(******************************************************************************)
INCLUDE "utils.ml"
+
+external init_lib : unit -> unit = "caml_init_lib"
+
+exception CPlusPlusError of string
+
+let () = Callback.register_exception "CPlusPlusError" (CPlusPlusError "")
+
+let () = init_lib ()
+
+
type tree
-type 'a node = int
+type 'a node = private int
type node_kind = [`Text | `Tree ]
-
-let compare_node : 'a node -> 'a node -> int = (-)
+
+external inode : 'a node -> int = "%identity"
+external nodei : int -> 'a node = "%identity"
+let compare_node x y = (inode x) - (inode y)
let equal_node : 'a node -> 'a node -> bool = (==)
-
-(* abstract type, values are pointers to a XMLTree C++ object *)
-external int_of_node : 'a node -> int = "%identity"
external parse_xml_uri : string -> int -> bool -> bool -> tree = "caml_call_shredder_uri"
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 tree_save : tree -> Unix.file_descr -> unit = "caml_xml_tree_save"
+external tree_load : Unix.file_descr -> tree = "caml_xml_tree_load"
external nullt : unit -> 'a node = "caml_xml_tree_nullt"
-let nil : 'a node = Obj.magic (-1)
+let nil : [`Tree ] node = nodei ~-1
+let nulldoc : [`Text ] node = nodei ~-1
+let root : [`Tree ] node = nodei 0
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
-
-
+let text_is_empty t n = (equal_node nulldoc n) || text_is_empty t n
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"
+external text_get_cached_text : tree -> [`Text] node -> string = "caml_text_collection_get_cached_text"
+
+external tree_root : tree -> [`Tree] node = "caml_xml_tree_root"
+let tree_is_nil x = equal_node x nil
-external tree_serialize : tree -> string -> unit = "caml_xml_tree_serialize"
+external tree_parent : tree -> [`Tree] node -> [`Tree] node = "caml_xml_tree_parent" "noalloc"
+external tree_parent_doc : tree -> [`Text ] node -> [`Tree ] node = "caml_xml_tree_parent_doc" "noalloc"
+(*external tree_prev_doc : tree -> [`Text ] node -> [`Tree ] node = "caml_xml_tree_prev_doc" "noalloc" *)
+external tree_first_child : tree -> [`Tree] node -> [`Tree] node = "caml_xml_tree_first_child" "noalloc"
+external tree_first_element : tree -> [`Tree] node -> [`Tree] node = "caml_xml_tree_first_element" "noalloc"
+external tree_tagged_child : tree -> [`Tree] node -> Tag.t -> [`Tree] node = "caml_xml_tree_tagged_child" "noalloc"
+external tree_next_sibling : tree -> [`Tree] node -> [`Tree] node = "caml_xml_tree_next_sibling" "noalloc"
+external tree_next_element : tree -> [`Tree] node -> [`Tree] node = "caml_xml_tree_next_element" "noalloc"
+external tree_tagged_sibling : tree -> [`Tree] node -> Tag.t -> [`Tree] node = "caml_xml_tree_tagged_sibling" "noalloc"
-external tree_unserialize : string -> tree = "caml_xml_tree_unserialize"
-
-external tree_root : tree -> [`Tree] node = "caml_xml_tree_root"
+external tree_prev_sibling : tree -> [`Tree] node -> [`Tree] node = "caml_xml_tree_prev_sibling" "noalloc"
+external tree_is_leaf : tree -> [`Tree] node -> bool = "caml_xml_tree_is_leaf" "noalloc"
+external tree_last_child : tree -> [`Tree] node -> [`Tree] node = "caml_xml_tree_last_child" "noalloc"
