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
27 external create : unit -> t = "caml_xml_tree_builder_create"
28 external open_document : t -> bool -> int -> bool -> int -> unit = "caml_xml_tree_builder_open_document"
29 external close_document : t -> tree = "caml_xml_tree_builder_close_document"
30 external open_tag : t -> string -> unit = "caml_xml_tree_builder_new_open_tag"
31 external close_tag : t -> string -> unit = "caml_xml_tree_builder_new_closing_tag"
32 external text : t -> string -> unit = "caml_xml_tree_builder_new_text"
36 if Buffer.length t > 0 then begin
38 text b (Buffer.contents t);
43 let output_attr b name value =
44 let atname = "<@>" ^ name in
51 let start_element_handler b t tag attr_list =
58 List.iter (fun (name, value) -> output_attr b name value) l;
62 let end_element_handler b t tag =
66 let character_data_handler _b t text =
67 Buffer.add_string t text
69 let create_parser () =
70 let buf = Buffer.create 512 in
71 let build = create () in
72 let parser_ = Expat.parser_create ~encoding:None in
78 Expat.set_start_element_handler parser_ (start_element_handler build buf);
79 Expat.set_end_element_handler parser_ (end_element_handler build buf);
80 Expat.set_character_data_handler parser_ (character_data_handler build buf);
81 open_document build !Options.index_empty_texts !Options.sample_factor
82 !Options.disable_text_collection !Options.text_index_type;
87 let parser_, finalizer = create_parser () in
88 Expat.parse parser_ s;
92 let in_chan = open_in file in
93 let buffer = String.create 4096 in
94 let parser_, finalizer = create_parser () in
98 let read = input in_chan buffer 0 4096 in
99 if read == 0 then raise End_of_file else
100 Expat.parse_sub parser_ buffer 0 read;
104 | End_of_file -> close_in in_chan
116 type bit_vector = string
118 external bool_of_int : int -> bool = "%identity"
120 let bit_vector_unsafe_get v i =
122 (((Char.code (String.unsafe_get v (i lsr 3))) lsr (i land 7)) land 1)
126 elements: Ptset.Int.t;
127 attributes: Ptset.Int.t;
128 attribute_array : Tag.t array;
129 children : Ptset.Int.t array;
130 siblings : Ptset.Int.t array;
131 descendants: Ptset.Int.t array;
132 followings: Ptset.Int.t array;
138 external parse_xml_uri : string -> int -> bool -> bool -> int -> tree = "caml_call_shredder_uri"
139 external parse_xml_string : string -> int -> bool -> bool -> int -> tree = "caml_call_shredder_string"
141 external tree_print_xml_fast3 : tree -> [`Tree ] Node.t -> Unix.file_descr -> unit = "caml_xml_tree_print"
142 let print_xml t n fd =
143 tree_print_xml_fast3 t.doc n fd
146 external tree_save : tree -> Unix.file_descr -> string -> unit = "caml_xml_tree_save"
147 external tree_load : Unix.file_descr -> string -> bool -> int -> tree = "caml_xml_tree_load"
149 external nullt : unit -> 'a Node.t = "caml_xml_tree_nullt"
151 let nil : [`Tree ] Node.t = Node.nil
152 let root : [`Tree ] Node.t = Node.null
156 external unordered_set_alloc : int -> unordered_set = "caml_unordered_set_alloc"
157 external unordered_set_length : unordered_set -> int = "caml_unordered_set_length"
158 external unordered_set_insert : unordered_set -> int -> unit = "caml_unordered_set_set" "noalloc"
160 module HPtset = Hashtbl.Make(Ptset.Int)
162 let vector_htbl = HPtset.create MED_H_SIZE
