1 (******************************************************************************)
2 (* SXSI : XPath evaluator *)
3 (* Kim Nguyen (Kim.Nguyen@nicta.com.au) *)
4 (* Copyright NICTA 2008 *)
5 (* Distributed under the terms of the LGPL (see LICENCE) *)
6 (******************************************************************************)
11 external init_lib : unit -> unit = "sxsi_cpp_init"
13 exception CPlusPlusError of string
15 let () = Callback.register_exception "CPlusPlusError" (CPlusPlusError "")
19 type node = [ `Tree ] Node.t
24 external register_tag : tree -> string -> Tag.t = "caml_xml_tree_register_tag"
26 external tag_name : tree -> Tag.t -> string = "caml_xml_tree_get_tag_name"
28 let tag t = (); fun s ->
31 | "<@>" -> Tag.attribute
32 | "" -> Tag.document_node
33 | "<@$>" -> Tag.attribute_data
34 | _ -> register_tag t s
38 if t == Tag.pcdata then "<$>"
39 else if t == Tag.attribute_data then "<@$>"
40 else if t == Tag.attribute then "<@>"
41 else if t == Tag.nullt then "<!NIL!>"
48 Tag.to_string = to_string t;
49 Tag.translate = translate
55 external create : unit -> t = "caml_xml_tree_builder_create"
56 external open_document : t -> bool -> int -> bool -> int -> unit = "caml_xml_tree_builder_open_document"
57 external close_document : t -> tree = "caml_xml_tree_builder_close_document"
58 external open_tag : t -> string -> unit = "caml_xml_tree_builder_new_open_tag"
59 external close_tag : t -> string -> unit = "caml_xml_tree_builder_new_closing_tag"
60 external text : t -> string -> unit = "caml_xml_tree_builder_new_text"
66 (c == '\n' || c == '\t' || c == ' ') && loop len (i+1)
70 loop (String.length s) 0
74 let event_counter = ref 0 in
77 if !event_counter land 0xffffff == 0 then
78 Printf.eprintf "Current position: %i\n%!" (Expat.get_current_byte_index parser_))
81 if Buffer.length t > 0 then begin
82 let s = Buffer.contents t in
91 let output_attr b name value =
92 let atname = "<@>" ^ name in
99 let start_element_handler parser_ b t tag attr_list =
100 display_count parser_;
107 List.iter (fun (name, value) -> output_attr b name value) l;
111 let end_element_handler parser_ b t tag =
112 display_count parser_;
116 let character_data_handler parser_ _ t text =
117 display_count parser_;
118 Buffer.add_string t text
120 let create_parser () =
121 let buf = Buffer.create 512 in
122 let build = create () in
123 let parser_ = Expat.parser_create ~encoding:None in
127 Printf.eprintf "Finished Parsing\n%!";
128 Printf.eprintf "Started Index construction\n%!";
129 let r = close_document build in
130 Printf.eprintf "Finished Index construction\n%!";
133 Expat.set_start_element_handler parser_ (start_element_handler parser_ build buf);
134 Expat.set_end_element_handler parser_ (end_element_handler parser_ build buf);
135 Expat.set_character_data_handler parser_ (character_data_handler parser_ build buf);
136 Printf.eprintf "Started Parsing\n%!";
137 open_document build !Options.index_empty_texts !Options.sample_factor
138 !Options.disable_text_collection !Options.text_index_type;
143 let parser_, finalizer = create_parser () in
144 Expat.parse parser_ s;
147 let parse_file file =
148 let in_chan = open_in file in
149 let buffer = String.create 4096 in
150 let parser_, finalizer = create_parser () in
154 let read = input in_chan buffer 0 4096 in
155 if read == 0 then raise End_of_file else
156 Expat.parse_sub parser_ buffer 0 read;
160 | End_of_file -> close_in in_chan
170 type bit_vector = string
