -> tree = "caml_call_shredder_uri"
external parse_xml_string : string -> int -> bool -> bool -> int -> tree = "caml_call_shredder_string"
+*)
external tree_print_xml_fast3 : tree -> [`Tree ] Node.t -> Unix.file_descr -> unit = "caml_xml_tree_print"
let print_xml t n fd =
tree_print_xml_fast3 t.doc n fd
external tree_next_element : tree -> [`Tree] Node.t -> [`Tree] Node.t = "caml_xml_tree_next_element" "noalloc"
let next_element t n = tree_next_element t.doc n
+external tree_next_node_before : tree -> [`Tree] Node.t -> [`Tree] Node.t -> [`Tree] Node.t = "caml_xml_tree_next_node_before" "noalloc"
+let next_node_before t n ctx = tree_next_node_before t.doc n ctx
+
external tree_tagged_following_sibling : tree -> [`Tree] Node.t -> Tag.t -> [`Tree] Node.t = "caml_xml_tree_tagged_following_sibling" "noalloc"
let tagged_following_sibling t n tag = tree_tagged_following_sibling t.doc n tag
let is_root t = t == root
let node_of_t t =
- let _ = Tag.init (Obj.magic t) in
+ 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
}
-let parse f str =
- node_of_t
- (f str !Options.sample_factor
- !Options.index_empty_texts
- !Options.disable_text_collection
- !Options.text_index_type
- )
-let parse_xml_uri str = parse parse_xml_uri str
-let parse_xml_string str = parse parse_xml_string str
+let parse_xml_uri str = node_of_t (TreeBuilder.parse_file str)
+let parse_xml_string str = node_of_t (TreeBuilder.parse_string str)
let size t = tree_size t.doc;;
-external pool : tree -> Tag.pool = "%identity"
-
let magic_string = "SXSI_INDEX"
let version_string = "3"
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';
(* 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 =
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");
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
-let tag_pool t = pool t.doc
-
let equal a b = a == b
let nts = function
projects / SXSI / xpathcomp.git / blobdiff