164 let unordered_set_of_set s =
166 HPtset.find vector_htbl s
169 let v = unordered_set_alloc (Ptset.Int.cardinal s) in
170 let _ = Ptset.Int.iter (fun e -> unordered_set_insert v e) s in
171 HPtset.add vector_htbl s v; v
173 let ptset_to_vector = unordered_set_of_set
175 (** tree interface *)
177 external tree_root : tree -> [`Tree] Node.t = "caml_xml_tree_root" "noalloc"
180 external tree_first_child : tree -> [`Tree] Node.t -> [`Tree] Node.t = "caml_xml_tree_first_child" "noalloc"
181 let first_child t n = tree_first_child t.doc n
183 external tree_first_element : tree -> [`Tree] Node.t -> [`Tree] Node.t = "caml_xml_tree_first_element" "noalloc"
184 let first_element t n = tree_first_element t.doc n
186 external tree_tagged_child : tree -> [`Tree] Node.t -> Tag.t -> [`Tree] Node.t = "caml_xml_tree_tagged_child" "noalloc"
187 let tagged_child t n tag = tree_tagged_child t.doc n tag
189 external tree_select_child : tree -> [`Tree ] Node.t -> unordered_set -> [`Tree] Node.t = "caml_xml_tree_select_child" "noalloc"
190 let select_child t n tag_set = tree_select_child t.doc n tag_set
192 external tree_last_child : tree -> [`Tree] Node.t -> [`Tree] Node.t = "caml_xml_tree_last_child" "noalloc"
193 let last_child t n = tree_last_child t.doc n
197 external tree_next_sibling : tree -> [`Tree] Node.t -> [`Tree] Node.t = "caml_xml_tree_next_sibling" "noalloc"
198 let next_sibling t n = tree_next_sibling t.doc n
200 external tree_next_element : tree -> [`Tree] Node.t -> [`Tree] Node.t = "caml_xml_tree_next_element" "noalloc"
201 let next_element t n = tree_next_element t.doc n
203 external tree_next_node_before : tree -> [`Tree] Node.t -> [`Tree] Node.t -> [`Tree] Node.t = "caml_xml_tree_next_node_before" "noalloc"
204 let next_node_before t n ctx = tree_next_node_before t.doc n ctx
206 external tree_tagged_following_sibling : tree -> [`Tree] Node.t -> Tag.t -> [`Tree] Node.t = "caml_xml_tree_tagged_following_sibling" "noalloc"
207 let tagged_following_sibling t n tag = tree_tagged_following_sibling t.doc n tag
209 external tree_select_following_sibling : tree -> [`Tree ] Node.t -> unordered_set -> [`Tree] Node.t = "caml_xml_tree_select_following_sibling" "noalloc"
210 let select_following_sibling t n tag_set = tree_select_following_sibling t.doc n tag_set
212 external tree_prev_sibling : tree -> [`Tree] Node.t -> [`Tree] Node.t = "caml_xml_tree_prev_sibling" "noalloc"
213 let prev_sibling t n = tree_prev_sibling t.doc n
217 external tree_tagged_descendant : tree -> [`Tree ] Node.t -> Tag.t -> [`Tree ] Node.t = "caml_xml_tree_tagged_descendant" "noalloc"
218 let tagged_descendant t n tag = tree_tagged_descendant t.doc n tag
220 external tree_tagged_next : tree -> [`Tree ] Node.t -> Tag.t -> [`Tree ] Node.t = "caml_xml_tree_tagged_next" "noalloc"
221 let tagged_next t n tag = tree_tagged_next t.doc n tag
223 external tree_select_descendant : tree -> [`Tree ] Node.t -> unordered_set -> [`Tree] Node.t = "caml_xml_tree_select_descendant" "noalloc"
224 let select_descendant t n tag_set = tree_select_descendant t.doc n tag_set
226 external tree_tagged_following_before : tree -> [`Tree ] Node.t -> Tag.t -> [`Tree ] Node.t -> [`Tree ] Node.t = "caml_xml_tree_tagged_following_before" "noalloc"
