(* Copyright NICTA 2008 *)
(* Distributed under the terms of the LGPL (see LICENCE) *)
(******************************************************************************)
-INCLUDE "debug.ml"
open Ata
+INCLUDE "utils.ml"
+let () = init_timer();;
+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 l = ref [] ;;
-let time f x =
- let t1 = Unix.gettimeofday () in
- let r = f x in
- let t2 = Unix.gettimeofday () in
- let t = (1000. *.(t2 -. t1)) in
- l:= t::!l;
- Printf.eprintf " %fms\n%!" t ;
- r
-;;
-let total_time () = List.fold_left (+.) 0. !l;;
+let test_loop tree tag =
+ let t' = Tree.tagged_descendant tree tag Tree.root in
+ let f = Hashtbl.create 4096
+ in
+ let jump t _ = Tree.tagged_following_below tree tag t Tree.root in
+ 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 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));
+ let t = (Obj.magic ((Obj.magic t) + 1)) in
+ loop t
+ in
+ loop root
+
+
+let test_loop2 tree tag =
+ let t' = Tree.tagged_descendant tree tag Tree.root 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 =
+ 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 _ = Printf.eprintf "Contains(bree)" in
+ let _ = time (Tree.test_contains doc) "bree" in
+ let _ = Printf.eprintf "Contains(brain)" in
+ let _ = time (Tree.test_contains doc) "brain" in
+ let _ = Printf.eprintf "Contains(brain)" in
+ let i = time (Tree.test_contains doc) "brain" in
+ let _ = Printf.eprintf "%i\nContains(Australia)" i in
+ let i = time (Tree.test_contains doc) "AUSTRALIA" in
+ let _ = Printf.eprintf "%i\n Contains(1930)" i in
+ let i = time (Tree.test_contains doc) "1930" in
+ let _ = Printf.eprintf "%i\n startswith(bar)" i in
+ let i = time (Tree.test_prefix doc) "bar" in
+ let _ = Printf.eprintf "%i\n endswith(LAND)" i in
+ let i = time (Tree.test_suffix doc) "LAND" in
+ let _ = Printf.eprintf "%i\n =(2001)" i in
+ let i = time (Tree.test_equals doc) "2001" in
+ let _ = Printf.eprintf "%i\n =(Nguyen)" i in
+ let i = time (Tree.test_equals doc) "Nguyen" in
+ Printf.eprintf "%i\n" i ;
+ ()
+type pointers
+external build_pointers : Tree.t -> pointers = "caml_build_pointers"
+external iter_pointers : pointers -> int = "caml_iter_pointers"
+external free_pointers : pointers -> unit = "caml_free_pointers"
-let main v query output =
- let _ = Tag.init (Tree.Binary.tag_pool v) in
+
+let main v query_string output =
+
+ let _ = Tag.init (Tree.tag_pool v) in
Printf.eprintf "Parsing query : ";
let query = try
time
- XPath.Parser.parse_string query
+ XPath.Parser.parse_string query_string
with
Ulexer.Loc.Exc_located ((x,y),e) -> Printf.eprintf "character %i-%i %s\n" x y (Printexc.to_string e);exit 1
- in
- Printf.eprintf "Compiling query : ";
- let auto = time XPath.Compile.compile query in
- XPath.Ast.print Format.err_formatter query;
- Printf.eprintf "Execution time : ";
- let result = time (BottomUpNew.run auto) v in
- Printf.eprintf "Number of nodes in the result set : %i\n" (List.length result);
- begin
- match output with
- | None -> ()
- | Some f ->
-
- Printf.eprintf "Serializing results : ";
- time( fun () ->
- let oc = open_out f in
- output_string oc "<?xml version=\"1.0\" encoding=\"UTF-8\"?>\n";
- List.iter (fun t -> Tree.Binary.print_xml_fast oc t;
- output_char oc '\n') result) ();
- end;
- Printf.eprintf "Total time : %fms\n Coherence : %i\n%!" (total_time())
+ 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 ~count:1 (Tree.benchmark_fcns) v in *)
