X-Git-Url: http://git.nguyen.vg/gitweb/?a=blobdiff_plain;f=main.ml;h=dc2cb3e7ac790b16506794c8ca29e6653e65644c;hb=329088598ab63bc2d67ff0dfc4f54e90f5d4f283;hp=ba9332859e8c4dd7efef221fd8ae34756fdcdeca;hpb=c10ce35cd399aff15a49f3b24a31b38cb2191da0;p=SXSI%2Fxpathcomp.git diff --git a/main.ml b/main.ml index ba93328..dc2cb3e 100644 --- a/main.ml +++ b/main.ml @@ -4,74 +4,186 @@ (* Copyright NICTA 2008 *) (* Distributed under the terms of the LGPL (see LICENCE) *) (******************************************************************************) -INCLUDE "debug.ml" -open Automaton +open Ata +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 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_desc tree tag Tree.root in + let f = Hashtbl.create 4096 + in + let jump t _ = Tree.tagged_foll_ctx 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 main filename query output = - (* Just a trick to allow the C++ code to print debugging stuff first *) - let v = time (fun () -> let v = Tree.Binary.parse_xml_uri filename; - in Printf.eprintf "Parsing document : %!";v - ) () - in - let _ = Tag.init (Tree.Binary.tag_pool v) in - MM(v,__LOCATION__); + +let test_loop2 tree tag = + let t' = Tree.tagged_desc tree tag Tree.root in + let f = Hashtbl.create 4096 + in + let jump t _ = Tree.tagged_foll_ctx 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 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; - Format.eprintf "\n%!"; - (* Format.eprintf "Internal rep of the tree is :\n%!"; - Tree.Binary.dump v; *) - Printf.eprintf "Execution time : "; - time (fun v -> ignore (TopDown.accept auto v)) v; - Printf.eprintf "Number of nodes in the result set : %i\n" (BST.cardinal auto.result); + 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 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_contains 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 - match output with - | None -> () - | Some f -> - - Printf.eprintf "Serializing results : "; - time( fun () -> - let oc = open_out f in - output_string oc "\n"; - BST.iter (fun t -> Tree.Binary.print_xml_fast oc t; - output_char oc '\n') auto.result) (); - end; - Printf.eprintf "Total time : %fms\n Coherence : %i\n%!" (total_time()) -;; + 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 argc = Array.length Sys.argv;; -if (argc < 3 || argc >4) -then - (prerr_endline ("usage : " ^ Sys.argv.(0) ^ " \'query\'[ ]"); - exit 1) + let r = time (bottom_up_count auto v )(snd test_list) in + let _ = Printf.eprintf "Number of nodes in the result set : %i\n%!" r + in () + 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 ( top_down_count 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 (GR.top_down 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 "\n";*) + GR.Results.iter (fun t -> + Tree.print_xml_fast3 v t oc; + (*output_char oc '\n'; *) + ) result) (); + end; + end; + let _ = Gc.set enabled_gc in + Printf.eprintf "Total running time : %fms\n%!" (total_time()) ;; +Options.parse_cmdline();; -main Sys.argv.(1) Sys.argv.(2) (if argc == 4 then Some Sys.argv.(3) else None) ;; - -Printf.eprintf "\n=================================================\nDEBUGGING\n%!"; -Tree.DEBUGTREE.print_stats Format.err_formatter;; - +let v = + if (Filename.check_suffix !Options.input_file ".srx") + then + begin + Printf.eprintf "Loading from file : "; + time (Tree.load ~sample:!Options.sample_factor ) + !Options.input_file; + end + 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 + Printf.eprintf "Writing file to disk : "; + time (Tree.save v) !Options.save_file; + end; + v +in + main v !Options.query !Options.output_file;;