(* Distributed under the terms of the LGPL (see LICENCE) *)
(******************************************************************************)
INCLUDE "debug.ml"
+INCLUDE "log.ml"
INCLUDE "utils.ml"
-
external init_lib : unit -> unit = "sxsi_cpp_init"
exception CPlusPlusError of string
type tree
+external register_tag : tree -> string -> Tag.t = "caml_xml_tree_register_tag"
+
+external tag_name : tree -> Tag.t -> string = "caml_xml_tree_get_tag_name"
+
+let tag t = (); fun s ->
+ match s with
+ | "<$>" -> Tag.pcdata
+ | "<@>" -> Tag.attribute
+ | "" -> Tag.document_node
+ | "<@$>" -> Tag.attribute_data
+ | _ -> register_tag t s
+
+let to_string d = ();
+ fun t ->
+ if t == Tag.pcdata then "<$>"
+ else if t == Tag.attribute_data then "<@$>"
+ else if t == Tag.attribute then "<@>"
+ else if t == Tag.nullt then "<!NIL!>"
+ else tag_name d t
+
+let translate x = x
+
+let mk_tag_ops t = {
+ Tag.tag = tag t;
+ Tag.to_string = to_string t;
+ Tag.translate = translate
+}
module TreeBuilder =
struct
type t
external create : unit -> t = "caml_xml_tree_builder_create"
- external open_document : t -> bool -> int -> bool -> int -> unit = "caml_xml_tree_builder_open_document"
+ external open_document : t -> int -> bool -> int -> unit = "caml_xml_tree_builder_open_document"
external close_document : t -> tree = "caml_xml_tree_builder_close_document"
- external open_tag : t -> string -> unit = "caml_xml_tree_builder_new_open_tag"
- external close_tag : t -> string -> unit = "caml_xml_tree_builder_new_closing_tag"
- external text : t -> string -> unit = "caml_xml_tree_builder_new_text"
+ external open_tag : t -> string -> unit = "caml_xml_tree_builder_open_tag"
+ external close_tag : t -> string -> unit = "caml_xml_tree_builder_close_tag"
+ external text : t -> string -> unit = "caml_xml_tree_builder_text"
+ external trim : string -> string = "caml_trim"
+
+ let is_whitespace s =
+ let rec loop len i =
+ if i < len then
+ let c = s.[i] in
+ (c == '\n' || c == '\t' || c == ' ') && loop len (i+1)
+ else
+ true
+ in
+ loop (String.length s) 0
+
+
+ let display_count =
+ let event_counter = ref 0 in
+ (fun parser_ ->
+ incr event_counter;
+ if !event_counter land 0xffffff == 0 then
+ Logger.print Format.err_formatter "Current position: %i@\n@?" (Expat.get_current_byte_index parser_))
let do_text b t =
if Buffer.length t > 0 then begin
- open_tag b "<$>";
- text b (Buffer.contents t);
- close_tag b "<$>";
+ let s = Buffer.contents t in
+ if (!Config.index_empty_texts) || not (is_whitespace s) then
+ begin
+ open_tag b "<$>";
+ text b s;
+ close_tag b "<$>";
+ end;
Buffer.clear t
end
let atname = "<@>" ^ name in
open_tag b atname;
open_tag b "<@$>";
- text b value;
- close_tag b "<@$>";
- close_tag b atname
+ text b value;
+ close_tag b "<@$>";
+ close_tag b atname
- let start_element_handler b t tag attr_list =
+ let start_element_handler parser_ b t tag attr_list =
do_text b t;
open_tag b tag;
match attr_list with
- [] -> ()
- | l ->
- open_tag b "<@>";
- List.iter (fun (name, value) -> output_attr b name value) l;
- close_tag b "<@>"
+ [] -> ()
+ | l ->
+ open_tag b "<@>";
+ List.iter (fun (name, value) -> output_attr b name value) l;
+ close_tag b "<@>"
- let end_element_handler b t tag =
+ let end_element_handler parser_ b t tag =
do_text b t;
close_tag b tag
- let character_data_handler _b t text =
+ let character_data_handler parser_ _ t text =
Buffer.add_string t text
let create_parser () =
let finalize () =
do_text build buf;
close_tag build "";
- close_document build
+ LOG ( __ "parsing" 2 "%s\n" "Finished parsing");
+ LOG ( __ "indexing" 2 "%s\n" "Starting index construction");
+ let r = close_document build
+ in
+ LOG ( __ "indexing" 2 "%s\n" "Finished index construction");
+ r
in
- Expat.set_start_element_handler parser_ (start_element_handler build buf);
- Expat.set_end_element_handler parser_ (end_element_handler build buf);
