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 = "caml_init_lib"
13 exception CPlusPlusError of string
15 let () = Callback.register_exception "CPlusPlusError" (CPlusPlusError "")
19 type node = [ `Tree ] Node.t
23 type bit_vector = string
25 external bool_of_int : int -> bool = "%identity"
27 let bit_vector_unsafe_get v i =
29 (((Char.code (String.unsafe_get v (i lsr 3))) lsr (i land 7)) land 1)
33 elements: Ptset.Int.t;
34 attributes: Ptset.Int.t;
35 attribute_array : Tag.t array;
36 children : Ptset.Int.t array;
37 siblings : Ptset.Int.t array;
38 descendants: Ptset.Int.t array;
39 followings: Ptset.Int.t array;
44 external parse_xml_uri : string -> int -> bool -> bool -> int -> tree = "caml_call_shredder_uri"
45 external parse_xml_string : string -> int -> bool -> bool -> int -> tree = "caml_call_shredder_string"
46 external tree_print_xml_fast3 : tree -> [`Tree ] Node.t -> Unix.file_descr -> unit = "caml_xml_tree_print"
47 let print_xml t n fd =
48 tree_print_xml_fast3 t.doc n fd
51 external tree_save : tree -> Unix.file_descr -> string -> unit = "caml_xml_tree_save"
52 external tree_load : Unix.file_descr -> string -> bool -> int -> tree = "caml_xml_tree_load"
54 external nullt : unit -> 'a Node.t = "caml_xml_tree_nullt"
56 let nil : [`Tree ] Node.t = Node.nil
57 let root : [`Tree ] Node.t = Node.null
61 external unordered_set_alloc : int -> unordered_set = "caml_unordered_set_alloc"
62 external unordered_set_length : unordered_set -> int = "caml_unordered_set_length"
63 external unordered_set_insert : unordered_set -> int -> unit = "caml_unordered_set_set" "noalloc"
65 module HPtset = Hashtbl.Make(Ptset.Int)
67 let vector_htbl = HPtset.create MED_H_SIZE
69 let unordered_set_of_set s =
71 HPtset.find vector_htbl s
74 let v = unordered_set_alloc (Ptset.Int.cardinal s) in
75 let _ = Ptset.Int.iter (fun e -> unordered_set_insert v e) s in
76 HPtset.add vector_htbl s v; v
78 let ptset_to_vector = unordered_set_of_set
82 external tree_root : tree -> [`Tree] Node.t = "caml_xml_tree_root" "noalloc"
85 external tree_first_child : tree -> [`Tree] Node.t -> [`Tree] Node.t = "caml_xml_tree_first_child" "noalloc"
86 let first_child t n = tree_first_child t.doc n
88 external tree_first_element : tree -> [`Tree] Node.t -> [`Tree] Node.t = "caml_xml_tree_first_element" "noalloc"
89 let first_element t n = tree_first_element t.doc n
91 external tree_tagged_child : tree -> [`Tree] Node.t -> Tag.t -> [`Tree] Node.t = "caml_xml_tree_tagged_child" "noalloc"
92 let tagged_child t n tag = tree_tagged_child t.doc n tag
94 external tree_select_child : tree -> [`Tree ] Node.t -> unordered_set -> [`Tree] Node.t = "caml_xml_tree_select_child" "noalloc"
95 let select_child t n tag_set = tree_select_child t.doc n tag_set
97 external tree_last_child : tree -> [`Tree] Node.t -> [`Tree] Node.t = "caml_xml_tree_last_child" "noalloc"
98 let last_child t n = tree_last_child t.doc n
102 external tree_next_sibling : tree -> [`Tree] Node.t -> [`Tree] Node.t = "caml_xml_tree_next_sibling" "noalloc"
103 let next_sibling t n = tree_next_sibling t.doc n
105 external tree_next_element : tree -> [`Tree] Node.t -> [`Tree] Node.t = "caml_xml_tree_next_element" "noalloc"
106 let next_element t n = tree_next_element t.doc n
108 external tree_tagged_following_sibling : tree -> [`Tree] Node.t -> Tag.t -> [`Tree] Node.t = "caml_xml_tree_tagged_following_sibling" "noalloc"
