(******************************************************************************) (* SXSI : XPath evaluator *) (* Kim Nguyen (Kim.Nguyen@nicta.com.au) *) (* Copyright NICTA 2008 *) (* Distributed under the terms of the LGPL (see LICENCE) *) (******************************************************************************) open Ata 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 enabled_gc = Gc.get() let disabled_gc = { Gc.get() with Gc.max_overhead = 1000000; Gc.space_overhead = 100 } 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_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 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 = 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 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 _ = Gc.set (disabled_gc) in *) if !Options.backward && ((snd test_list) != `NOTHING )then 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 result = time (top_down auto) v in let rcount = 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 -> Printf.eprintf "Serializing results : "; time( fun () -> let oc = open_out f in output_string oc "\n"; IdSet.iter (fun t -> Tree.print_xml_fast oc t; output_char oc '\n'; output_string oc "----------\n"; ) result) (); end; end; let _ = Gc.set enabled_gc in Printf.eprintf "Total running time : %fms\n%!" (total_time()) ;; Options.parse_cmdline();; let v = if (Filename.check_suffix !Options.input_file ".srx") then begin Printf.eprintf "Loading from file : "; time (Tree.load ~sample:!Options.sample_factor ) (Filename.chop_suffix !Options.input_file ".srx"); 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;;