- Expat.set_character_data_handler parser_ (character_data_handler build buf);
- open_document build !Options.index_empty_texts !Options.sample_factor
- !Options.disable_text_collection !Options.text_index_type;
+ Expat.set_start_element_handler parser_ (start_element_handler parser_ build buf);
+ Expat.set_end_element_handler parser_ (end_element_handler parser_ build buf);
+ Expat.set_character_data_handler parser_ (character_data_handler parser_ build buf);
+ LOG ( __ "parsing" 2 "%s\n" "Started parsing");
+ open_document build !Config.sample_factor !Config.disable_text_collection !Config.text_index_type;
open_tag build "";
parser_, finalize
let in_chan = open_in file in
let buffer = String.create 4096 in
let parser_, finalizer = create_parser () in
- let () =
+ let parse () =
try
while true do
let read = input in_chan buffer 0 4096 in
if read == 0 then raise End_of_file else
Expat.parse_sub parser_ buffer 0 read;
- done
+ done
with
- | End_of_file -> close_in in_chan
- | e -> raise e
+ | End_of_file -> close_in in_chan
+ | e -> raise e
in
- finalizer ()
-
-
+ Utils.time ~msg:"Parsing XML file" parse ();
+ Utils.time ~msg:"Creating tree and text-collection index" finalizer ()
end
type bit_vector = string
external bool_of_int : int -> bool = "%identity"
+external int_of_bool : bool -> int = "%identity"
let bit_vector_unsafe_get v i =
bool_of_int
(((Char.code (String.unsafe_get v (i lsr 3))) lsr (i land 7)) land 1)
+let chr (c:int) : char = Obj.magic (c land 0xff)
+let bit_vector_unsafe_set v i b =
+ let j = i lsr 3 in
+ let c = Char.code v.[j] in
+ let bit = int_of_bool b in
+ let mask = bit lsl (i land 7) in
+ if b then v.[j] <- chr (c lor mask) else v.[j] <- (chr (c land (lnot mask)))
+
+let bit_vector_create n =
+ let len = if n <= 0 then 0 else (n - 1) / 8 + 1 in
+ String.make len '\000'
+
+type tag_list
+
+external tag_list_alloc : int -> tag_list = "caml_tag_list_alloc"
+external tag_list_set : tag_list -> int -> Tag.t -> unit = "caml_tag_list_set" "noalloc"
type t = {
doc : tree;
elements: Ptset.Int.t;
attributes: Ptset.Int.t;
- attribute_array : Tag.t array;
+ attribute_array : tag_list;
children : Ptset.Int.t array;
siblings : Ptset.Int.t array;
descendants: Ptset.Int.t array;
}
-
+let tag_operations t = mk_tag_ops t.doc
(*
-external parse_xml_uri : string -> int -> bool -> bool -> int -> tree = "caml_call_shredder_uri"
-external parse_xml_string : string -> int -> bool -> bool -> int -> tree = "caml_call_shredder_string"
+ external parse_xml_uri : string -> int -> bool -> bool -> int -> tree = "caml_call_shredder_uri"
+ external parse_xml_string : string -> int -> bool -> bool -> int -> tree = "caml_call_shredder_string"
*)
external tree_print_xml_fast3 : tree -> [`Tree ] Node.t -> Unix.file_descr -> unit = "caml_xml_tree_print"
let print_xml t n fd =
let nil : [`Tree ] Node.t = Node.nil
let root : [`Tree ] Node.t = Node.null
-type unordered_set
-
-external unordered_set_alloc : int -> unordered_set = "caml_unordered_set_alloc"
-external unordered_set_length : unordered_set -> int = "caml_unordered_set_length"
-external unordered_set_insert : unordered_set -> int -> unit = "caml_unordered_set_set" "noalloc"
module HPtset = Hashtbl.Make(Ptset.Int)
let vector_htbl = HPtset.create MED_H_SIZE
+let reinit () = HPtset.clear vector_htbl
-let unordered_set_of_set s =
+let tag_list_of_set s =
try
HPtset.find vector_htbl s
with
- Not_found ->
- let v = unordered_set_alloc (Ptset.Int.cardinal s) in
- let _ = Ptset.Int.iter (fun e -> unordered_set_insert v e) s in
- HPtset.add vector_htbl s v; v
-
-let ptset_to_vector = unordered_set_of_set
+ Not_found ->
+ let v = tag_list_alloc (Ptset.Int.cardinal s + 1) in
+ let i = ref 0 in
+ let () = Ptset.Int.iter (fun e -> tag_list_set v !i e; incr i) s in
+ let () = tag_list_set v !i Tag.nullt in
+ HPtset.add vector_htbl s v; v
(** tree interface *)
