INCLUDE "utils.ml"
let () = init_timer();;
-let enabled_gc = Gc.get()
-let disabled_gc = { Gc.get() with
- Gc.max_overhead = 1000000;
- Gc.space_overhead = 100 }
+let default_gc = Gc.get()
+let tuned_gc = { Gc.get() with
+ Gc.minor_heap_size = 4*1024*1024;
+ Gc.major_heap_increment = 1024*1024;
+ Gc.max_overhead = 1000000;
+ }
let hash x = 131*x/(x-1+1)
let test_loop tree tag =
let f = Hashtbl.create 4096
in
let jump t _ = Tree.tagged_following_below tree tag t Tree.root in
- let g t ctx =
+ let g t ctx =
if t == Tree.nil then 0
else 1+ ((Hashtbl.find f (hash 101)) (jump t ctx) ctx)
in
Hashtbl.add f (hash 101) g;
(Hashtbl.find f (hash 101)) t' Tree.root
-let test_full tree =
+let test_full tree =
let root = Tree.root in
let fin = Tree.closing tree root in
let rec loop t = if t <= fin then
let tag = Tree.tag tree t in
(* let _ = Tag.to_string tag in *)
- if tag == Tag.pcdata then (ignore (Tree.get_text tree t));
+ if tag == Tag.pcdata then (ignore (Tree.get_text tree t));
let t = (Obj.magic ((Obj.magic t) + 1)) in
loop t
in
let f = Hashtbl.create 4096
in
let jump t _ = Tree.tagged_following_below tree tag t Tree.root in
- let rec g t ctx =
+ let rec g t ctx =
if t == Tree.nil then 0
else 1+ (match (Hashtbl.find f (hash 101)) with
`Foo ->g (jump t ctx) ctx
- )
+ )
in
Hashtbl.add f (hash 101) `Foo;
g t' Tree.root
-let test_text doc =
+let test_text doc =
let _ = Printf.eprintf "Contains(bree)" in
let _ = time (Tree.test_contains doc) "bree" in
let _ = Printf.eprintf "Contains(brain)" in
()
type pointers
-external build_pointers : Tree.t -> pointers = "caml_build_pointers"
+type order = PREORDER | INORDER | POSTORDER
+external build_pointers : Tree.t -> order -> pointers = "caml_build_pointers"
external iter_pointers : pointers -> int = "caml_iter_pointers"
external free_pointers : pointers -> unit = "caml_free_pointers"
let main v query_string output =
-
+
let _ = Tag.init (Tree.tag_pool v) in
- Printf.eprintf "Parsing query : ";
+ Printf.eprintf "Parsing query : ";
let query = try
time
XPath.Parser.parse_string query_string
in
let _ = Printf.eprintf "Number of nodes %i\n%!" (Tree.size v) in
(* let _ = test_text v in *)
-(* let _ = Tree.stats v in
- let _ = Printf.eprintf "Timing first_child/next_sibling %!" in
- let _ = time (Tree.benchmark_fcns) v in
- let _ = Printf.eprintf "Timing last_child/prev_sibling %!" in
+(* let _ = Tree.stats v in *)
+ let _ = Printf.eprintf "\nTiming first_child/next_sibling on sxsi %!" in
+ let c = time (Tree.benchmark_fcns) v in
+ let _ = Printf.eprintf "Traversed %i nodes\n" c in
+ let _ = Printf.eprintf "\nTiming first_element/next_element on sxsi %!" in
+ let c = time (Tree.benchmark_fene) v in
+ let _ = Printf.eprintf "Traversed %i nodes\n" c in
+ let _ = Printf.eprintf "\nTiming last_child/prev_sibling on sxsi %!" in
let _ = time (Tree.benchmark_lcps) v in
- let _ = Printf.eprintf "Timing jump to a %!" in
- let _ = time (Tree.benchmark_jump v) (Tag.tag "a") in
- let _ = Printf.eprintf "Timing pointer allocation %!" in
- let pointers = time (build_pointers) v in
- let _ = Printf.eprintf "Timing pointer iteration %!" in
+ let tag = "keyword" in
+ let _ = Printf.eprintf "\nTiming jump to <%s> on sxsi %!" tag in
+ let _ = time (Tree.benchmark_jump v) (Tag.tag tag) in