109 let tagged_following_sibling t n tag = tree_tagged_following_sibling t.doc n tag
111 external tree_select_following_sibling : tree -> [`Tree ] Node.t -> unordered_set -> [`Tree] Node.t = "caml_xml_tree_select_following_sibling" "noalloc"
112 let select_following_sibling t n tag_set = tree_select_following_sibling t.doc n tag_set
114 external tree_prev_sibling : tree -> [`Tree] Node.t -> [`Tree] Node.t = "caml_xml_tree_prev_sibling" "noalloc"
115 let prev_sibling t n = tree_prev_sibling t.doc n
119 external tree_tagged_descendant : tree -> [`Tree ] Node.t -> Tag.t -> [`Tree ] Node.t = "caml_xml_tree_tagged_descendant" "noalloc"
120 let tagged_descendant t n tag = tree_tagged_descendant t.doc n tag
122 external tree_tagged_next : tree -> [`Tree ] Node.t -> Tag.t -> [`Tree ] Node.t = "caml_xml_tree_tagged_next" "noalloc"
123 let tagged_next t n tag = tree_tagged_next t.doc n tag
125 external tree_select_descendant : tree -> [`Tree ] Node.t -> unordered_set -> [`Tree] Node.t = "caml_xml_tree_select_descendant" "noalloc"
126 let select_descendant t n tag_set = tree_select_descendant t.doc n tag_set
128 external tree_tagged_following_before : tree -> [`Tree ] Node.t -> Tag.t -> [`Tree ] Node.t -> [`Tree ] Node.t = "caml_xml_tree_tagged_following_before" "noalloc"
129 let tagged_following_before t n tag ctx = tree_tagged_following_before t.doc n tag ctx
131 external tree_select_following_before : tree -> [`Tree ] Node.t -> unordered_set -> [`Tree] Node.t -> [`Tree] Node.t = "caml_xml_tree_select_following_before" "noalloc"
132 let select_following_before t n tag_set ctx = tree_select_following_before t.doc n tag_set ctx
134 external tree_parent : tree -> [`Tree] Node.t -> [`Tree] Node.t = "caml_xml_tree_parent" "noalloc"
135 let parent t n = tree_parent t.doc n
137 external tree_binary_parent : tree -> [`Tree] Node.t -> [`Tree] Node.t = "caml_xml_tree_binary_parent"
139 let binary_parent t n = tree_binary_parent t.doc n
142 external tree_tag : tree -> [`Tree] Node.t -> Tag.t = "caml_xml_tree_tag" "noalloc"
143 let tag t n = tree_tag t.doc n
145 external tree_is_first_child : tree -> [ `Tree ] Node.t -> bool = "caml_xml_tree_is_first_child" "noalloc"
146 let is_first_child t n = tree_is_first_child t.doc n
148 external tree_is_right_descendant : tree -> [ `Tree ] Node.t -> [`Tree] Node.t -> bool =
149 "caml_xml_tree_is_right_descendant" "noalloc"
151 let is_right_descendant t n1 n2 = tree_is_right_descendant t.doc n1 n2
154 let node_tags t = Ptset.Int.add Tag.document_node t.descendants.(Tag.document_node)
156 let attribute_tags t = t.attributes
158 let element_tags t = t.elements
161 t.children.(tag), t.descendants.(tag), t.siblings.(tag), t.followings.(tag)
164 let dump_tag_table t =
165 eprintf "Child tags:\n%!";
167 (fun tag set -> eprintf "%s: %a\n%!"
168 (Tag.to_string tag) TagSet.print (TagSet.inj_positive set))
170 eprintf "-----------------------------\n%!";
171 eprintf "Descendant tags:\n%!";
173 (fun tag set -> eprintf "%s: %a\n%!"
174 (Tag.to_string tag) TagSet.print (TagSet.inj_positive set))
176 eprintf "-----------------------------\n%!";
177 eprintf "Sibling tags:\n%!";
179 (fun tag set -> eprintf "%s: %a\n%!"
180 (Tag.to_string tag) TagSet.print (TagSet.inj_positive set))
182 eprintf "-----------------------------\n%!";
183 eprintf "Following tags:\n%!";
185 (fun tag set -> eprintf "%s: %a\n%!"
186 (Tag.to_string tag) TagSet.print (TagSet.inj_positive set))
188 eprintf "-----------------------------\n%!"
