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 (******************************************************************************)
10 let () = init_timer();;
12 let enabled_gc = Gc.get()
13 let disabled_gc = { Gc.get() with
14 Gc.max_overhead = 1000000;
15 Gc.space_overhead = 100 }
16 let hash x = 131*x/(x-1+1)
18 let test_loop tree tag =
19 let t' = Tree.tagged_descendant tree tag Tree.root in
20 let f = Hashtbl.create 4096
22 let jump t _ = Tree.tagged_following_below tree tag t Tree.root in
24 if t == Tree.nil then 0
25 else 1+ ((Hashtbl.find f (hash 101)) (jump t ctx) ctx)
27 Hashtbl.add f (hash 101) g;
28 (Hashtbl.find f (hash 101)) t' Tree.root
31 let root = Tree.root in
32 let fin = Tree.closing tree root in
33 let rec loop t = if t <= fin then
34 let tag = Tree.tag tree t in
35 (* let _ = Tag.to_string tag in *)
36 if tag == Tag.pcdata then (ignore (Tree.get_text tree t));
37 let t = (Obj.magic ((Obj.magic t) + 1)) in
43 let test_loop2 tree tag =
44 let t' = Tree.tagged_descendant tree tag Tree.root in
45 let f = Hashtbl.create 4096
47 let jump t _ = Tree.tagged_following_below tree tag t Tree.root in
49 if t == Tree.nil then 0
50 else 1+ (match (Hashtbl.find f (hash 101)) with
51 `Foo ->g (jump t ctx) ctx
54 Hashtbl.add f (hash 101) `Foo;
58 let _ = Printf.eprintf "Contains(bree)" in
59 let _ = time (Tree.test_contains doc) "bree" in
60 let _ = Printf.eprintf "Contains(brain)" in
61 let _ = time (Tree.test_contains doc) "brain" in
62 let _ = Printf.eprintf "Contains(brain)" in
63 let i = time (Tree.test_contains doc) "brain" in
64 let _ = Printf.eprintf "%i\nContains(Australia)" i in
65 let i = time (Tree.test_contains doc) "AUSTRALIA" in
66 let _ = Printf.eprintf "%i\n Contains(1930)" i in
67 let i = time (Tree.test_contains doc) "1930" in
68 let _ = Printf.eprintf "%i\n startswith(bar)" i in
69 let i = time (Tree.test_prefix doc) "bar" in
70 let _ = Printf.eprintf "%i\n endswith(LAND)" i in
71 let i = time (Tree.test_suffix doc) "LAND" in
72 let _ = Printf.eprintf "%i\n =(2001)" i in
73 let i = time (Tree.test_equals doc) "2001" in
74 let _ = Printf.eprintf "%i\n =(Nguyen)" i in
75 let i = time (Tree.test_equals doc) "Nguyen" in
76 Printf.eprintf "%i\n" i ;
80 external build_pointers : Tree.t -> pointers = "caml_build_pointers"
81 external iter_pointers : pointers -> int = "caml_iter_pointers"
82 external free_pointers : pointers -> unit = "caml_free_pointers"
85 let main v query_string output =
87 let _ = Tag.init (Tree.tag_pool v) in
88 Printf.eprintf "Parsing query : ";
91 XPath.Parser.parse_string query_string
93 Ulexer.Loc.Exc_located ((x,y),e) -> Printf.eprintf "character %i-%i %s\n" x y (Printexc.to_string e);exit 1
95 let _ = Printf.eprintf "Number of nodes %i\n%!" (Tree.size v) in
96 (* let _ = test_text v in *)
97 (* let _ = Tree.stats v in
98 let _ = Printf.eprintf "Timing first_child/next_sibling %!" in
99 let _ = time (Tree.benchmark_fcns) v in
100 let _ = Printf.eprintf "Timing last_child/prev_sibling %!" in
101 let _ = time (Tree.benchmark_lcps) v in
102 let _ = Printf.eprintf "Timing jump to a %!" in
103 let _ = time (Tree.benchmark_jump v) (Tag.tag "a") in
104 let _ = Printf.eprintf "Timing pointer allocation %!" in
105 let pointers = time (build_pointers) v in
106 let _ = Printf.eprintf "Timing pointer iteration %!" in
107 let i = time (iter_pointers) pointers in
108 let _ = Printf.eprintf "Traversed %i pointers\n\nTiming pointer deallocation %!" i in
109 let _ = time (free_pointers) pointers in *)
110 (* let _ = Printf.eprintf "Timing //keyword :" in
111 let r = time (test_loop v) (Tag.tag "keyword") in
112 let _ = Printf.eprintf "Count is %i\n%!" r in
113 let _ = Printf.eprintf "Timing //keyword 2:" in
114 let r = time (test_loop2 v) (Tag.tag "keyword") in
115 let _ = Printf.eprintf "Count is %i\n%!" r in
116 let _ = Printf.eprintf "Timing //node() :" in
117 let _ = time (test_full) v in *)
118 XPath.Ast.print Format.err_formatter query;
119 Format.fprintf Format.err_formatter "\n%!";