+external tree_is_first_child : tree -> [`Tree] node -> bool = "caml_xml_tree_is_first_child" "noalloc"
-let tree_is_nil x = equal_node x nil
-external tree_parent : tree -> [`Tree] node -> [`Tree] node = "caml_xml_tree_parent"
-external tree_parent_doc : tree -> [`Text ] node -> [`Tree ] node = "caml_xml_tree_parent_doc"
-external tree_prev_doc : tree -> [`Text ] node -> [`Tree ] node = "caml_xml_tree_prev_doc"
-external tree_first_child : tree -> [`Tree] node -> [`Tree] node = "caml_xml_tree_first_child"
-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"
+external tree_tag_id : tree -> [`Tree ] node -> Tag.t = "caml_xml_tree_tag_id" "noalloc"
let tree_is_last t n = equal_node nil (tree_next_sibling t n)
-external tree_prev_text : tree -> [`Tree] node -> [`Text ] node = "caml_xml_tree_prev_text"
+(*external tree_prev_text : tree -> [`Tree] node -> [`Text ] node = "caml_xml_tree_prev_text" "noalloc" *)
-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_my_text : tree -> [`Tree] node -> [`Text ] node = "caml_xml_tree_my_text" "noalloc"
+(*external tree_next_text : tree -> [`Tree] node -> [`Text ] node = "caml_xml_tree_next_text" "noalloc" *)
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) ))
+let text_size tree = inode (snd ( tree_doc_ids tree root ))
-let get_cached_text t x =
- if x == -1 then ""
+let text_get_cached_text t (x:[`Text] node) =
+ if x == nulldoc then ""
else
- get_cached_text t x
-
-
-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"
-external tree_tagged_desc : tree -> [`Tree ] node -> Tag.t -> [`Tree ] node = "caml_xml_tree_tagged_desc"
-external tree_tagged_foll_below : tree -> [`Tree ] node -> Tag.t -> [`Tree ] node -> [`Tree ] node = "caml_xml_tree_tagged_foll_below"
-external tree_subtree_tags : tree -> [`Tree ] node -> Tag.t -> int = "caml_xml_tree_subtree_tags"
-(*
-external tree_select_below : tree -> [`Tree ] node -> Ptset.int_vector -> Ptset.int_vector -> [`Tree ] node = "caml_xml_tree_select_below"
-external tree_select_desc_only : tree -> [`Tree ] node -> Ptset.int_vector -> [`Tree ] node = "caml_xml_tree_select_desc_only"
-external tree_select_next : tree -> [`Tree ] node -> Ptset.int_vector -> Ptset.int_vector -> [`Tree ] node -> [`Tree ] node = "caml_xml_tree_select_next"
-external tree_select_foll_only : tree -> [`Tree ] node -> Ptset.int_vector -> [`Tree ] node -> [`Tree ] node = "caml_xml_tree_select_foll_only"
-external tree_select_desc_or_foll_only : tree -> [`Tree ] node -> Ptset.int_vector -> [`Tree ] node -> [`Tree ] node = "caml_xml_tree_select_foll_only" *)
-
-type descr =
- | Nil
- | Node of [`Tree] node
- | Text of [`Text] node * [`Tree] node
+ text_get_cached_text t x
+
+
+external tree_text_xml_id : tree -> [`Text ] node -> int = "caml_xml_tree_text_xml_id" "noalloc"
+external tree_node_xml_id : tree -> [`Tree ] node -> int = "caml_xml_tree_node_xml_id" "noalloc"
+external tree_is_ancestor : tree -> [`Tree ] node -> [`Tree ] node -> bool = "caml_xml_tree_is_ancestor" "noalloc"
+external tree_tagged_desc : tree -> [`Tree ] node -> Tag.t -> [`Tree ] node = "caml_xml_tree_tagged_desc" "noalloc"