191 external tree_subtree_tags : tree -> [`Tree] Node.t -> Tag.t -> int = "caml_xml_tree_subtree_tags" "noalloc"
192 let subtree_tags t n tag = tree_subtree_tags t.doc n tag
194 external tree_subtree_size : tree -> [`Tree] Node.t -> int = "caml_xml_tree_subtree_size" "noalloc"
195 let subtree_size t n = tree_subtree_size t.doc n
197 let subtree_elements t node =
198 let size = tree_subtree_size t.doc node - 1 in
200 else let size = size - (tree_subtree_tags t.doc node Tag.pcdata) in
201 if size < 2 then size else
202 let acc = ref size in
203 for i = 0 to Array.length t.attribute_array - 1 do
204 acc := !acc - tree_subtree_tags t.doc node t.attribute_array.(i)
208 external tree_closing : tree -> [`Tree] Node.t -> [`Tree] Node.t = "caml_xml_tree_closing" "noalloc"
209 let closing t n = tree_closing t.doc n
211 external tree_num_tags : tree -> int = "caml_xml_tree_num_tags" "noalloc"
212 let num_tags t = tree_num_tags t.doc
214 external tree_size : tree -> int = "caml_xml_tree_size" "noalloc"
215 let size t = tree_size t.doc
220 let rec loop left node acc_d total_d num_leaves =
222 (acc_d+total_d,if left then num_leaves+1 else num_leaves)
224 let d,td = loop true (tree_first_child tree node) (acc_d+1) total_d num_leaves in
225 loop false (tree_next_sibling tree node) (acc_d) d td
227 let a,b = loop true root 0 0 0
229 Printf.eprintf "Average depth: %f, number of leaves %i\n%!" ((float_of_int a)/. (float_of_int b)) b
237 external hash : t -> int = "%identity"
238 external uid : t -> Uid.t = "%identity"
239 external equal : t -> t -> bool = "%eq"
240 external make : t -> int = "%identity"
241 external node : t -> int = "%identity"
242 external stats : unit -> unit = "%identity"
245 let to_ptset s = fold (Ptset.Int.add) s Ptset.Int.empty
249 Hashtbl.Make(struct type t = TagS.t * TagS.t
251 HASHINT2(Uid.to_int x.TagS.Node.id,
252 Uid.to_int y.TagS.Node.id)
259 Hashtbl.Make(struct type t = Tag.t * TagS.t
261 HASHINT2(x, Uid.to_int y.TagS.Node.id)
268 let add_cache = TagTSCache.create 1023
269 let union_cache = TSTSCache.create 1023
270 let subset_cache = TSTSCache.create 1023
273 TSTSCache.clear union_cache;
274 TSTSCache.clear subset_cache;
275 TagTSCache.clear add_cache
278 (x == y) || (x == TagS.empty) ||
279 if y == TagS.empty then false
283 TSTSCache.find subset_cache key
286 let z = TagS.subset x y in
287 TSTSCache.add subset_cache key z; z
289 let order ((x, y) as z) =
290 if x.TagS.Node.id <= y.TagS.Node.id then z
294 if _subset x y then y
295 else if _subset y x then x
297 let key = order (x, y) in
299 TSTSCache.find union_cache key
302 let z = TagS.union x y in
303 TSTSCache.add union_cache key z; z
308 TagTSCache.find add_cache key
311 let z = TagS.add t s in
312 TagTSCache.add add_cache key z;z
314 let child_sibling_labels tree =
315 let table_c = Array.create (tree_num_tags tree) TagS.empty in
316 let table_n = Array.copy table_c in
318 if node == nil then TagS.empty
320 let children = loop (tree_first_child tree node) in
321 let tag = tree_tag tree node in
323 let tc = table_c.(tag) in
324 if _subset children tc then ()
325 else table_c.(tag) <- _union tc children
327 let siblings = loop (tree_next_sibling tree node) in
329 let tn = table_n.(tag) in
330 if _subset siblings tn then ()
331 else table_n.(tag) <- _union tn siblings
338 let descendant_labels tree =
339 let table_d = Array.create (tree_num_tags tree) TagS.empty in
341 if node == nil then TagS.empty else
342 let d1 = loop (tree_first_child tree node) in
343 let d2 = loop (tree_next_sibling tree node) in
344 let tag = tree_tag tree node in
346 let td = table_d.(tag) in
347 if _subset d1 td then ()
348 else table_d.(tag) <- _union td d1;