227 let tagged_following_before t n tag ctx = tree_tagged_following_before t.doc n tag ctx
229 external tree_select_following_before : tree -> [`Tree ] Node.t -> unordered_set -> [`Tree] Node.t -> [`Tree] Node.t = "caml_xml_tree_select_following_before" "noalloc"
230 let select_following_before t n tag_set ctx = tree_select_following_before t.doc n tag_set ctx
232 external tree_parent : tree -> [`Tree] Node.t -> [`Tree] Node.t = "caml_xml_tree_parent" "noalloc"
233 let parent t n = tree_parent t.doc n
235 external tree_binary_parent : tree -> [`Tree] Node.t -> [`Tree] Node.t = "caml_xml_tree_binary_parent"
237 let binary_parent t n = tree_binary_parent t.doc n
240 external tree_tag : tree -> [`Tree] Node.t -> Tag.t = "caml_xml_tree_tag" "noalloc"
241 let tag t n = tree_tag t.doc n
243 external tree_is_first_child : tree -> [ `Tree ] Node.t -> bool = "caml_xml_tree_is_first_child" "noalloc"
244 let is_first_child t n = tree_is_first_child t.doc n
246 external tree_is_right_descendant : tree -> [ `Tree ] Node.t -> [`Tree] Node.t -> bool =
247 "caml_xml_tree_is_right_descendant" "noalloc"
249 let is_right_descendant t n1 n2 = tree_is_right_descendant t.doc n1 n2
252 let node_tags t = Ptset.Int.add Tag.document_node t.descendants.(Tag.document_node)
254 let attribute_tags t = t.attributes
256 let element_tags t = t.elements
259 t.children.(tag), t.descendants.(tag), t.siblings.(tag), t.followings.(tag)
262 let dump_tag_table t =
263 eprintf "Child tags:\n%!";
265 (fun tag set -> eprintf "%s: %a\n%!"
266 (Tag.to_string tag) TagSet.print (TagSet.inj_positive set))
268 eprintf "-----------------------------\n%!";
269 eprintf "Descendant tags:\n%!";
271 (fun tag set -> eprintf "%s: %a\n%!"
272 (Tag.to_string tag) TagSet.print (TagSet.inj_positive set))
274 eprintf "-----------------------------\n%!";
275 eprintf "Sibling tags:\n%!";
277 (fun tag set -> eprintf "%s: %a\n%!"
278 (Tag.to_string tag) TagSet.print (TagSet.inj_positive set))
280 eprintf "-----------------------------\n%!";
281 eprintf "Following tags:\n%!";
283 (fun tag set -> eprintf "%s: %a\n%!"
284 (Tag.to_string tag) TagSet.print (TagSet.inj_positive set))
286 eprintf "-----------------------------\n%!"
289 external tree_subtree_tags : tree -> [`Tree] Node.t -> Tag.t -> int = "caml_xml_tree_subtree_tags" "noalloc"
290 let subtree_tags t n tag = tree_subtree_tags t.doc n tag
292 external tree_subtree_size : tree -> [`Tree] Node.t -> int = "caml_xml_tree_subtree_size" "noalloc"
293 let subtree_size t n = tree_subtree_size t.doc n
295 let subtree_elements t node =
296 let size = tree_subtree_size t.doc node - 1 in
298 else let size = size - (tree_subtree_tags t.doc node Tag.pcdata) in
299 if size < 2 then size else
300 let acc = ref size in
301 for i = 0 to Array.length t.attribute_array - 1 do
302 acc := !acc - tree_subtree_tags t.doc node t.attribute_array.(i)
306 external tree_closing : tree -> [`Tree] Node.t -> [`Tree] Node.t = "caml_xml_tree_closing" "noalloc"
307 let closing t n = tree_closing t.doc n
309 external tree_num_tags : tree -> int = "caml_xml_tree_num_tags" "noalloc"
310 let num_tags t = tree_num_tags t.doc
312 external tree_size : tree -> int = "caml_xml_tree_size" "noalloc"
313 let size t = tree_size t.doc
318 let rec loop left node acc_d total_d num_leaves =