172 external bool_of_int : int -> bool = "%identity"
174 let bit_vector_unsafe_get v i =
176 (((Char.code (String.unsafe_get v (i lsr 3))) lsr (i land 7)) land 1)
180 elements: Ptset.Int.t;
181 attributes: Ptset.Int.t;
182 attribute_array : Tag.t array;
183 children : Ptset.Int.t array;
184 siblings : Ptset.Int.t array;
185 descendants: Ptset.Int.t array;
186 followings: Ptset.Int.t array;
190 let tag_operations t = mk_tag_ops t.doc
192 external parse_xml_uri : string -> int -> bool -> bool -> int -> tree = "caml_call_shredder_uri"
193 external parse_xml_string : string -> int -> bool -> bool -> int -> tree = "caml_call_shredder_string"
195 external tree_print_xml_fast3 : tree -> [`Tree ] Node.t -> Unix.file_descr -> unit = "caml_xml_tree_print"
196 let print_xml t n fd =
197 tree_print_xml_fast3 t.doc n fd
200 external tree_save : tree -> Unix.file_descr -> string -> unit = "caml_xml_tree_save"
201 external tree_load : Unix.file_descr -> string -> bool -> int -> tree = "caml_xml_tree_load"
203 external nullt : unit -> 'a Node.t = "caml_xml_tree_nullt"
205 let nil : [`Tree ] Node.t = Node.nil
206 let root : [`Tree ] Node.t = Node.null
210 external unordered_set_alloc : int -> unordered_set = "caml_unordered_set_alloc"
211 external unordered_set_length : unordered_set -> int = "caml_unordered_set_length"
212 external unordered_set_insert : unordered_set -> int -> unit = "caml_unordered_set_set" "noalloc"
214 module HPtset = Hashtbl.Make(Ptset.Int)
216 let vector_htbl = HPtset.create MED_H_SIZE
218 let unordered_set_of_set s =
220 HPtset.find vector_htbl s
223 let v = unordered_set_alloc (Ptset.Int.cardinal s) in
224 let _ = Ptset.Int.iter (fun e -> unordered_set_insert v e) s in
225 HPtset.add vector_htbl s v; v
227 let ptset_to_vector = unordered_set_of_set
229 (** tree interface *)
231 external tree_root : tree -> [`Tree] Node.t = "caml_xml_tree_root" "noalloc"
234 external tree_first_child : tree -> [`Tree] Node.t -> [`Tree] Node.t = "caml_xml_tree_first_child" "noalloc"
235 let first_child t n = tree_first_child t.doc n
237 external tree_first_element : tree -> [`Tree] Node.t -> [`Tree] Node.t = "caml_xml_tree_first_element" "noalloc"
238 let first_element t n = tree_first_element t.doc n
240 external tree_tagged_child : tree -> [`Tree] Node.t -> Tag.t -> [`Tree] Node.t = "caml_xml_tree_tagged_child" "noalloc"
241 let tagged_child t n tag = tree_tagged_child t.doc n tag
243 external tree_select_child : tree -> [`Tree ] Node.t -> unordered_set -> [`Tree] Node.t = "caml_xml_tree_select_child" "noalloc"
244 let select_child t n tag_set = tree_select_child t.doc n tag_set
246 external tree_last_child : tree -> [`Tree] Node.t -> [`Tree] Node.t = "caml_xml_tree_last_child" "noalloc"
247 let last_child t n = tree_last_child t.doc n
250 external tree_next_sibling : tree -> [`Tree] Node.t -> [`Tree] Node.t = "caml_xml_tree_next_sibling" "noalloc"
251 let next_sibling t n = tree_next_sibling t.doc n
253 external tree_next_element : tree -> [`Tree] Node.t -> [`Tree] Node.t = "caml_xml_tree_next_element" "noalloc"
254 let next_element t n = tree_next_element t.doc n
256 external tree_next_node_before : tree -> [`Tree] Node.t -> [`Tree] Node.t -> [`Tree] Node.t = "caml_xml_tree_next_node_before" "noalloc"
257 let next_node_before t n ctx = tree_next_node_before t.doc n ctx
259 external tree_tagged_following_sibling : tree -> [`Tree] Node.t -> Tag.t -> [`Tree] Node.t = "caml_xml_tree_tagged_following_sibling" "noalloc"