350 _add tag (_union d1 d2)
355 let collect_labels tree =
356 let table_f = Array.create (tree_num_tags tree) TagS.empty in
357 let table_n = Array.copy table_f in
358 let table_c = Array.copy table_f in
359 let table_d = Array.copy table_f in
360 let rec loop node foll_siblings descendants followings =
361 if node == nil then foll_siblings, descendants, followings else
362 let tag = tree_tag tree node in
364 let tf = table_f.(tag) in
365 if _subset followings tf then ()
366 else table_f.(tag) <- _union tf followings in
368 let tn = table_n.(tag) in
369 if _subset foll_siblings tn then ()
370 else table_n.(tag) <- _union tn foll_siblings in
371 let children, n_descendants, n_followings =
372 loop (tree_last_child tree node) TagS.empty TagS.empty followings
375 let tc = table_c.(tag) in
376 if _subset children tc then ()
377 else table_c.(tag) <- _union tc children
380 let td = table_d.(tag) in
381 if _subset n_descendants td then ()
382 else table_d.(tag) <- _union td n_descendants
384 loop (tree_prev_sibling tree node)
385 (_add tag foll_siblings)
386 (_add tag (_union n_descendants descendants))
387 (_add tag n_followings)
389 ignore (loop root TagS.empty TagS.empty TagS.empty);
390 table_f, table_n, table_c, table_d
393 let is_nil t = t == nil
394 let is_node t = t != nil
395 let is_root t = t == root
398 let _ = Tag.init (Obj.magic t) in
399 let f, n, c, d = time collect_labels t ~msg:"Building tag relationship table" in
400 let c = Array.map TagS.to_ptset c in
401 let n = Array.map TagS.to_ptset n in
402 let f = Array.map TagS.to_ptset f in
403 let d = Array.map TagS.to_ptset d in
404 let () = clear_cache () in
405 let attributes = Ptset.Int.add Tag.attribute d.(Tag.attribute) in
406 let elements = Ptset.Int.add Tag.document_node
407 (Ptset.Int.remove Tag.pcdata
408 (Ptset.Int.diff d.(Tag.document_node) attributes))
411 attributes = attributes;
412 attribute_array = Array.of_list (Ptset.Int.elements attributes);
423 (f str !Options.sample_factor
424 !Options.index_empty_texts
425 !Options.disable_text_collection
426 !Options.text_index_type
429 let parse_xml_uri str = parse parse_xml_uri str
430 let parse_xml_string str = parse parse_xml_string str
432 let size t = tree_size t.doc;;
434 external pool : tree -> Tag.pool = "%identity"
436 let magic_string = "SXSI_INDEX"
437 let version_string = "3"
440 Unix.lseek fd 0 Unix.SEEK_CUR
442 let pr_pos fd = Printf.eprintf "At position %i\n%!" (pos fd)
445 let sl = String.length s in
446 let ssl = Printf.sprintf "%020i" sl in
447 ignore (Unix.write fd ssl 0 20);
448 ignore (Unix.write fd s 0 (String.length s))
450 let rec really_read fd buffer start length =
451 if length <= 0 then () else
452 match Unix.read fd buffer start length with
453 0 -> raise End_of_file
454 | r -> really_read fd buffer (start + r) (length - r);;
457 let buffer = String.create 20 in
458 let _ = really_read fd buffer 0 20 in
459 let size = int_of_string buffer in
460 let buffer = String.create size in
461 let _ = really_read fd buffer 0 size in
464 let save_tag_table channel t =
465 let t = Array.map (fun s -> Array.of_list (Ptset.Int.elements s)) t in
466 Marshal.to_channel channel t []
469 let fd = Unix.openfile str [ Unix.O_WRONLY;Unix.O_TRUNC;Unix.O_CREAT] 0o644 in
470 let out_c = Unix.out_channel_of_descr fd in
471 let _ = set_binary_mode_out out_c true in
472 output_string out_c magic_string;
473 output_char out_c '\n';
474 output_string out_c version_string;
475 output_char out_c '\n';
476 save_tag_table out_c t.children;
477 save_tag_table out_c t.siblings;
478 save_tag_table out_c t.descendants;
479 save_tag_table out_c t.followings;
480 (* we need to move the fd to the correct position *)