+ (* let _ = Printf.eprintf "\nTiming pointer allocation (preorder) %!" in
+ let pointers = time (build_pointers v) PREORDER in
+ let _ = Printf.eprintf "\nTiming pointer iteration %!" in
let i = time (iter_pointers) pointers in
- let _ = Printf.eprintf "Traversed %i pointers\n\nTiming pointer deallocation %!" i in
+ let _ = Printf.eprintf "Traversed %i pointers\nTiming pointer deallocation %!" i in
+ let _ = time (free_pointers) pointers in
+
+
+ let _ = Printf.eprintf "\nTiming pointer allocation (inorder) %!" in
+ let pointers = time (build_pointers v) INORDER in
+ let _ = Printf.eprintf "\nTiming pointer iteration %!" in
+ let i = time (iter_pointers) pointers in
+ let _ = Printf.eprintf "Traversed %i pointers\nTiming pointer deallocation %!" i in
+ let _ = time (free_pointers) pointers in
+
+ let _ = Printf.eprintf "\nTiming pointer allocation (postorder) %!" in
+ let pointers = time (build_pointers v) POSTORDER in
+ let _ = Printf.eprintf "\nTiming pointer iteration %!" in
+ let i = time (iter_pointers) pointers in
+ let _ = Printf.eprintf "Traversed %i pointers\nTiming pointer deallocation %!" i in
let _ = time (free_pointers) pointers in *)
+
+ let _ = Printf.eprintf "\nTiming iterative_traversal on sxsi %!" in
+ let c = time (Tree.benchmark_iter) v in
+ let _ = Printf.eprintf "Traversed %i nodes\n" c in
+
+
+
+
+
+
(* let _ = Printf.eprintf "Timing //keyword :" in
let r = time (test_loop v) (Tag.tag "keyword") in
- let _ = Printf.eprintf "Count is %i\n%!" r in
+ let _ = Printf.eprintf "Count is %i\n%!" r in
let _ = Printf.eprintf "Timing //keyword 2:" in
let r = time (test_loop2 v) (Tag.tag "keyword") in
- let _ = Printf.eprintf "Count is %i\n%!" r in
+ let _ = Printf.eprintf "Count is %i\n%!" r in
let _ = Printf.eprintf "Timing //node() :" in
let _ = time (test_full) v in *)
XPath.Ast.print Format.err_formatter query;
Format.fprintf Format.err_formatter "\n%!";
Printf.eprintf "Compiling query : ";
- let auto,ltags,contains = time (XPath.Compile.compile ~querystring:query_string) query in
+ let auto,ltags,contains = time (XPath.Compile.compile ~querystring:query_string) query in
let _ = Ata.dump Format.err_formatter auto in
let _ = Printf.eprintf "%!" in
- let jump_to =
+ let jump_to =
match contains with
None -> (max_int,`NOTHING)
- | Some ((op,s)) ->
- let r = Tree.count v s
+ | Some ((op,s)) ->
+ let r = Tree.count v s
in
Printf.eprintf "%i documents in the TextCollection\n" (Tree.text_size v);
Printf.eprintf "Global count is %i, using " r;
- if r < !Options.tc_threshold then begin
+ if r < !Options.tc_threshold then begin
Printf.eprintf "TextCollection contains\nCalling global contains : ";
time (Tree.init_textfun op v) s;
end
in
let test_list = jump_to in
(*
- let test_list =
+ let test_list =
if (!Options.backward) then begin
Printf.eprintf "Finding min occurences : ";
- time
+ time
( List.fold_left (fun ((min_occ,kind)as acc) (tag,_) ->
let numtags = Tree.subtree_tags v tag Tree.root in
if ((numtags < min_occ) && numtags >= 2)
begin
let _ = Gc.full_major();Gc.compact() in
let _ = Printf.eprintf "%!" in
-(* let _ = Gc.set (disabled_gc) in *)
- if !Options.backward && ((snd test_list) != `NOTHING )then
+ let _ = Gc.set (tuned_gc) in
