(* Copyright NICTA 2008 *)
(* Distributed under the terms of the LGPL (see LICENCE) *)
(******************************************************************************)
-(*INCLUDE "debug.ml" *)
+INCLUDE "utils.ml"
type tree
type 'a node = int
external parse_xml_string : string -> int -> bool -> bool -> tree = "caml_call_shredder_string"
external save_tree : tree -> string -> unit = "caml_xml_tree_save"
-external load_tree : string -> int -> tree = "caml_xml_tree_load"
+external load_tree : string -> int -> tree = "caml_xml_tree_load"
external nullt : unit -> 'a node = "caml_xml_tree_nullt"
let nil : 'a node = Obj.magic (-1)
-external text_get_text : tree -> [`Text] node -> string = "caml_text_collection_get_text"
+external text_get_tc_text : tree -> [`Text] node -> string = "caml_text_collection_get_text"
external text_is_empty : tree -> [`Text ] node -> bool = "caml_text_collection_empty_text"
let text_is_empty t n =
(equal_node nil n) || text_is_empty t n
-external get_cached_text : tree -> [`Text ] node -> string = "caml_text_collection_get_cached_text"
-
-
-let text_get_text t n =
- if equal_node nil n then ""
- else get_cached_text t n
-external text_size : tree -> int = "caml_text_collection_size"
+
external text_is_contains : tree -> string -> bool = "caml_text_collection_is_contains"
external text_count_contains : tree -> string -> int = "caml_text_collection_count_contains"
external text_count : tree -> string -> int = "caml_text_collection_count"
external text_contains : tree -> string -> [`Text ] node array = "caml_text_collection_contains"
+external text_unsorted_contains : tree -> string -> unit = "caml_text_collection_unsorted_contains"
+external get_cached_text : tree -> [`Text] node -> string = "caml_text_collection_get_cached_text"
+let get_cached_text t x =
+ if x == -1 then ""
+ else get_cached_text t x
-
-external tree_serialize : tree -> string -> unit = "caml_xml_tree_serialize"
+external tree_serialize : tree -> string -> unit = "caml_xml_tree_serialize"
external tree_unserialize : string -> tree = "caml_xml_tree_unserialize"
external tree_next_sibling : tree -> [`Tree] node -> [`Tree] node = "caml_xml_tree_next_sibling"
external tree_prev_sibling : tree -> [`Tree] node -> [`Tree] node = "caml_xml_tree_prev_sibling"
external tree_is_leaf : tree -> [`Tree] node -> bool = "caml_xml_tree_is_leaf"
-
+external tree_last_child : tree -> [`Tree] node -> [`Tree] node = "caml_xml_tree_last_child"
+external tree_is_first_child : tree -> [`Tree] node -> bool = "caml_xml_tree_is_first_child"
+
(* external tag : tree -> [`Tree ] node -> T = "caml_xml_tree_tag"*)
external tree_tag_id : tree -> [`Tree ] node -> Tag.t = "caml_xml_tree_tag_id"
+
let tree_is_last t n = equal_node nil (tree_next_sibling t n)
external tree_my_text : tree -> [`Tree] node -> [`Text ] node = "caml_xml_tree_my_text"
external tree_next_text : tree -> [`Tree] node -> [`Text ] node = "caml_xml_tree_next_text"
external tree_doc_ids : tree -> [`Tree ] node -> [`Text ] node * [`Text ] node = "caml_xml_tree_doc_ids"
+
+let text_size tree = int_of_node (snd ( tree_doc_ids tree (Obj.magic 0) ))
+
external tree_text_xml_id : tree -> [`Text ] node -> int = "caml_xml_tree_text_xml_id"
external tree_node_xml_id : tree -> [`Tree ] node -> int = "caml_xml_tree_node_xml_id"
external tree_is_ancestor : tree -> [`Tree ] node -> [`Tree ] node -> bool = "caml_xml_tree_is_ancestor"