320 (acc_d+total_d,if left then num_leaves+1 else num_leaves)
322 let d,td = loop true (tree_first_child tree node) (acc_d+1) total_d num_leaves in
323 loop false (tree_next_sibling tree node) (acc_d) d td
325 let a,b = loop true root 0 0 0
327 Printf.eprintf "Average depth: %f, number of leaves %i\n%!" ((float_of_int a)/. (float_of_int b)) b
335 external hash : t -> int = "%identity"
336 external uid : t -> Uid.t = "%identity"
337 external equal : t -> t -> bool = "%eq"
338 external make : t -> int = "%identity"
339 external node : t -> int = "%identity"
340 external stats : unit -> unit = "%identity"
343 let to_ptset s = fold (Ptset.Int.add) s Ptset.Int.empty
347 Hashtbl.Make(struct type t = TagS.t * TagS.t
349 HASHINT2(Uid.to_int x.TagS.Node.id,
350 Uid.to_int y.TagS.Node.id)
357 Hashtbl.Make(struct type t = Tag.t * TagS.t
359 HASHINT2(x, Uid.to_int y.TagS.Node.id)
366 let add_cache = TagTSCache.create 1023
367 let union_cache = TSTSCache.create 1023
368 let subset_cache = TSTSCache.create 1023
371 TSTSCache.clear union_cache;
372 TSTSCache.clear subset_cache;
373 TagTSCache.clear add_cache
376 (x == y) || (x == TagS.empty) ||
377 if y == TagS.empty then false
381 TSTSCache.find subset_cache key
384 let z = TagS.subset x y in
385 TSTSCache.add subset_cache key z; z
387 let order ((x, y) as z) =
388 if x.TagS.Node.id <= y.TagS.Node.id then z
392 if _subset x y then y
393 else if _subset y x then x
395 let key = order (x, y) in
397 TSTSCache.find union_cache key
400 let z = TagS.union x y in
401 TSTSCache.add union_cache key z; z
406 TagTSCache.find add_cache key
409 let z = TagS.add t s in
410 TagTSCache.add add_cache key z;z
412 let child_sibling_labels tree =
413 let table_c = Array.create (tree_num_tags tree) TagS.empty in
414 let table_n = Array.copy table_c in
416 if node == nil then TagS.empty
418 let children = loop (tree_first_child tree node) in
419 let tag = tree_tag tree node in
421 let tc = table_c.(tag) in
422 if _subset children tc then ()
423 else table_c.(tag) <- _union tc children
425 let siblings = loop (tree_next_sibling tree node) in
427 let tn = table_n.(tag) in
428 if _subset siblings tn then ()
429 else table_n.(tag) <- _union tn siblings
436 let descendant_labels tree =
437 let table_d = Array.create (tree_num_tags tree) TagS.empty in
439 if node == nil then TagS.empty else
440 let d1 = loop (tree_first_child tree node) in
441 let d2 = loop (tree_next_sibling tree node) in
442 let tag = tree_tag tree node in
444 let td = table_d.(tag) in
445 if _subset d1 td then ()
446 else table_d.(tag) <- _union td d1;
448 _add tag (_union d1 d2)
453 let collect_labels tree =
454 let table_f = Array.create (tree_num_tags tree) TagS.empty in
455 let table_n = Array.copy table_f in
456 let table_c = Array.copy table_f in
457 let table_d = Array.copy table_f in
458 let rec loop node foll_siblings descendants followings =
459 if node == nil then foll_siblings, descendants, followings else
460 let tag = tree_tag tree node in
462 let tf = table_f.(tag) in
463 if _subset followings tf then ()
464 else table_f.(tag) <- _union tf followings in
466 let tn = table_n.(tag) in
467 if _subset foll_siblings tn then ()
468 else table_n.(tag) <- _union tn foll_siblings in
469 let children, n_descendants, n_followings =