+external tree_tagged_foll_below : tree -> [`Tree ] node -> Tag.t -> [`Tree ] node -> [`Tree ] node = "caml_xml_tree_tagged_foll_below" "noalloc"
+external tree_subtree_tags : tree -> [`Tree ] node -> Tag.t -> int = "caml_xml_tree_subtree_tags" "noalloc"
+
+
+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"
+
+external tree_select_child : tree -> [`Tree ] node -> unordered_set -> [`Tree] node = "caml_xml_tree_select_child" "noalloc"
+external tree_select_foll_sibling : tree -> [`Tree ] node -> unordered_set -> [`Tree] node = "caml_xml_tree_select_foll_sibling" "noalloc"
+external tree_select_desc : tree -> [`Tree ] node -> unordered_set -> [`Tree] node = "caml_xml_tree_select_desc" "noalloc"
+external tree_select_foll_below : tree -> [`Tree ] node -> unordered_set -> [`Tree] node -> [`Tree] node = "caml_xml_tree_select_foll_below" "noalloc"
+
+
+module HPtset = Hashtbl.Make(Ptset.Int)
+
+let vector_htbl = HPtset.create MED_H_SIZE
+
+let ptset_to_vector 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
+
-type t = { doc : tree;
- node : descr;
- ttable : (Tag.t,(Ptset.Int.t*Ptset.Int.t)) Hashtbl.t;
- }
+type t = {
+ doc : tree;
+ ttable : (Tag.t,(Ptset.Int.t*Ptset.Int.t*Ptset.Int.t*Ptset.Int.t)) Hashtbl.t;
+}
let text_size t = text_size t.doc
+
module MemUnion = Hashtbl.Make (struct
type t = Ptset.Int.t*Ptset.Int.t
let equal (x,y) (z,t) = (Ptset.Int.equal x z)&&(Ptset.Int.equal y t)
if x < y then HASHINT2(x,y) else HASHINT2(y,x)
end)
+module MemAdd = Hashtbl.Make (
+ struct
+ type t = Tag.t*Ptset.Int.t
+ let equal (x,y) (z,t) = (x == z)&&(Ptset.Int.equal y t)
+ let hash (x,y) = HASHINT2(x,Ptset.Int.hash y)
+ end)
+
let collect_tags tree =
let h_union = MemUnion.create BIG_H_SIZE in
let pt_cup s1 s2 =
in
MemUnion.add h_union (s1,s2) s;s
in
- let h_add = Hashtbl.create BIG_H_SIZE in
- let pt_add t s =
- let k = HASHINT2(Tag.hash t,Ptset.Int.hash s) in
- try
- Hashtbl.find h_add k
- with
+ let h_add = MemAdd.create BIG_H_SIZE in
+ let pt_add t s =
+ try MemAdd.find h_add (t,s)
+ with
| Not_found -> let r = Ptset.Int.add t s in
- Hashtbl.add h_add k r;r
+ MemAdd.add h_add (t,s) r;r
in
let h = Hashtbl.create BIG_H_SIZE in
- let sing = Ptset.Int.singleton Tag.pcdata in
- let update t sb sa =
- let sbelow,safter =
+ let update t sc sb ss sa =
+ let schild,sbelow,ssibling,safter =
try
Hashtbl.find h t
with
| Not_found ->
- (sing,sing)
+ (Ptset.Int.empty,Ptset.Int.empty,Ptset.Int.empty,Ptset.Int.empty)
in
- Hashtbl.replace h t (pt_cup sbelow sb, pt_cup safter sa)
+ Hashtbl.replace h t
+ (pt_cup sc schild,pt_cup sbelow sb, pt_cup ssibling ss, pt_cup safter sa)
in
- let rec loop id acc =
- if equal_node id nil
- then (Ptset.Int.empty,acc)
+ let rec loop_right id acc_sibling acc_after=
+ if id == nil
+ then (acc_sibling,acc_after)
+ else
+ let sibling2,after2 = loop_right (tree_next_sibling tree id) acc_sibling acc_after in
+ let child1,below1 = loop_left (tree_first_child tree id) after2 in
+ let tag = tree_tag_id tree id in
+ update tag child1 below1 sibling2 after2;
+ (pt_add tag sibling2, (pt_add tag (pt_cup after2 below1)))
+ and loop_left id acc_after =
+ if id == nil
+ then (Ptset.Int.empty,Ptset.Int.empty)
else