type t = { doc : tree;
node : descr;
- ttable : (Tag.t,(Ptset.t*Ptset.t)) Hashtbl.t;
+ ttable : (Tag.t,(Ptset.t*Ptset.t)) Hashtbl.t;
}
-
-let update h t sb sa =
- let sbelow,safter =
- try
- Hashtbl.find h t
- with
- | Not_found -> Ptset.empty,Ptset.empty
- in
- Hashtbl.replace h t (Ptset.union sbelow sb, Ptset.union safter sa)
+
+let text_size t = text_size t.doc
let collect_tags tree =
+ let h_union = Hashtbl.create 511 in
+ let pt_cup s1 s2 =
+ (* special case, since this is a union we want hash(s1,s2) = hash(s2,s1) *)
+ let x = Ptset.hash s1
+ and y = Ptset.hash s2 in
+ let h = if x < y then HASHINT2(x,y) else HASHINT2(y,x) in
+ try
+ Hashtbl.find h_union h
+ with
+ | Not_found -> let s = Ptset.union s1 s2
+ in
+ Hashtbl.add h_union h s;s
+ in
+ let h_add = Hashtbl.create 511 in
+ let pt_add t s =
+ let k = HASHINT2(Tag.hash t,Ptset.hash s) in
+ try
+ Hashtbl.find h_add k
+ with
+ | Not_found -> let r = Ptset.add t s in
+ Hashtbl.add h_add k r;r
+ in
let h = Hashtbl.create 511 in
+ let sing = Ptset.singleton Tag.pcdata in
+ let update t sb sa =
+ let sbelow,safter =
+ try
+ Hashtbl.find h t
+ with
+ | Not_found ->
+ (sing,sing)
+ in
+ Hashtbl.replace h t (pt_cup sbelow sb, pt_cup safter sa)
+ in
let rec loop id acc =
if equal_node id nil
- then (Ptset.singleton Tag.pcdata, Ptset.add Tag.pcdata acc)
+ then (Ptset.empty,acc)
else
let below2,after2 = loop (tree_next_sibling tree id) acc in
let below1,after1 = loop (tree_first_child tree id) after2 in
let tag = tree_tag_id tree id in
- update h tag below1 after2;
- Ptset.add tag (Ptset.union below1 below2), (Ptset.add tag after1)
+ update tag below1 after2;
+ pt_add tag (pt_cup below1 below2), (pt_add tag after1)
in
let b,a = loop (tree_root tree) Ptset.empty in
- update h Tag.pcdata b a;
+ update Tag.pcdata b a;
h
+
+
let contains_array = ref [| |]
-
+let contains_index = Hashtbl.create 4096
+let in_array _ i =
+ try
+ Hashtbl.find contains_index i
+ with
+ Not_found -> false
+
let init_contains t s =
let a = text_contains t.doc s
in
Array.fast_sort (compare) a;
- contains_array := a
+ contains_array := a;
+ Array.iter (fun x -> Hashtbl.add contains_index x true) !contains_array
+let count_contains t s = text_count_contains t.doc s
+let unsorted_contains t s = text_unsorted_contains t.doc s
+
let init_naive_contains t s =
let i,j = tree_doc_ids t.doc (tree_root t.doc)
in
let rec loop n acc l =
if n >= j then acc,l
else
- let s = text_get_text t.doc n
+ let s = get_cached_text t.doc n
in
if matching s
then loop (n+1) (n::acc) (l+1)
let compare = compare_node end)
end
-let is_nil t = match t.node with
- | Nil -> true
- | Node(i) -> equal_node i nil
- | _ -> false
-
-let is_node t =
-match t.node with
- | Node(i) -> not(equal_node i nil)
- | _ -> false
+let is_nil t = t.node == Nil
+let is_node t = t.node != Nil
-let node_of_t t =
+let node_of_t t =
let _ = Tag.init (Obj.magic t) in
let table = collect_tags t
in
Ptset.iter (fun i -> Printf.eprintf "'%s' " (Tag.to_string i)) sa;
Printf.eprintf "} \n----------------------------------\n";
) table in
-*)
+ let i,j = tree_doc_ids t (tree_root t) in
+ Printf.eprintf "%i docs, range from %i to %i\n%!" (Array.length s) i j;
+ Array.iter (fun i -> print_endline (">>>" ^ i ^ "<<<")) s; *)
{ doc= t;
node = Node(tree_root t);
ttable = table;
}
+let finalize _ = Printf.eprintf "Release the string list !\n%!"