external tree_tagged_child : tree -> [`Tree] Node.t -> Tag.t -> [`Tree] Node.t = "caml_xml_tree_tagged_child" "noalloc"
let tagged_child t n tag = tree_tagged_child t.doc n tag
-external tree_select_child : tree -> [`Tree ] Node.t -> unordered_set -> [`Tree] Node.t = "caml_xml_tree_select_child" "noalloc"
+external tree_select_child : tree -> [`Tree ] Node.t -> tag_list -> [`Tree] Node.t = "caml_xml_tree_select_child" "noalloc"
let select_child t n tag_set = tree_select_child t.doc n tag_set
external tree_last_child : tree -> [`Tree] Node.t -> [`Tree] Node.t = "caml_xml_tree_last_child" "noalloc"
let last_child t n = tree_last_child t.doc n
-
external tree_next_sibling : tree -> [`Tree] Node.t -> [`Tree] Node.t = "caml_xml_tree_next_sibling" "noalloc"
let next_sibling t n = tree_next_sibling t.doc n
external tree_next_element : tree -> [`Tree] Node.t -> [`Tree] Node.t = "caml_xml_tree_next_element" "noalloc"
let next_element t n = tree_next_element t.doc n
-external tree_next_node_before : tree -> [`Tree] Node.t -> [`Tree] Node.t -> [`Tree] Node.t = "caml_xml_tree_next_node_before" "noalloc"
-let next_node_before t n ctx = tree_next_node_before t.doc n ctx
-external tree_tagged_following_sibling : tree -> [`Tree] Node.t -> Tag.t -> [`Tree] Node.t = "caml_xml_tree_tagged_following_sibling" "noalloc"
-let tagged_following_sibling t n tag = tree_tagged_following_sibling t.doc n tag
+external tree_tagged_sibling : tree -> [`Tree] Node.t -> Tag.t -> [`Tree] Node.t = "caml_xml_tree_tagged_sibling" "noalloc"
+let tagged_sibling t n tag = tree_tagged_sibling t.doc n tag
-external tree_select_following_sibling : tree -> [`Tree ] Node.t -> unordered_set -> [`Tree] Node.t = "caml_xml_tree_select_following_sibling" "noalloc"
-let select_following_sibling t n tag_set = tree_select_following_sibling t.doc n tag_set
+
+external tree_select_sibling : tree -> [`Tree ] Node.t -> tag_list -> [`Tree] Node.t = "caml_xml_tree_select_sibling" "noalloc"
+let select_sibling t n tag_set = tree_select_sibling t.doc n tag_set
external tree_prev_sibling : tree -> [`Tree] Node.t -> [`Tree] Node.t = "caml_xml_tree_prev_sibling" "noalloc"
let prev_sibling t n = tree_prev_sibling t.doc n
external tree_tagged_next : tree -> [`Tree ] Node.t -> Tag.t -> [`Tree ] Node.t = "caml_xml_tree_tagged_next" "noalloc"
let tagged_next t n tag = tree_tagged_next t.doc n tag
-external tree_select_descendant : tree -> [`Tree ] Node.t -> unordered_set -> [`Tree] Node.t = "caml_xml_tree_select_descendant" "noalloc"
+external tree_select_descendant : tree -> [`Tree ] Node.t -> tag_list -> [`Tree] Node.t = "caml_xml_tree_select_descendant" "noalloc"
let select_descendant t n tag_set = tree_select_descendant t.doc n tag_set
external tree_tagged_following_before : tree -> [`Tree ] Node.t -> Tag.t -> [`Tree ] Node.t -> [`Tree ] Node.t = "caml_xml_tree_tagged_following_before" "noalloc"
let tagged_following_before t n tag ctx = tree_tagged_following_before t.doc n tag ctx
-external tree_select_following_before : tree -> [`Tree ] Node.t -> unordered_set -> [`Tree] Node.t -> [`Tree] Node.t = "caml_xml_tree_select_following_before" "noalloc"
+external tree_select_following_before : tree -> [`Tree ] Node.t -> tag_list -> [`Tree] Node.t -> [`Tree] Node.t = "caml_xml_tree_select_following_before" "noalloc"
let select_following_before t n tag_set ctx = tree_select_following_before t.doc n tag_set ctx
external tree_parent : tree -> [`Tree] Node.t -> [`Tree] Node.t = "caml_xml_tree_parent" "noalloc"
let parent t n = tree_parent t.doc n
-external tree_binary_parent : tree -> [`Tree] Node.t -> [`Tree] Node.t = "caml_xml_tree_binary_parent"
- "noalloc"
-let binary_parent t n = tree_binary_parent t.doc n
-
-
external tree_tag : tree -> [`Tree] Node.t -> Tag.t = "caml_xml_tree_tag" "noalloc"
let tag t n = tree_tag t.doc n
open Format
let dump_tag_table t =
- eprintf "Child tags:\n%!";
- Array.iteri
- (fun tag set -> eprintf "%s: %a\n%!"