+ if !Options.backward && ((snd test_list) != `NOTHING )then
if !Options.count_only then
let r = time_mem (bottom_up_count auto v )(snd test_list) in
- let _ = Printf.eprintf "Number of nodes in the result set : %i\n%!" r
+ let _ = Printf.eprintf "Number of nodes in the result set : %i\n%!" r
in ()
else begin
let r = time_mem (bottom_up auto v )(snd test_list) in
match output with
| None -> ()
- | Some f ->
+ | Some f ->
Printf.eprintf "Serializing results : ";
time( fun () ->
(*let oc = open_out f in *)
let oc = Unix.openfile f [ Unix.O_WRONLY;Unix.O_TRUNC;Unix.O_CREAT] 0o644 in
(*output_string oc "<?xml version=\"1.0\" encoding=\"UTF-8\"?>\n";*)
- IdSet.iter (fun t ->
+ IdSet.iter (fun t ->
Tree.print_xml_fast3 v t oc;
- (*output_char oc '\n'; *)
+ (*output_char oc '\n'; *)
) r) ();
end
-
+
else
- let _ =
- if !Options.backward then Printf.eprintf "WARNING: couldn't find a jumping point, running top-down\n"
+ let _ =
+ if !Options.backward then Printf.eprintf "WARNING: couldn't find a jumping point, running top-down\n"
in
if !Options.count_only then
- let r = time_mem ( top_down_count auto ) v in
+ let r = time ~count:5 ( top_down_count1 auto ) v in
let _ = Printf.eprintf "Number of nodes in the result set : %i\n%!" r
in ()
- else
- let module GR = Ata.Test(struct let doc = v end) in
- let result = time_mem (GR.top_down auto) v in
+ else
+ let module GR = Ata(*.Test(struct let doc = v end) *) in
+ let result = time ~count:5 (GR.top_down1 auto) v in
let _ = Printf.eprintf "Counting results " in
- let rcount = time (GR.Results.length) result in
+ let rcount = time (IdSet.length) result in
Printf.eprintf "Number of nodes in the result set : %i\n" rcount;
Printf.eprintf "\n%!";
begin
match output with
| None -> ()
- | Some f ->
+ | Some f ->
Printf.eprintf "Serializing results : ";
- time( fun () ->
- (*let oc = open_out f in *)
- let oc = Unix.openfile f [ Unix.O_WRONLY;Unix.O_TRUNC;Unix.O_CREAT] 0o644 in
- (*output_string oc "<?xml version=\"1.0\" encoding=\"UTF-8\"?>\n";*)
- let t1 = ref (Unix.gettimeofday()) in
- let count = ref 1 in
- let old_count = ref 1 in
- GR.Results.iter (fun t ->
- incr count;
- begin
- if (!count mod 15) == 0
- then
- let t2 = Unix.gettimeofday() in
- let _ = Printf.eprintf "Printing %i elements in %f ms\n"
- (!count - !old_count) (1000. *.(t2 -. !t1))
- in
- ( old_count := !count; t1 := Unix.gettimeofday())
- end;
- Tree.print_xml_fast3 v t oc;
- (*output_char oc '\n'; *)
- ) result) ();
- end;
+ let oc =
+ Unix.openfile f [ Unix.O_WRONLY;Unix.O_TRUNC;Unix.O_CREAT] 0o644
+ in
+ time(
+ IdSet.iter (
+ fun t ->
+
+ Tree.print_xml_fast3 v t oc;
+
+ )) result ;
+ end;
end;
- let _ = Gc.set enabled_gc in
- Printf.eprintf "Total running time : %fms\n%!" (total_time())
+ Printf.eprintf "Total running time : %fms\n%!" (total_time())
;;
Options.parse_cmdline();;
let v =
if (Filename.check_suffix !Options.input_file ".srx")
- then
+ then
begin
Printf.eprintf "Loading from file : ";
time (Tree.load ~sample:!Options.sample_factor ~load_text:(not !Options.count_only))
!Options.input_file;
end
- else
- let v =
+ else
+ let v =
time (fun () -> let v = Tree.parse_xml_uri !Options.input_file;
in Printf.eprintf "Parsing document : %!";v
- ) ()
+ ) ()
in
if !Options.save_file <> ""
then begin