470 loop (tree_last_child tree node) TagS.empty TagS.empty followings
473 let tc = table_c.(tag) in
474 if _subset children tc then ()
475 else table_c.(tag) <- _union tc children
478 let td = table_d.(tag) in
479 if _subset n_descendants td then ()
480 else table_d.(tag) <- _union td n_descendants
482 loop (tree_prev_sibling tree node)
483 (_add tag foll_siblings)
484 (_add tag (_union n_descendants descendants))
485 (_add tag n_followings)
487 ignore (loop root TagS.empty TagS.empty TagS.empty);
488 table_f, table_n, table_c, table_d
491 let is_nil t = t == nil
492 let is_node t = t != nil
493 let is_root t = t == root
496 let _ = Tag.init (Obj.magic t) in
497 let f, n, c, d = time collect_labels t ~msg:"Building tag relationship table" in
498 let c = Array.map TagS.to_ptset c in
499 let n = Array.map TagS.to_ptset n in
500 let f = Array.map TagS.to_ptset f in
501 let d = Array.map TagS.to_ptset d in
502 let () = clear_cache () in
503 let attributes = Ptset.Int.add Tag.attribute d.(Tag.attribute) in
504 let elements = Ptset.Int.add Tag.document_node
505 (Ptset.Int.remove Tag.pcdata
506 (Ptset.Int.diff d.(Tag.document_node) attributes))
509 attributes = attributes;
510 attribute_array = Array.of_list (Ptset.Int.elements attributes);
520 let parse_xml_uri str = node_of_t (TreeBuilder.parse_file str)
521 let parse_xml_string str = node_of_t (TreeBuilder.parse_string str)
523 let size t = tree_size t.doc;;
525 external pool : tree -> Tag.pool = "%identity"
527 let magic_string = "SXSI_INDEX"
528 let version_string = "3"
531 Unix.lseek fd 0 Unix.SEEK_CUR
533 let pr_pos fd = Printf.eprintf "At position %i\n%!" (pos fd)
536 let sl = String.length s in
537 let ssl = Printf.sprintf "%020i" sl in
538 ignore (Unix.write fd ssl 0 20);
539 ignore (Unix.write fd s 0 (String.length s))
541 let rec really_read fd buffer start length =
542 if length <= 0 then () else
543 match Unix.read fd buffer start length with
544 0 -> raise End_of_file
545 | r -> really_read fd buffer (start + r) (length - r);;
548 let buffer = String.create 20 in
549 let _ = really_read fd buffer 0 20 in
550 let size = int_of_string buffer in
551 let buffer = String.create size in
552 let _ = really_read fd buffer 0 size in
555 let save_tag_table channel t =
556 let t = Array.map (fun s -> Array.of_list (Ptset.Int.elements s)) t in
557 Marshal.to_channel channel t []
560 let fd = Unix.openfile str [ Unix.O_WRONLY;Unix.O_TRUNC;Unix.O_CREAT] 0o644 in
561 let out_c = Unix.out_channel_of_descr fd in
562 let _ = set_binary_mode_out out_c true in
563 output_string out_c magic_string;
564 output_char out_c '\n';
565 output_string out_c version_string;
566 output_char out_c '\n';
567 save_tag_table out_c t.children;
568 save_tag_table out_c t.siblings;
569 save_tag_table out_c t.descendants;
570 save_tag_table out_c t.followings;
571 (* we need to move the fd to the correct position *)
573 ignore (Unix.lseek fd (pos_out out_c) Unix.SEEK_SET);
574 tree_save t.doc fd str;
577 let load_tag_table channel =
578 let table : int array array = Marshal.from_channel channel in
579 Array.map (fun a -> Ptset.Int.from_list (Array.to_list a)) table
581 let load ?(sample=64) ?(load_text=true) str =
582 let fd = Unix.openfile str [ Unix.O_RDONLY ] 0o644 in
583 let in_c = Unix.in_channel_of_descr fd in