- (Tag.to_string tag) TagSet.print (TagSet.inj_positive set))
- t.children;
- eprintf "-----------------------------\n%!";
- eprintf "Descendant tags:\n%!";
- Array.iteri
- (fun tag set -> eprintf "%s: %a\n%!"
- (Tag.to_string tag) TagSet.print (TagSet.inj_positive set))
- t.descendants;
- eprintf "-----------------------------\n%!";
- eprintf "Sibling tags:\n%!";
- Array.iteri
- (fun tag set -> eprintf "%s: %a\n%!"
- (Tag.to_string tag) TagSet.print (TagSet.inj_positive set))
- t.siblings;
- eprintf "-----------------------------\n%!";
- eprintf "Following tags:\n%!";
- Array.iteri
- (fun tag set -> eprintf "%s: %a\n%!"
- (Tag.to_string tag) TagSet.print (TagSet.inj_positive set))
- t.followings;
- eprintf "-----------------------------\n%!"
-
+ let tag = ref 0 in
+ let printer ppf set =
+ Logger.print ppf "%s: %a"
+ (Tag.to_string !tag) TagSet.print (TagSet.inj_positive set);
+ incr tag
+ in
+ let set_printer msg set =
+ tag := 0;
+ Logger.print err_formatter "%s :@\n" msg;
+ Pretty.pp_print_array ~sep:pp_force_newline printer err_formatter set;
+ Logger.print err_formatter "-----------------------------@\n";
+ in
+ set_printer "Child tags" t.children;
+ set_printer "Descendant tags" t.descendants;
+ set_printer "Sibling tags" t.siblings;
+ set_printer "Following tags" t.followings
external tree_subtree_tags : tree -> [`Tree] Node.t -> Tag.t -> int = "caml_xml_tree_subtree_tags" "noalloc"
let subtree_tags t n tag = tree_subtree_tags t.doc n tag
external tree_subtree_size : tree -> [`Tree] Node.t -> int = "caml_xml_tree_subtree_size" "noalloc"
let subtree_size t n = tree_subtree_size t.doc n
+let rec iter_array_tag i a len tree node acc =
+ if i == len then acc
+ else
+ iter_array_tag (i+1) a len tree node
+ (acc - (tree_subtree_tags tree node a.(i)))
+
+external tree_subtree_elements : tree -> [`Tree] Node.t -> int = "caml_xml_tree_subtree_elements" "noalloc"
+
+let subtree_elements t node =
+ tree_subtree_elements t.doc node
+(*
let subtree_elements t node =
let size = tree_subtree_size t.doc node - 1 in
- if size == 0 then 0
- else let size = size - (tree_subtree_tags t.doc node Tag.pcdata) in
- if size < 2 then size else
- let acc = ref size in
- for i = 0 to Array.length t.attribute_array - 1 do
- acc := !acc - tree_subtree_tags t.doc node t.attribute_array.(i)
- done;
- !acc
+ if size <= 0 then 0
+ else let size = size - (tree_subtree_tags t.doc node Tag.pcdata) in
+ if size < 3 then size else
+ let a = t.attribute_array in
+ iter_array_tag 0 a (Array.length a) t.doc node size
+*)
external tree_closing : tree -> [`Tree] Node.t -> [`Tree] Node.t = "caml_xml_tree_closing" "noalloc"
let closing t n = tree_closing t.doc n
let size t = tree_size t.doc
-let stats t =
- let tree = t.doc in
- let rec loop left node acc_d total_d num_leaves =
- if node == nil then
- (acc_d+total_d,if left then num_leaves+1 else num_leaves)
- else
- let d,td = loop true (tree_first_child tree node) (acc_d+1) total_d num_leaves in
- loop false (tree_next_sibling tree node) (acc_d) d td
- in
- let a,b = loop true root 0 0 0
- in
- Printf.eprintf "Average depth: %f, number of leaves %i\n%!" ((float_of_int a)/. (float_of_int b)) b
-;;
+
module TagS =
- struct
- include Ptset.Make (
- struct type t = int
- type data = t
- external hash : t -> int = "%identity"
- external uid : t -> Uid.t = "%identity"
- external equal : t -> t -> bool = "%eq"
- external make : t -> int = "%identity"
- external node : t -> int = "%identity"
- external stats : unit -> unit = "%identity"
- end
- )
- let to_ptset s = fold (Ptset.Int.add) s Ptset.Int.empty
- end
+struct
+ include Ptset.Make (
+ struct type t = int
+ type data = t
+ external hash : t -> int = "%identity"
+ external uid : t -> Uid.t = "%identity"