+ (* let _ = Printf.eprintf "Timing last_child/prev_sibling %!" 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 i = time (iter_pointers) pointers in
+ let _ = Printf.eprintf "Traversed %i pointers\n\nTiming pointer deallocation %!" i in
+ let _ = time (free_pointers) pointers 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 "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 "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 _ = Ata.dump Format.err_formatter auto in
+ let _ = Printf.eprintf "%!" in
+ let jump_to =
+ match contains with
+ None -> (max_int,`NOTHING)
+ | 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
+ Printf.eprintf "TextCollection contains\nCalling global contains : ";
+ time (Tree.init_textfun op v) s;
+ end
+ else begin
+ Printf.eprintf "Naive contains\nCalling global contains : ";
+ time (Tree.init_naive_contains v) s
+ end;(r,`CONTAINS(s))
+ in
+ let test_list = jump_to in
+ (*
+ let test_list =
+ if (!Options.backward) then begin
+ Printf.eprintf "Finding min occurences : ";
+ 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)
+ then (numtags,`TAG(tag))
+ else acc) jump_to) ltags
+ end
+ else (max_int,`NOTHING)
+ in*)
+ let _ = if (snd test_list) != `NOTHING then
+ let occ,s1,s2 = match test_list with
+ | (x,`TAG (tag)) -> (x, "tag", (Tag.to_string tag))
+ | (x,`CONTAINS(s)) -> (x, "contains()", s)
+ | _ -> assert false
+ in
+ Printf.eprintf "Will jump to %s %s occuring %i time\n%!" s1 s2 occ
+ in
+ Printf.eprintf "Execution time %s : "
+ (if !Options.count_only then "(counting only)" else if !Options.backward then "(bottomup)" else "");
+ begin
+ let _ = Gc.full_major();Gc.compact() in
+ let _ = Printf.eprintf "%!" in
+ 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
+ in ()
+ else begin
+ let r = time_mem (bottom_up auto v )(snd test_list) in
+ let _ = Printf.eprintf "Number of nodes in the result set : %i\n%!" (IdSet.length r)
+ in
+ match output with
+
+ | None -> ()
+ | 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 ->
+ Tree.print_xml_fast3 v t oc;
+ (*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"
+ in
+ if !Options.count_only then
+ 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 ~count:5 (GR.top_down1 auto) v in
+ let _ = Printf.eprintf "Counting results " in
+ let rcount = time (GR.Results.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 ->
+ 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;
+ end;
+ Printf.eprintf "Total running time : %fms\n%!" (total_time())
;;
-
Options.parse_cmdline();;
-let v =
+let v =
if (Filename.check_suffix !Options.input_file ".srx")
then
begin
Printf.eprintf "Loading from file : ";
- time (Tree.Binary.load ~sample:!Options.sample_factor )
- (Filename.chop_suffix !Options.input_file ".srx");
- end
+ time (Tree.load ~sample:!Options.sample_factor ~load_text:(not !Options.count_only))
+ !Options.input_file;
+ end
else
let v =
- time (fun () -> let v = Tree.Binary.parse_xml_uri !Options.input_file;
+ time (fun () -> let v = Tree.parse_xml_uri !Options.input_file;
in Printf.eprintf "Parsing document : %!";v
) ()
in
if !Options.save_file <> ""
then begin
Printf.eprintf "Writing file to disk : ";
- time (Tree.Binary.save v) !Options.save_file;
+ time (Tree.save v) !Options.save_file;
end;
v
in
main v !Options.query !Options.output_file;;
-IFDEF DEBUG
-THEN
-Printf.eprintf "\n=================================================\nDEBUGGING\n%!";
-Format.eprintf "\nAutomaton is:\n%!";
-Ata.dump Format.err_formatter auto;
-Tree.DEBUGTREE.print_stats Format.err_formatter;;
-Gc.full_major()
-ENDIF
+
+