260 let tagged_following_sibling t n tag = tree_tagged_following_sibling t.doc n tag
262 external tree_select_following_sibling : tree -> [`Tree ] Node.t -> unordered_set -> [`Tree] Node.t = "caml_xml_tree_select_following_sibling" "noalloc"
263 let select_following_sibling t n tag_set = tree_select_following_sibling t.doc n tag_set
265 external tree_prev_sibling : tree -> [`Tree] Node.t -> [`Tree] Node.t = "caml_xml_tree_prev_sibling" "noalloc"
266 let prev_sibling t n = tree_prev_sibling t.doc n
270 external tree_tagged_descendant : tree -> [`Tree ] Node.t -> Tag.t -> [`Tree ] Node.t = "caml_xml_tree_tagged_descendant" "noalloc"
271 let tagged_descendant t n tag = tree_tagged_descendant t.doc n tag
273 external tree_tagged_next : tree -> [`Tree ] Node.t -> Tag.t -> [`Tree ] Node.t = "caml_xml_tree_tagged_next" "noalloc"
274 let tagged_next t n tag = tree_tagged_next t.doc n tag
276 external tree_select_descendant : tree -> [`Tree ] Node.t -> unordered_set -> [`Tree] Node.t = "caml_xml_tree_select_descendant" "noalloc"
277 let select_descendant t n tag_set = tree_select_descendant t.doc n tag_set
279 external tree_tagged_following_before : tree -> [`Tree ] Node.t -> Tag.t -> [`Tree ] Node.t -> [`Tree ] Node.t = "caml_xml_tree_tagged_following_before" "noalloc"
280 let tagged_following_before t n tag ctx = tree_tagged_following_before t.doc n tag ctx
282 external tree_select_following_before : tree -> [`Tree ] Node.t -> unordered_set -> [`Tree] Node.t -> [`Tree] Node.t = "caml_xml_tree_select_following_before" "noalloc"
283 let select_following_before t n tag_set ctx = tree_select_following_before t.doc n tag_set ctx
285 external tree_parent : tree -> [`Tree] Node.t -> [`Tree] Node.t = "caml_xml_tree_parent" "noalloc"
286 let parent t n = tree_parent t.doc n
288 external tree_binary_parent : tree -> [`Tree] Node.t -> [`Tree] Node.t = "caml_xml_tree_binary_parent"
290 let binary_parent t n = tree_binary_parent t.doc n
293 external tree_tag : tree -> [`Tree] Node.t -> Tag.t = "caml_xml_tree_tag" "noalloc"
294 let tag t n = tree_tag t.doc n
296 external tree_is_first_child : tree -> [ `Tree ] Node.t -> bool = "caml_xml_tree_is_first_child" "noalloc"
297 let is_first_child t n = tree_is_first_child t.doc n
299 external tree_is_right_descendant : tree -> [ `Tree ] Node.t -> [`Tree] Node.t -> bool =
300 "caml_xml_tree_is_right_descendant" "noalloc"
302 let is_right_descendant t n1 n2 = tree_is_right_descendant t.doc n1 n2
305 let node_tags t = Ptset.Int.add Tag.document_node t.descendants.(Tag.document_node)
307 let attribute_tags t = t.attributes
309 let element_tags t = t.elements
312 t.children.(tag), t.descendants.(tag), t.siblings.(tag), t.followings.(tag)
315 let dump_tag_table t =
316 eprintf "Child tags:\n%!";
318 (fun tag set -> eprintf "%s: %a\n%!"
319 (Tag.to_string tag) TagSet.print (TagSet.inj_positive set))
321 eprintf "-----------------------------\n%!";
322 eprintf "Descendant tags:\n%!";
324 (fun tag set -> eprintf "%s: %a\n%!"
325 (Tag.to_string tag) TagSet.print (TagSet.inj_positive set))
327 eprintf "-----------------------------\n%!";
328 eprintf "Sibling tags:\n%!";
330 (fun tag set -> eprintf "%s: %a\n%!"
331 (Tag.to_string tag) TagSet.print (TagSet.inj_positive set))
333 eprintf "-----------------------------\n%!";
334 eprintf "Following tags:\n%!";
336 (fun tag set -> eprintf "%s: %a\n%!"
337 (Tag.to_string tag) TagSet.print (TagSet.inj_positive set))
339 eprintf "-----------------------------\n%!"