+ external equal : t -> t -> bool = "%eq"
+ external make : t -> int = "%identity"
+ external node : t -> int = "%identity"
+ external stats : unit -> unit = "%identity"
+ external init : unit -> unit = "%identity"
+ end
+ )
+ let to_ptset s = fold (Ptset.Int.add) s Ptset.Int.empty
+end
module TSTSCache =
Hashtbl.Make(struct type t = TagS.t * TagS.t
let equal u v =
let u1,u2 = u
and v1,v2 = v in
- u1 == v1 && u2 == v2
- end)
+ u1 == v1 && u2 == v2
+ end)
module TagTSCache =
Hashtbl.Make(struct type t = Tag.t * TagS.t
let hash (x, y) =
let equal u v =
let u1,u2 = u
and v1,v2 = v in
- u1 == v1 && u2 == v2
- end)
+ u1 == v1 && u2 == v2
+ end)
let add_cache = TagTSCache.create 1023
let union_cache = TSTSCache.create 1023
if y == TagS.empty then false
else
let key = (x, y) in
- try
- TSTSCache.find subset_cache key
- with
- | Not_found ->
- let z = TagS.subset x y in
- TSTSCache.add subset_cache key z; z
+ try
+ TSTSCache.find subset_cache key
+ with
+ | Not_found ->
+ let z = TagS.subset x y in
+ TSTSCache.add subset_cache key z; z
let order ((x, y) as z) =
if x.TagS.Node.id <= y.TagS.Node.id then z
if _subset x y then y
else if _subset y x then x
else
- let key = order (x, y) in
- try
- TSTSCache.find union_cache key
+ let key = order (x, y) in
+ try
+ TSTSCache.find union_cache key
with
- | Not_found ->
- let z = TagS.union x y in
- TSTSCache.add union_cache key z; z
+ | Not_found ->
+ let z = TagS.union x y in
+ TSTSCache.add union_cache key z; z
let _add t s =
let key = (t,s) in
- try
- TagTSCache.find add_cache key
- with
- | Not_found ->
- let z = TagS.add t s in
- TagTSCache.add add_cache key z;z
+ try
+ TagTSCache.find add_cache key
+ with
+ | Not_found ->
+ let z = TagS.add t s in
+ TagTSCache.add add_cache key z;z
let child_sibling_labels tree =
let table_c = Array.create (tree_num_tags tree) TagS.empty in
let siblings = loop (tree_next_sibling tree node) in
let () =
let tn = table_n.(tag) in
- if _subset siblings tn then ()
- else table_n.(tag) <- _union tn siblings
+ if _subset siblings tn then ()
+ else table_n.(tag) <- _union tn siblings
in
- _add tag siblings
+ _add tag siblings
in
- ignore (loop root);
- table_c, table_n
+ ignore (loop root);
+ table_c, table_n
let descendant_labels tree =
let table_d = Array.create (tree_num_tags tree) TagS.empty in
let tag = tree_tag tree node in
let () =
let td = table_d.(tag) in
- if _subset d1 td then ()
- else table_d.(tag) <- _union td d1;
+ if _subset d1 td then ()
+ else table_d.(tag) <- _union td d1;
in
- _add tag (_union d1 d2)
+ _add tag (_union d1 d2)
in
- ignore (loop root);
- table_d
+ ignore (loop root);
+ table_d
let collect_labels tree =
let table_f = Array.create (tree_num_tags tree) TagS.empty in
let tag = tree_tag tree node in
let () =
let tf = table_f.(tag) in
- if _subset followings tf then ()
- else table_f.(tag) <- _union tf followings in
+ if _subset followings tf then ()
+ else table_f.(tag) <- _union tf followings in
let () =
let tn = table_n.(tag) in
- if _subset foll_siblings tn then ()
- else table_n.(tag) <- _union tn foll_siblings in
+ if _subset foll_siblings tn then ()
+ else table_n.(tag) <- _union tn foll_siblings in
let children, n_descendants, n_followings =
loop (tree_last_child tree node) TagS.empty TagS.empty followings
in
let () =
let tc = table_c.(tag) in
- if _subset children tc then ()
- else table_c.(tag) <- _union tc children
+ if _subset children tc then ()
+ else table_c.(tag) <- _union tc children
in
let () =
let td = table_d.(tag) in
- if _subset n_descendants td then ()
- else table_d.(tag) <- _union td n_descendants
+ if _subset n_descendants td then ()
+ else table_d.(tag) <- _union td n_descendants
in
- loop (tree_prev_sibling tree node)
- (_add tag foll_siblings)