+;;
+
+let parse f str =
+ node_of_t
+ (f str
+ !Options.sample_factor
+ !Options.index_empty_texts
+ !Options.disable_text_collection)
+
+let parse_xml_uri str = parse parse_xml_uri str
+let parse_xml_string str = parse parse_xml_string str
-
-let parse_xml_uri str = node_of_t
- (parse_xml_uri str
- !Options.sample_factor
- !Options.index_empty_texts
- !Options.disable_text_collection)
-
-let parse_xml_string str = node_of_t
- (parse_xml_string str
- !Options.sample_factor
- !Options.index_empty_texts
- !Options.disable_text_collection)
external pool : tree -> Tag.pool = "%identity"
-let save t str = save_tree t.doc str
+
+let save t str = (save_tree t.doc str)
+;;
+
let load ?(sample=64) str =
node_of_t (load_tree str sample)
let equal a b = (compare a b) == 0
-let norm (n : [`Tree ] node ) = if tree_is_nil n then Nil else Node (n)
+let norm (n : [`Tree ] node ) = if n == -1 then Nil else Node (n)
let nts = function
Nil -> "Nil"
| Text (i,j) -> Printf.sprintf "Text (%i, %i)" i j
| Node (i) -> Printf.sprintf "Node (%i)" i
+let dump_node t = nts t.node
+
let mk_nil t = { t with node = Nil }
let root n = { n with node = norm (tree_root n.doc) }
| Node(t) -> (int_of_node t) == 0
| _ -> false
+let is_left n = match n.node with
+ | Node(t) -> (tree_is_first_child n.doc t) && (equal_node nil (tree_prev_text n.doc t))
+ | Text(_,t) -> tree_is_nil t || tree_is_first_child n.doc t
+ | _ -> false
+
+let is_below_right t1 t2 =
+ match (t1.node,t2.node) with
+ | Nil,_ | _,Nil -> false
+ | Node(i1), Node(i2) ->
+ tree_is_ancestor t1.doc (tree_parent t1.doc i1) i2
+ && not (tree_is_ancestor t1.doc i1 i2)
+ | Text(_,i1),Node(i2) -> i1 == i2 ||
+ (tree_is_ancestor t1.doc (tree_parent t1.doc i1) i2 && i1 < i2)
+ | Text(_,i1),Text(i,_) ->
+ let x,y = tree_doc_ids t1.doc i1 in
+ i >= x && i <= y
+ | Node(i1), Text(i,_) ->
+ let i2 = tree_next_sibling t1.doc i1 in
+ let x,y = tree_doc_ids t1.doc i2 in
+ i >= x && i <= y
+
let parent n =
let node' =
match n.node with (* inlined parent *)
| Node i -> { n with node= norm(tree_next_sibling n.doc i) }
| _ -> { n with node = Nil }
-let node_sibling_ctx n _ =
+let node_sibling_ctx n _ =
match n.node with
| Node i -> { n with node= norm(tree_next_sibling n.doc i) }
| _ -> { n with node = Nil }
let text_below t =
let l = Array.length !contains_array in
- if l = 0 then { t with node=Nil }
- else
match t.node with
- | Node(n) ->
+ | Node(n) ->
let i,j = tree_doc_ids t.doc n in
- let id = array_find !contains_array i j
+ let id = if l == 0 then i else (array_find !contains_array i j)
in
+(* Printf.printf "Looking for text below node %i with tag %s in range %i %i, in array : [|\n%!"