- let below2,after2 = loop (tree_next_sibling tree id) acc in
- let below1,after1 = loop (tree_first_child tree id) after2 in
+ let sibling2,after2 = loop_right (tree_next_sibling tree id) Ptset.Int.empty acc_after in
+ let child1,below1 = loop_left (tree_first_child tree id) after2 in
let tag = tree_tag_id tree id in
- update tag below1 after2;
- pt_add tag (pt_cup below1 below2), (pt_add tag after1)
+ update tag child1 below1 sibling2 after2;
+ (pt_add tag sibling2,(pt_add tag (pt_cup after2 below1)))
in
- let b,a = loop (tree_root tree) Ptset.Int.empty in
- update Tag.pcdata b a;
- h
-
-
-
-
+ let _ = loop_left (tree_root tree) Ptset.Int.empty in h
+
+
+
let contains_array = ref [| |]
let contains_index = Hashtbl.create 4096
let rec loop n acc l =
if n >= j then acc,l
else
- let s = get_cached_text t.doc n
+ let s = text_get_cached_text t.doc n
in
if matching s
- then loop (n+1) (n::acc) (l+1)
- else loop (n+1) acc l
+ then loop (nodei ((inode n)+1)) (n::acc) (l+1)
+ else loop (nodei ((inode n)+1)) acc l
in
let acc,l = loop i [] 0 in
- let a = Array.create l nil in
+ let a = Array.create l nulldoc in
let _ = List.fold_left (fun cpt e -> a.(cpt) <- e; (cpt-1)) (l-1) acc
in
contains_array := a
let compare = compare_node end)
end
-let is_nil t = t.node == Nil
+let is_nil t = t == nil
-let is_node t = t.node != Nil
+let is_node t = t != nil
+let is_root t = t == root
let node_of_t t =
let _ = Tag.init (Obj.magic t) in
let table = collect_tags t
in
-(*
- let _ = Hashtbl.iter (fun t (sb,sa) ->
- Printf.eprintf "'%s' -> { " (Tag.to_string t);
- Ptset.iter (fun i -> Printf.eprintf "'%s' " (Tag.to_string i)) sb;
- Printf.eprintf "}\n { ";
- 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%!"
;;
external pool : tree -> Tag.pool = "%identity"
-let save t str = (save_tree t.doc str)
+let magic_string = "SXSI_INDEX"
+let version_string = "1"
+
+let pos fd =
+ Unix.lseek fd 0 Unix.SEEK_CUR
+
+let pr_pos fd = Printf.eprintf "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))
+
+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);;
+
+let read fd =
+ let buffer = String.create 20 in
+ let _ = really_read fd buffer 0 20 in
+ let size = int_of_string buffer in
+ let buffer = String.create size in
+ let _ = really_read fd buffer 0 size in
+ buffer
+
+
+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 _ = 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';
+ Marshal.to_channel out_c t.ttable [ ];
+ (* 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;
+ close_out out_c
;;
let load ?(sample=64) str =
- node_of_t (load_tree str sample)
-
+ let fd = Unix.openfile str [ Unix.O_RDONLY ] 0o644 in
+ let in_c = Unix.in_channel_of_descr fd in
+ let _ = set_binary_mode_in in_c true in
+ (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 table : (Tag.t,(Ptset.Int.t*Ptset.Int.t*Ptset.Int.t*Ptset.Int.t)) Hashtbl.t =
+ Marshal.from_channel in_c
+ in
+ let ntable = Hashtbl.create (Hashtbl.length table) in
+ Hashtbl.iter (fun k (s1,s2,s3,s4) ->
+ let ss1 = Ptset.Int.fold (Ptset.Int.add) s1 Ptset.Int.empty
+ and ss2 = Ptset.Int.fold (Ptset.Int.add) s2 Ptset.Int.empty
+ and ss3 = Ptset.Int.fold (Ptset.Int.add) s3 Ptset.Int.empty