342 external tree_subtree_tags : tree -> [`Tree] Node.t -> Tag.t -> int = "caml_xml_tree_subtree_tags" "noalloc"
343 let subtree_tags t n tag = tree_subtree_tags t.doc n tag
345 external tree_subtree_size : tree -> [`Tree] Node.t -> int = "caml_xml_tree_subtree_size" "noalloc"
346 let subtree_size t n = tree_subtree_size t.doc n
348 let subtree_elements t node =
349 let size = tree_subtree_size t.doc node - 1 in
351 else let size = size - (tree_subtree_tags t.doc node Tag.pcdata) in
352 if size < 2 then size else
353 let acc = ref size in
354 for i = 0 to Array.length t.attribute_array - 1 do
355 acc := !acc - tree_subtree_tags t.doc node t.attribute_array.(i)
359 external tree_closing : tree -> [`Tree] Node.t -> [`Tree] Node.t = "caml_xml_tree_closing" "noalloc"
360 let closing t n = tree_closing t.doc n
362 external tree_num_tags : tree -> int = "caml_xml_tree_num_tags" "noalloc"
363 let num_tags t = tree_num_tags t.doc
365 external tree_size : tree -> int = "caml_xml_tree_size" "noalloc"
366 let size t = tree_size t.doc
371 let rec loop left node acc_d total_d num_leaves =
373 (acc_d+total_d,if left then num_leaves+1 else num_leaves)
375 let d,td = loop true (tree_first_child tree node) (acc_d+1) total_d num_leaves in
376 loop false (tree_next_sibling tree node) (acc_d) d td
378 let a,b = loop true root 0 0 0
380 Printf.eprintf "Average depth: %f, number of leaves %i\n%!" ((float_of_int a)/. (float_of_int b)) b
388 external hash : t -> int = "%identity"
389 external uid : t -> Uid.t = "%identity"
390 external equal : t -> t -> bool = "%eq"
391 external make : t -> int = "%identity"
392 external node : t -> int = "%identity"
393 external stats : unit -> unit = "%identity"
396 let to_ptset s = fold (Ptset.Int.add) s Ptset.Int.empty
400 Hashtbl.Make(struct type t = TagS.t * TagS.t
402 HASHINT2(Uid.to_int x.TagS.Node.id,
403 Uid.to_int y.TagS.Node.id)
410 Hashtbl.Make(struct type t = Tag.t * TagS.t
412 HASHINT2(x, Uid.to_int y.TagS.Node.id)
419 let add_cache = TagTSCache.create 1023
420 let union_cache = TSTSCache.create 1023
421 let subset_cache = TSTSCache.create 1023
424 TSTSCache.clear union_cache;
425 TSTSCache.clear subset_cache;
426 TagTSCache.clear add_cache
429 (x == y) || (x == TagS.empty) ||
430 if y == TagS.empty then false
434 TSTSCache.find subset_cache key
437 let z = TagS.subset x y in
438 TSTSCache.add subset_cache key z; z
440 let order ((x, y) as z) =
441 if x.TagS.Node.id <= y.TagS.Node.id then z
445 if _subset x y then y
446 else if _subset y x then x
448 let key = order (x, y) in
450 TSTSCache.find union_cache key
453 let z = TagS.union x y in
454 TSTSCache.add union_cache key z; z
459 TagTSCache.find add_cache key
462 let z = TagS.add t s in
463 TagTSCache.add add_cache key z;z
465 let child_sibling_labels tree =
466 let table_c = Array.create (tree_num_tags tree) TagS.empty in
467 let table_n = Array.copy table_c in
469 if node == nil then TagS.empty
471 let children = loop (tree_first_child tree node) in
472 let tag = tree_tag tree node in
474 let tc = table_c.(tag) in
475 if _subset children tc then ()
476 else table_c.(tag) <- _union tc children
478 let siblings = loop (tree_next_sibling tree node) in
480 let tn = table_n.(tag) in
481 if _subset siblings tn then ()
482 else table_n.(tag) <- _union tn siblings
489 let descendant_labels tree =
490 let table_d = Array.create (tree_num_tags tree) TagS.empty in
492 if node == nil then TagS.empty else
493 let d1 = loop (tree_first_child tree node) in
494 let d2 = loop (tree_next_sibling tree node) in
495 let tag = tree_tag tree node in
497 let td = table_d.(tag) in
498 if _subset d1 td then ()
499 else table_d.(tag) <- _union td d1;
501 _add tag (_union d1 d2)