584 let _ = set_binary_mode_in in_c true in
586 (let ms = input_line in_c in if ms <> magic_string then failwith "Invalid index file");
587 (let vs = input_line in_c in if vs <> version_string then failwith "Invalid version file");
588 let c = load_tag_table in_c in
589 let s = load_tag_table in_c in
590 let d = load_tag_table in_c in
591 let f = load_tag_table in_c in
594 let c, s, d, f = time ~msg:"Loading tag table"(load_table) () in
595 ignore(Unix.lseek fd (pos_in in_c) Unix.SEEK_SET);
596 let xml_tree = tree_load fd str load_text sample in
597 let () = Tag.init (Obj.magic xml_tree) in
598 let attributes = Ptset.Int.add Tag.attribute d.(Tag.attribute) in
599 let elements = Ptset.Int.add Tag.document_node
600 (Ptset.Int.remove Tag.pcdata
601 (Ptset.Int.diff d.(Tag.document_node) attributes))
603 let tree = { doc = xml_tree;
604 attributes = attributes;
605 attribute_array = Array.of_list (Ptset.Int.elements attributes);
618 let tag_pool t = pool t.doc
620 let equal a b = a == b
624 | i -> Printf.sprintf "Node (%i)" i
626 let dump_node t = nts (Node.to_int t)
630 type query_result = { bv : bit_vector;
634 external tree_flush : tree -> Unix.file_descr -> unit = "caml_xml_tree_flush"
635 let flush t fd = tree_flush t.doc fd
637 external text_prefix : tree -> string -> query_result = "caml_text_collection_prefix_bv"
638 let text_prefix t s = text_prefix t.doc s
640 external text_suffix : tree -> string -> query_result = "caml_text_collection_suffix_bv"
641 let text_suffix t s = text_suffix t.doc s
643 external text_equals : tree -> string -> query_result = "caml_text_collection_equals_bv"
644 let text_equals t s = text_equals t.doc s
646 external text_contains : tree -> string -> query_result = "caml_text_collection_contains_bv"
647 let text_contains t s = text_contains t.doc s
650 module Predicate = Hcons.Make (
653 type t = (_t -> node -> bool) ref
654 let hash t = Hashtbl.hash t
655 let equal t1 t2 = t1 == t2
658 let string_of_query query =
660 | `Prefix -> "starts-with"
661 | `Suffix -> "ends-with"
662 | `Equals -> "equals"
663 | `Contains -> "contains"
666 let query_fun = function
667 | `Prefix -> text_prefix
668 | `Suffix -> text_suffix
669 | `Equals -> text_equals
670 | `Contains -> text_contains
673 let _pred_cache = Hashtbl.create 17
675 let mk_pred query s =
676 let f = query_fun query in
677 let memo = ref (fun _ _ -> failwith "Undefined") in
678 memo := begin fun tree node ->
680 try Hashtbl.find _pred_cache (query,s) with
682 time ~count:1 ~msg:(Printf.sprintf "Computing text query %s(%s)"
683 (string_of_query query) s)
686 let bv = results.bv in
687 memo := begin fun _ n ->
689 bit_vector_unsafe_get bv (Node.to_int n)
691 D_TRACE_(Printf.eprintf "Result of memoized call to query %s is %b for node %i\n" s b (Node.to_int n));
694 let b = bit_vector_unsafe_get bv (Node.to_int node) in
695 D_TRACE_(Printf.eprintf "Result is %b for node %i\n" b (Node.to_int node));
701 let full_text_prefix t s = (text_prefix t s).pos
703 let full_text_suffix t s = (text_suffix t s).pos
705 let full_text_equals t s = (text_equals t s).pos
707 let full_text_contains t s = (text_contains t s).pos
709 let full_text_query q t s =
710 let res = (query_fun q) t s in
711 Hashtbl.replace _pred_cache (q,s) res;