482 ignore (Unix.lseek fd (pos_out out_c) Unix.SEEK_SET);
483 tree_save t.doc fd str;
486 let load_tag_table channel =
487 let table : int array array = Marshal.from_channel channel in
488 Array.map (fun a -> Ptset.Int.from_list (Array.to_list a)) table
490 let load ?(sample=64) ?(load_text=true) str =
491 let fd = Unix.openfile str [ Unix.O_RDONLY ] 0o644 in
492 let in_c = Unix.in_channel_of_descr fd in
493 let _ = set_binary_mode_in in_c true in
495 (let ms = input_line in_c in if ms <> magic_string then failwith "Invalid index file");
496 (let vs = input_line in_c in if vs <> version_string then failwith "Invalid version file");
497 let c = load_tag_table in_c in
498 let s = load_tag_table in_c in
499 let d = load_tag_table in_c in
500 let f = load_tag_table in_c in
503 let c, s, d, f = time ~msg:"Loading tag table"(load_table) () in
504 ignore(Unix.lseek fd (pos_in in_c) Unix.SEEK_SET);
505 let xml_tree = tree_load fd str load_text sample in
506 let () = Tag.init (Obj.magic xml_tree) in
507 let attributes = Ptset.Int.add Tag.attribute d.(Tag.attribute) in
508 let elements = Ptset.Int.add Tag.document_node
509 (Ptset.Int.remove Tag.pcdata
510 (Ptset.Int.diff d.(Tag.document_node) attributes))
512 let tree = { doc = xml_tree;
513 attributes = attributes;
514 attribute_array = Array.of_list (Ptset.Int.elements attributes);
527 let tag_pool t = pool t.doc
529 let equal a b = a == b
533 | i -> Printf.sprintf "Node (%i)" i
535 let dump_node t = nts (Node.to_int t)
539 type query_result = { bv : bit_vector;
543 external tree_flush : tree -> Unix.file_descr -> unit = "caml_xml_tree_flush"
544 let flush t fd = tree_flush t.doc fd
546 external text_prefix : tree -> string -> query_result = "caml_text_collection_prefix_bv"
547 let text_prefix t s = text_prefix t.doc s
549 external text_suffix : tree -> string -> query_result = "caml_text_collection_suffix_bv"
550 let text_suffix t s = text_suffix t.doc s
552 external text_equals : tree -> string -> query_result = "caml_text_collection_equals_bv"
553 let text_equals t s = text_equals t.doc s
555 external text_contains : tree -> string -> query_result = "caml_text_collection_contains_bv"
556 let text_contains t s = text_contains t.doc s
559 module Predicate = Hcons.Make (
562 type t = (_t -> node -> bool) ref
563 let hash t = Hashtbl.hash t
564 let equal t1 t2 = t1 == t2
567 let string_of_query query =
569 | `Prefix -> "starts-with"
570 | `Suffix -> "ends-with"
571 | `Equals -> "equals"
572 | `Contains -> "contains"
575 let query_fun = function
576 | `Prefix -> text_prefix
577 | `Suffix -> text_suffix
578 | `Equals -> text_equals
579 | `Contains -> text_contains
582 let _pred_cache = Hashtbl.create 17
584 let mk_pred query s =
585 let f = query_fun query in
586 let memo = ref (fun _ _ -> failwith "Undefined") in
587 memo := begin fun tree node ->
589 try Hashtbl.find _pred_cache (query,s) with
591 time ~count:1 ~msg:(Printf.sprintf "Computing text query %s(%s)"
592 (string_of_query query) s)
595 let bv = results.bv in
596 memo := begin fun _ n ->
598 bit_vector_unsafe_get bv (Node.to_int n)
600 D_TRACE_(Printf.eprintf "Result of memoized call to query %s is %b for node %i\n" s b (Node.to_int n));
603 let b = bit_vector_unsafe_get bv (Node.to_int node) in
604 D_TRACE_(Printf.eprintf "Result is %b for node %i\n" b (Node.to_int node));
610 let full_text_prefix t s = (text_prefix t s).pos
612 let full_text_suffix t s = (text_suffix t s).pos
614 let full_text_equals t s = (text_equals t s).pos
616 let full_text_contains t s = (text_contains t s).pos
618 let full_text_query q t s =
619 let res = (query_fun q) t s in
620 Hashtbl.replace _pred_cache (q,s) res;