120 Printf.eprintf "Compiling query : ";
121 let auto,ltags,contains = time (XPath.Compile.compile ~querystring:query_string) query in
122 let _ = Ata.dump Format.err_formatter auto in
123 let _ = Printf.eprintf "%!" in
126 None -> (max_int,`NOTHING)
128 let r = Tree.count v s
130 Printf.eprintf "%i documents in the TextCollection\n" (Tree.text_size v);
131 Printf.eprintf "Global count is %i, using " r;
132 if r < !Options.tc_threshold then begin
133 Printf.eprintf "TextCollection contains\nCalling global contains : ";
134 time (Tree.init_textfun op v) s;
137 Printf.eprintf "Naive contains\nCalling global contains : ";
138 time (Tree.init_naive_contains v) s
141 let test_list = jump_to in
144 if (!Options.backward) then begin
145 Printf.eprintf "Finding min occurences : ";
147 ( List.fold_left (fun ((min_occ,kind)as acc) (tag,_) ->
148 let numtags = Tree.subtree_tags v tag Tree.root in
149 if ((numtags < min_occ) && numtags >= 2)
150 then (numtags,`TAG(tag))
151 else acc) jump_to) ltags
153 else (max_int,`NOTHING)
155 let _ = if (snd test_list) != `NOTHING then
156 let occ,s1,s2 = match test_list with
157 | (x,`TAG (tag)) -> (x, "tag", (Tag.to_string tag))
158 | (x,`CONTAINS(s)) -> (x, "contains()", s)
161 Printf.eprintf "Will jump to %s %s occuring %i time\n%!" s1 s2 occ
163 Printf.eprintf "Execution time %s : "
164 (if !Options.count_only then "(counting only)" else if !Options.backward then "(bottomup)" else "");
166 let _ = Gc.full_major();Gc.compact() in
167 let _ = Printf.eprintf "%!" in
168 (* let _ = Gc.set (disabled_gc) in *)
169 if !Options.backward && ((snd test_list) != `NOTHING )then
170 if !Options.count_only then
171 let r = time_mem (bottom_up_count auto v )(snd test_list) in
172 let _ = Printf.eprintf "Number of nodes in the result set : %i\n%!" r
175 let r = time_mem (bottom_up auto v )(snd test_list) in
176 let _ = Printf.eprintf "Number of nodes in the result set : %i\n%!" (IdSet.length r)
182 Printf.eprintf "Serializing results : ";
184 (*let oc = open_out f in *)
185 let oc = Unix.openfile f [ Unix.O_WRONLY;Unix.O_TRUNC;Unix.O_CREAT] 0o644 in
186 (*output_string oc "<?xml version=\"1.0\" encoding=\"UTF-8\"?>\n";*)
188 Tree.print_xml_fast3 v t oc;
189 (*output_char oc '\n'; *)
195 if !Options.backward then Printf.eprintf "WARNING: couldn't find a jumping point, running top-down\n"
197 if !Options.count_only then
198 let r = time_mem ( top_down_count auto ) v in
199 let _ = Printf.eprintf "Number of nodes in the result set : %i\n%!" r
202 let module GR = Ata.Test(struct let doc = v end) in
203 let result = time_mem (GR.top_down auto) v in
204 let _ = Printf.eprintf "Counting results " in
205 let rcount = time (GR.Results.length) result in
206 Printf.eprintf "Number of nodes in the result set : %i\n" rcount;
207 Printf.eprintf "\n%!";
212 Printf.eprintf "Serializing results : ";
214 (*let oc = open_out f in *)
215 let oc = Unix.openfile f [ Unix.O_WRONLY;Unix.O_TRUNC;Unix.O_CREAT] 0o644 in
216 (*output_string oc "<?xml version=\"1.0\" encoding=\"UTF-8\"?>\n";*)
217 let t1 = ref (Unix.gettimeofday()) in
219 let old_count = ref 1 in
220 GR.Results.iter (fun t ->
223 if (!count mod 15) == 0
225 let t2 = Unix.gettimeofday() in
226 let _ = Printf.eprintf "Printing %i elements in %f ms\n"
227 (!count - !old_count) (1000. *.(t2 -. !t1))
229 ( old_count := !count; t1 := Unix.gettimeofday())
231 Tree.print_xml_fast3 v t oc;
232 (*output_char oc '\n'; *)
236 let _ = Gc.set enabled_gc in
237 Printf.eprintf "Total running time : %fms\n%!" (total_time())
240 Options.parse_cmdline();;
243 if (Filename.check_suffix !Options.input_file ".srx")
246 Printf.eprintf "Loading from file : ";
247 time (Tree.load ~sample:!Options.sample_factor ~load_text:(not !Options.count_only))
252 time (fun () -> let v = Tree.parse_xml_uri !Options.input_file;
253 in Printf.eprintf "Parsing document : %!";v
256 if !Options.save_file <> ""
258 Printf.eprintf "Writing file to disk : ";
259 time (Tree.save v) !Options.save_file;
263 main v !Options.query !Options.output_file;;