- (_add tag (_union n_descendants descendants))
- (_add tag n_followings)
+ loop (tree_prev_sibling tree node)
+ (_add tag foll_siblings)
+ (_add tag (_union n_descendants descendants))
+ (_add tag n_followings)
in
- ignore (loop root TagS.empty TagS.empty TagS.empty);
- table_f, table_n, table_c, table_d
+ ignore (loop root TagS.empty TagS.empty TagS.empty);
+ table_f, table_n, table_c, table_d
let is_nil t = t == nil
let is_root t = t == root
let node_of_t t =
- let _ = Tag.init (Obj.magic t) in
- let f, n, c, d = time collect_labels t ~msg:"Building tag relationship table" in
+ LOG ( __ "indexing" 2 "%s\n" "Initializing tag structure");
+ let _ = Tag.init (mk_tag_ops t) in
+ LOG ( __ "indexing" 2 "%s\n" "Starting tag table construction");
+ let f, n, c, d = Utils.time ~msg:"Building tag relationship table" collect_labels t in
let c = Array.map TagS.to_ptset c in
let n = Array.map TagS.to_ptset n in
let f = Array.map TagS.to_ptset f in
(Ptset.Int.remove Tag.pcdata
(Ptset.Int.diff d.(Tag.document_node) attributes))
in
- { doc= t;
- attributes = attributes;
- attribute_array = Array.of_list (Ptset.Int.elements attributes);
- elements = elements;
- children = c;
- siblings = n;
- descendants = d;
- followings = f
+ { doc= t;
+ attributes = attributes;
+ attribute_array = tag_list_of_set attributes;
+ elements = elements;
+ children = c;
+ siblings = n;
+ descendants = d;
+ followings = f
- }
+ }
let parse_xml_uri str = node_of_t (TreeBuilder.parse_file str)
let size t = tree_size t.doc;;
-external pool : tree -> Tag.pool = "%identity"
-
let magic_string = "SXSI_INDEX"
-let version_string = "3"
+let version_string = "4"
let pos fd =
Unix.lseek fd 0 Unix.SEEK_CUR
-let pr_pos fd = Printf.eprintf "At position %i\n%!" (pos fd)
+let pr_pos fd = Logger.print err_formatter "At position %i@\n" (pos fd)
let write fd s =
let sl = String.length s in
let ssl = Printf.sprintf "%020i" sl in
- ignore (Unix.write fd ssl 0 20);
- ignore (Unix.write fd s 0 (String.length s))
+ ignore (Unix.write fd ssl 0 20);
+ ignore (Unix.write fd s 0 (String.length s))
let rec really_read fd buffer start length =
if length <= 0 then () else
match Unix.read fd buffer start length with
- 0 -> raise End_of_file
- | r -> really_read fd buffer (start + r) (length - r);;
+ 0 -> raise End_of_file
+ | r -> really_read fd buffer (start + r) (length - r);;
let read fd =
let buffer = String.create 20 in
let size = int_of_string buffer in
let buffer = String.create size in
let _ = really_read fd buffer 0 size in
- buffer
+ buffer
let save_tag_table channel t =
let t = Array.map (fun s -> Array.of_list (Ptset.Int.elements s)) t in
- Marshal.to_channel channel t []
+ Marshal.to_channel channel t []
let save t str =
let fd = Unix.openfile str [ Unix.O_WRONLY;Unix.O_TRUNC;Unix.O_CREAT] 0o644 in
let out_c = Unix.out_channel_of_descr fd in
+ let index_prefix = Filename.chop_suffix str ".srx" in
let _ = set_binary_mode_out out_c true in
- output_string out_c magic_string;
- output_char out_c '\n';
- output_string out_c version_string;
- output_char out_c '\n';
- save_tag_table out_c t.children;
- save_tag_table out_c t.siblings;
- save_tag_table out_c t.descendants;
- save_tag_table out_c t.followings;
+ output_string out_c magic_string;
+ output_char out_c '\n';
+ output_string out_c version_string;
+ output_char out_c '\n';
+ save_tag_table out_c t.children;
+ save_tag_table out_c t.siblings;
+ save_tag_table out_c t.descendants;
+ save_tag_table out_c t.followings;
(* we need to move the fd to the correct position *)
- flush out_c;
- ignore (Unix.lseek fd (pos_out out_c) Unix.SEEK_SET);
- tree_save t.doc fd str;
- close_out out_c
+ flush out_c;
+ ignore (Unix.lseek fd (pos_out out_c) Unix.SEEK_SET);