506 let collect_labels tree =
507 let table_f = Array.create (tree_num_tags tree) TagS.empty in
508 let table_n = Array.copy table_f in
509 let table_c = Array.copy table_f in
510 let table_d = Array.copy table_f in
511 let rec loop node foll_siblings descendants followings =
512 if node == nil then foll_siblings, descendants, followings else
513 let tag = tree_tag tree node in
515 let tf = table_f.(tag) in
516 if _subset followings tf then ()
517 else table_f.(tag) <- _union tf followings in
519 let tn = table_n.(tag) in
520 if _subset foll_siblings tn then ()
521 else table_n.(tag) <- _union tn foll_siblings in
522 let children, n_descendants, n_followings =
523 loop (tree_last_child tree node) TagS.empty TagS.empty followings
526 let tc = table_c.(tag) in
527 if _subset children tc then ()
528 else table_c.(tag) <- _union tc children
531 let td = table_d.(tag) in
532 if _subset n_descendants td then ()
533 else table_d.(tag) <- _union td n_descendants
535 loop (tree_prev_sibling tree node)
536 (_add tag foll_siblings)
537 (_add tag (_union n_descendants descendants))
538 (_add tag n_followings)
540 ignore (loop root TagS.empty TagS.empty TagS.empty);
541 table_f, table_n, table_c, table_d
544 let is_nil t = t == nil
545 let is_node t = t != nil
546 let is_root t = t == root
549 eprintf "Initializing tag structure\n%!";
550 let _ = Tag.init (mk_tag_ops t) in
551 eprintf "Starting tag table construction\n%!";
552 let f, n, c, d = time collect_labels t ~msg:"Building tag relationship table" in
553 let c = Array.map TagS.to_ptset c in
554 let n = Array.map TagS.to_ptset n in
555 let f = Array.map TagS.to_ptset f in
556 let d = Array.map TagS.to_ptset d in
557 let () = clear_cache () in
558 let attributes = Ptset.Int.add Tag.attribute d.(Tag.attribute) in
559 let elements = Ptset.Int.add Tag.document_node
560 (Ptset.Int.remove Tag.pcdata
561 (Ptset.Int.diff d.(Tag.document_node) attributes))
564 attributes = attributes;
565 attribute_array = Array.of_list (Ptset.Int.elements attributes);
575 let parse_xml_uri str = node_of_t (TreeBuilder.parse_file str)
576 let parse_xml_string str = node_of_t (TreeBuilder.parse_string str)
578 let size t = tree_size t.doc;;
580 let magic_string = "SXSI_INDEX"
581 let version_string = "3"
584 Unix.lseek fd 0 Unix.SEEK_CUR
586 let pr_pos fd = Printf.eprintf "At position %i\n%!" (pos fd)
589 let sl = String.length s in
590 let ssl = Printf.sprintf "%020i" sl in
591 ignore (Unix.write fd ssl 0 20);
592 ignore (Unix.write fd s 0 (String.length s))
594 let rec really_read fd buffer start length =
595 if length <= 0 then () else
596 match Unix.read fd buffer start length with
597 0 -> raise End_of_file
598 | r -> really_read fd buffer (start + r) (length - r);;
601 let buffer = String.create 20 in
602 let _ = really_read fd buffer 0 20 in
603 let size = int_of_string buffer in
604 let buffer = String.create size in
605 let _ = really_read fd buffer 0 size in
608 let save_tag_table channel t =
609 let t = Array.map (fun s -> Array.of_list (Ptset.Int.elements s)) t in
610 Marshal.to_channel channel t []
613 let fd = Unix.openfile str [ Unix.O_WRONLY;Unix.O_TRUNC;Unix.O_CREAT] 0o644 in
614 let out_c = Unix.out_channel_of_descr fd in
615 let index_prefix = Filename.chop_suffix str ".srx" in
616 let _ = set_binary_mode_out out_c true in
617 output_string out_c magic_string;
618 output_char out_c '\n';
619 output_string out_c version_string;
620 output_char out_c '\n';
621 save_tag_table out_c t.children;
622 save_tag_table out_c t.siblings;
623 save_tag_table out_c t.descendants;
624 save_tag_table out_c t.followings;
625 (* we need to move the fd to the correct position *)
627 ignore (Unix.lseek fd (pos_out out_c) Unix.SEEK_SET);