+ n (Tag.to_string (tree_tag_id t.doc n)) i j;
+ Array.iter (fun i -> Printf.printf "%i " (int_of_node i )) !contains_array;
+ Printf.printf "|]\nResult is %i\n%!" id; *)
if id == nil then
{ t with node=Nil }
else
{ t with node = Text(id, tree_next_sibling t.doc (tree_prev_doc t.doc id)) }
- | _ -> { t with node = Nil }
+ | _ -> (*Printf.printf "Here\n%!"; *)
+ { t with node = Nil }
let text_next t root =
let l = Array.length !contains_array in
- if l = 0 then { t with node=Nil }
- else
let inf = match t.node with
- | Node(n) -> snd(tree_doc_ids t.doc n)+1
+ | Node(n) -> snd(tree_doc_ids t.doc n)+1
| Text(i,_) -> i+1
| _ -> assert false
in
match root.node with
| Node (n) ->
- let _,j = tree_doc_ids t.doc n in
- let id = array_find !contains_array inf j
+ let _,j = tree_doc_ids t.doc n in
+ let id = if l == 0 then if inf > j then nil else inf
+ else array_find !contains_array inf j
in
if id == nil then { t with node= Nil }
else
let rec loop ?(print_right=true) t =
match t.node with
| Nil -> ()
- | Text(i,n) -> output_string outc (text_get_text t.doc i);
+ | Text(i,n) -> output_string outc (get_cached_text t.doc i);
if print_right
then loop (right t)
| Node (n) ->
| Node(_) ->
let value =
match (left a).node with
- | Text(i,_) -> text_get_text a.doc i
+ | Text(i,_) -> (get_cached_text a.doc i)
| _ -> assert false
in
output_char outc ' ';
snd(Hashtbl.find t.ttable tag)
let tags t tag = Hashtbl.find t.ttable tag
+
+let tagged_lowest t tag =
+ let rec loop_lowest i =
+ let j = tree_tagged_desc t.doc i tag in
+ if tree_is_nil j then i else loop_lowest j
+ in
+ match t.node with
+ | Node i ->
+ let j = loop_lowest i in
+ { t with
+ node = norm(
+ if tree_is_nil j then
+ if (tree_tag_id t.doc i) == tag
+ then i
+ else j
+ else j) }
+ | Nil -> t
+ | _ -> assert false
+
+
+let tagged_next t tag =
+ match t.node with
+ | Node(i) ->
+ let n = tree_tagged_foll_below t.doc i tag (Obj.magic 0)
+ in
+ if tree_is_nil n then mk_nil t
+ else
+ tagged_lowest { t with node = Node n } tag
+ | Nil -> t
+ | _ -> assert false
+
+let rec binary_parent t =
+ let res =
+ match t.node with
+ | Node(0) -> { t with node = Nil }
+ | Node(i) ->
+ let j = tree_prev_sibling t.doc i in
+ if tree_is_nil j then
+ let idoc = tree_prev_text t.doc i in
+ if equal_node nil idoc then
+ { t with node = Node (tree_parent t.doc i) }
+ else
+ { t with node = Text(idoc,i) }
+ else
+ let idoc = tree_prev_text t.doc i in
+ if equal_node nil idoc then
+ { t with node = Node (j) }
+ else { t with node = Text(idoc,i) }
+ | Text(d,i) ->
+ if tree_is_nil i then
+ let n = tree_parent_doc t.doc d in
+ let lc = tree_last_child t.doc n in
+ if tree_is_nil lc then {t with node = Node n }
+ else { t with node = Node lc }
+ else
+ let j = tree_prev_sibling t.doc i in
+ if tree_is_nil j then
+ { t with node = Node (tree_parent t.doc i) }
+ else { t with node = Node j }
+ | Nil -> t
+ in match res.node with
+ | Text(idoc,t) ->
+ if (Array.length !contains_array) != 0
+ then if in_array !contains_array idoc then res
+ else binary_parent res
+ else res
+ | _ -> res
+
+let benchmark_text t =
+ let doc = t.doc in
+ match (root t).node with
+ | Node i -> let _,size = tree_doc_ids doc i in
+ Printf.eprintf "%i will take ~ %i seconds\n%!"
+ size (size/10000) ;
+ let a = Array.create size "" in
+ for i = 0 to size
+ do
+ a.(i) <- text_get_tc_text t.doc (i+1)
+ done; a
+ | _ -> assert false
+
+let doc_ids (t:t) : (int*int) =
+ (Obj.magic (
+ match t.node with
+ | Node i -> tree_doc_ids t.doc i
+ | Text (i,_) -> (i,i)
+ | Nil -> (nil,nil)
+ ))
+
+let subtree_tags t tag = match t.node with
+ | Nil -> 0
+ | Node(i) -> tree_subtree_tags t.doc i tag
+ | Text(_,i) -> tree_subtree_tags t.doc i tag