+ tree_save t.doc fd index_prefix;
+ close_out out_c
;;
let load_tag_table channel =
let table : int array array = Marshal.from_channel channel in
- Array.map (fun a -> Ptset.Int.from_list (Array.to_list a)) table
+ Array.map (fun a -> Ptset.Int.from_list (Array.to_list a)) table
let load ?(sample=64) ?(load_text=true) str =
let fd = Unix.openfile str [ Unix.O_RDONLY ] 0o644 in
let in_c = Unix.in_channel_of_descr fd in
+ let index_prefix = Filename.chop_suffix str ".srx" in
let _ = set_binary_mode_in in_c true in
let load_table () =
(let ms = input_line in_c in if ms <> magic_string then failwith "Invalid index file");
- (let vs = input_line in_c in if vs <> version_string then failwith "Invalid version file");
+ (let vs = input_line in_c in if vs <> version_string then failwith "Unsupported index format");
let c = load_tag_table in_c in
let s = load_tag_table in_c in
let d = load_tag_table in_c in
let f = load_tag_table in_c in
- c,s,d,f
+ c,s,d,f
in
let c, s, d, f = time ~msg:"Loading tag table"(load_table) () in
ignore(Unix.lseek fd (pos_in in_c) Unix.SEEK_SET);
- let xml_tree = tree_load fd str load_text sample in
- let () = Tag.init (Obj.magic xml_tree) in
- let attributes = Ptset.Int.add Tag.attribute d.(Tag.attribute) in
- let elements = Ptset.Int.add Tag.document_node
- (Ptset.Int.remove Tag.pcdata
- (Ptset.Int.diff d.(Tag.document_node) attributes))
- in
- let tree = { doc = xml_tree;
- attributes = attributes;
- attribute_array = Array.of_list (Ptset.Int.elements attributes);
- elements = elements;
- children = c;
- siblings = s;
- descendants = d;
- followings = f
+ let xml_tree = tree_load fd index_prefix load_text sample in
+ let () = Tag.init (Obj.magic xml_tree) in
+ let attributes = Ptset.Int.add Tag.attribute d.(Tag.attribute) in
+ let elements = Ptset.Int.add Tag.document_node
+ (Ptset.Int.remove Tag.pcdata
+ (Ptset.Int.diff d.(Tag.document_node) attributes))
+ in
+ let tree = { doc = xml_tree;
+ attributes = attributes;
+ attribute_array = tag_list_of_set attributes;
+ elements = elements;
+ children = c;
+ siblings = s;
+ descendants = d;
+ followings = f
}
in close_in in_c;
tree
-let tag_pool t = pool t.doc
-
let equal a b = a == b
let nts = function
- -1 -> "Nil"
+-1 -> "Nil"
| i -> Printf.sprintf "Node (%i)" i
let dump_node t = nts (Node.to_int t)
external tree_flush : tree -> Unix.file_descr -> unit = "caml_xml_tree_flush"
let flush t fd = tree_flush t.doc fd
-external text_prefix : tree -> string -> query_result = "caml_text_collection_prefix_bv"
-let text_prefix t s = text_prefix t.doc s
+external text_prefix : tree -> string -> bool -> query_result = "caml_text_collection_prefix_bv"
+let text_prefix t s b = text_prefix t.doc s b
-external text_suffix : tree -> string -> query_result = "caml_text_collection_suffix_bv"
-let text_suffix t s = text_suffix t.doc s
+external text_suffix : tree -> string -> bool -> query_result = "caml_text_collection_suffix_bv"
+let text_suffix t s b = text_suffix t.doc s b
-external text_equals : tree -> string -> query_result = "caml_text_collection_equals_bv"
-let text_equals t s = text_equals t.doc s
+external text_equals : tree -> string -> bool -> query_result = "caml_text_collection_equals_bv"
+let text_equals t s b = text_equals t.doc s b
-external text_contains : tree -> string -> query_result = "caml_text_collection_contains_bv"
-let text_contains t s = text_contains t.doc s
+external text_contains : tree -> string -> bool -> query_result = "caml_text_collection_contains_bv"
+let text_contains t s b = text_contains t.doc s b
module Predicate = Hcons.Make (
type t = (_t -> node -> bool) ref
let hash t = Hashtbl.hash t
let equal t1 t2 = t1 == t2
-end)
+ end)
let string_of_query query =
- match query with
- | `Prefix -> "starts-with"