628 tree_save t.doc fd index_prefix;
631 let load_tag_table channel =
632 let table : int array array = Marshal.from_channel channel in
633 Array.map (fun a -> Ptset.Int.from_list (Array.to_list a)) table
635 let load ?(sample=64) ?(load_text=true) str =
636 let fd = Unix.openfile str [ Unix.O_RDONLY ] 0o644 in
637 let in_c = Unix.in_channel_of_descr fd in
638 let index_prefix = Filename.chop_suffix str ".srx" in
639 let _ = set_binary_mode_in in_c true in
641 (let ms = input_line in_c in if ms <> magic_string then failwith "Invalid index file");
642 (let vs = input_line in_c in if vs <> version_string then failwith "Invalid version file");
643 let c = load_tag_table in_c in
644 let s = load_tag_table in_c in
645 let d = load_tag_table in_c in
646 let f = load_tag_table in_c in
649 let c, s, d, f = time ~msg:"Loading tag table"(load_table) () in
650 ignore(Unix.lseek fd (pos_in in_c) Unix.SEEK_SET);
651 let xml_tree = tree_load fd index_prefix load_text sample in
652 let () = Tag.init (Obj.magic xml_tree) in
653 let attributes = Ptset.Int.add Tag.attribute d.(Tag.attribute) in
654 let elements = Ptset.Int.add Tag.document_node
655 (Ptset.Int.remove Tag.pcdata
656 (Ptset.Int.diff d.(Tag.document_node) attributes))
658 let tree = { doc = xml_tree;
659 attributes = attributes;
660 attribute_array = Array.of_list (Ptset.Int.elements attributes);
673 let equal a b = a == b
677 | i -> Printf.sprintf "Node (%i)" i
679 let dump_node t = nts (Node.to_int t)
683 type query_result = { bv : bit_vector;
687 external tree_flush : tree -> Unix.file_descr -> unit = "caml_xml_tree_flush"
688 let flush t fd = tree_flush t.doc fd
690 external text_prefix : tree -> string -> bool -> query_result = "caml_text_collection_prefix_bv"
691 let text_prefix t s b = text_prefix t.doc s b
693 external text_suffix : tree -> string -> bool -> query_result = "caml_text_collection_suffix_bv"
694 let text_suffix t s b = text_suffix t.doc s b
696 external text_equals : tree -> string -> bool -> query_result = "caml_text_collection_equals_bv"
697 let text_equals t s b = text_equals t.doc s b
699 external text_contains : tree -> string -> bool -> query_result = "caml_text_collection_contains_bv"
700 let text_contains t s b = text_contains t.doc s b
703 module Predicate = Hcons.Make (
706 type t = (_t -> node -> bool) ref
707 let hash t = Hashtbl.hash t
708 let equal t1 t2 = t1 == t2
711 let string_of_query query =
713 | `Prefix -> "starts-with"
714 | `Suffix -> "ends-with"
715 | `Equals -> "equals"
716 | `Contains -> "contains"
719 let query_fun = function
720 | `Prefix -> text_prefix
721 | `Suffix -> text_suffix
722 | `Equals -> text_equals
723 | `Contains -> text_contains
726 let _pred_cache = Hashtbl.create 17
728 let mk_pred query s =
729 let f = query_fun query in
730 let memo = ref (fun _ _ -> failwith "Undefined") in
731 memo := begin fun tree node ->
733 try Hashtbl.find _pred_cache (query,s) with
735 time ~count:1 ~msg:(Printf.sprintf "Computing text query %s(%s)"
736 (string_of_query query) s)
739 let bv = results.bv in
740 memo := begin fun _ n ->
742 bit_vector_unsafe_get bv (Node.to_int n)
744 D_TRACE_(Printf.eprintf "Result of memoized call to query %s is %b for node %i\n" s b (Node.to_int n));
747 let b = bit_vector_unsafe_get bv (Node.to_int node) in
748 D_TRACE_(Printf.eprintf "Result is %b for node %i\n" b (Node.to_int node));
754 let full_text_prefix t s = (text_prefix t s true).pos
756 let full_text_suffix t s = (text_suffix t s true).pos
758 let full_text_equals t s = (text_equals t s true).pos
760 let full_text_contains t s = (text_contains t s true).pos
762 let full_text_query q t s =
763 let res = (query_fun q) t s true in
764 Hashtbl.replace _pred_cache (q,s) res;