- | `Suffix -> "ends-with"
- | `Equals -> "equals"
- | `Contains -> "contains"
+ match query with
+ | `Prefix -> "starts-with"
+ | `Suffix -> "ends-with"
+ | `Equals -> "equals"
+ | `Contains -> "contains"
;;
let query_fun = function
- | `Prefix -> text_prefix
- | `Suffix -> text_suffix
- | `Equals -> text_equals
- | `Contains -> text_contains
+ | `Prefix -> text_prefix
+ | `Suffix -> text_suffix
+ | `Equals -> text_equals
+ | `Contains -> text_contains
;;
let _pred_cache = Hashtbl.create 17
;;
let mk_pred query s =
- let f = query_fun query in
+ LOG ( __ "bottom-up" 3 "Calling mk_pred for '%s'\n" s);
+ let f = query_fun query in
let memo = ref (fun _ _ -> failwith "Undefined") in
memo := begin fun tree node ->
let results =
try Hashtbl.find _pred_cache (query,s) with
- Not_found ->
- time ~count:1 ~msg:(Printf.sprintf "Computing text query %s(%s)"
- (string_of_query query) s)
- (f tree) s
+ Not_found ->
+ time ~count:1 ~msg:(Printf.sprintf "Computing text query %s(%s)"
+ (string_of_query query) s)
+ (f tree) s true
in
let bv = results.bv in
memo := begin fun _ n ->
- let b =
- bit_vector_unsafe_get bv (Node.to_int n)
- in
- D_TRACE_(Printf.eprintf "Result of memoized call to query %s is %b for node %i\n" s b (Node.to_int n));
- b
+ let r = bit_vector_unsafe_get bv (Node.to_int n) in
+ LOG( __ "bottom-up" 3 "Running predicate on node %i = %b@\n" (Node.to_int n) r);
+ r
end;
- let b = bit_vector_unsafe_get bv (Node.to_int node) in
- D_TRACE_(Printf.eprintf "Result is %b for node %i\n" b (Node.to_int node));
- b
+ let r = bit_vector_unsafe_get bv (Node.to_int node) in
+ LOG( __ "bottom-up" 3 "Running predicate on node %i = %b@\n" (Node.to_int node) r);
+ r
end;
Predicate.make memo
-let full_text_prefix t s = (text_prefix t s).pos
+let full_text_prefix t s = (text_prefix t s true).pos
-let full_text_suffix t s = (text_suffix t s).pos
+let full_text_suffix t s = (text_suffix t s true).pos
-let full_text_equals t s = (text_equals t s).pos
+let full_text_equals t s = (text_equals t s true).pos
-let full_text_contains t s = (text_contains t s).pos
+let full_text_contains t s = (text_contains t s true).pos
let full_text_query q t s =
- let res = (query_fun q) t s in
+ let res = (query_fun q) t s true in
Hashtbl.replace _pred_cache (q,s) res;
res.pos
+
+let stats tree =
+ let h = Hashtbl.create 1024 in
+ let depth = ref 0 in
+ let numleaves = ref 0 in
+ let numtexts = ref 0 in
+ let rec traverse tree t p d =
+ if is_nil t then
+ let oldc =
+ try
+ Hashtbl.find h p
+ with Not_found -> 0
+ in
+ Hashtbl.replace h p (oldc + 1);
+ if d > !depth then depth := d;
+ incr numleaves
+ else
+ let label = tree_tag tree t in
+ if label == Tag.pcdata || label == Tag.attribute_data then incr numtexts;
+ iter_siblings tree t (label::p) (d+1)
+ and iter_siblings tree t p d =
+ if is_nil t then () else
+ let fs = tree_first_child tree t in
+ traverse tree fs p d;
+ let ns = tree_next_sibling tree t in
+ iter_siblings tree ns p d
+ in
+ traverse tree.doc root [] 0;
+ let sumdepth = Hashtbl.fold (fun p c acc -> (List.length p) * c + acc) h 0 in
+ let alltags = Ptset.Int.union tree.elements tree.attributes in
+ Logger.print err_formatter "Statistics :@\n\
+Average depth: %f@\n\
+Longest path: %i@\n\
+Number of distinct paths: %i@\n\
+Number of nodes: %i@\n\
+Number of leaves: %i@\n\
+Number of pcdata/cdata nodes: %i@\n\
+Number of distinct tags: %i@\n\
+Largest tag id: %i@\n@?"
+ (float_of_int sumdepth /. float_of_int !numleaves)
+ !depth
+ (Hashtbl.length h)
+ (tree_subtree_size tree.doc root)
+ !numleaves
+ !numtexts
+ (Ptset.Int.cardinal alltags)
+ (Ptset.Int.max_elt alltags)
+
+
+type tree_pointer = tree
+let get_tree_pointer x = x.doc
projects / SXSI / xpathcomp.git / blobdiff