+ and ss4 = Ptset.Int.fold (Ptset.Int.add) s4 Ptset.Int.empty
+ in Hashtbl.add ntable k (ss1,ss2,ss3,ss4)
+ ) table;
+ Hashtbl.clear table;
+ (* The in_channel read a chunk of fd, so we might be after
+ the start of the XMLTree save file. Reset to the correct
+ position *)
+ ignore(Unix.lseek fd (pos_in in_c) Unix.SEEK_SET);
+ let tree = { doc = tree_load fd;
+ ttable = ntable;}
+ in close_in in_c;
+ tree
+
let tag_pool t = pool t.doc
-let compare a b = match a.node,b.node with
- | Nil, Nil -> 0
- | Nil,_ -> 1
- | _ , Nil -> -1
- | Node(i),Node(j) -> compare_node i j
- | Text(i,_), Text(j,_) -> compare_node i j
- | Node(i), Text(_,j) -> compare_node i j
- | Text(_,i), Node(j) -> compare_node i j
-
-let equal a b = (compare a b) == 0
-
-
-let norm (n : [`Tree ] node ) = if n == -1 then Nil else Node (n)
-
+let compare = compare_node
+
+let equal a b = a == b
+
let nts = function
- Nil -> "Nil"
- | Text (i,j) -> Printf.sprintf "Text (%i, %i)" i j
- | Node (i) -> Printf.sprintf "Node (%i)" i
+ -1 -> "Nil"
+ | i -> Printf.sprintf "Node (%i)" i
-let dump_node t = nts t.node
+let dump_node t = nts (inode t)
-let mk_nil t = { t with node = Nil }
-let root n = { n with node = norm (tree_root n.doc) }
-
-let is_root n = match n.node with
- | 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(t) when (int_of_node t)== 0 -> Nil
- | Node(t) ->
- let txt = tree_prev_text n.doc t in
- if text_is_empty n.doc txt then
- let ps = tree_prev_sibling n.doc t in
- if tree_is_nil ps
- then
- Node(tree_parent n.doc t)
- else Node(ps)
- else
- Text(txt,t)
- | Text(i,t) ->
- let ps = tree_prev_doc n.doc i in
- if tree_is_nil ps
- then Node (tree_parent_doc n.doc i)
- else Node(ps)
- | _ -> failwith "parent"
- in
- { n with node = node' }
-
-let node_child n =
- match n.node with
- | Node i -> { n with node= norm(tree_first_child n.doc i) }
- | _ -> { n with node = Nil }
-
-let node_sibling n =
- match n.node with
- | Node i -> { n with node= norm(tree_next_sibling n.doc i) }
- | _ -> { n with node = Nil }
-
-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 first_child n =
- let node' =
- match n.node with
- | Node (t) ->
- let fs = tree_first_child n.doc t in
- if equal_node nil fs
- then
- let txt = tree_my_text n.doc t in
- if equal_node nil txt
- then Nil
- else Text(txt,nil)
- else
- let txt = tree_prev_text n.doc fs in
- if equal_node nil txt
- then Node(fs)
- else Text(txt, fs)
- | Text(_,_) -> Nil
- | Nil -> failwith "first_child"
- in
- { n with node = node'}
-let next_sibling n =
- let node' =
- match n.node with
- | Text (_,ns) -> norm ns
- | Node(t) ->
- let ns = tree_next_sibling n.doc t in
- let txt = tree_next_text n.doc t in
- if equal_node nil txt
- then norm ns
- else Text(txt, ns)
- | Nil -> failwith "next_sibling"
- in
- { n with node = node'}
-
-let next_sibling_ctx n _ = next_sibling n
-
-let left = first_child
-let right = next_sibling
-
-let id t =
- match t.node with
- | Node(n) -> tree_node_xml_id t.doc n
- | Text(i,_) -> tree_text_xml_id t.doc i
- | _ -> -1
-
-let tag t =
- match t.node with
- | Text(_) -> Tag.pcdata
- | Node(n) -> tree_tag_id t.doc n
- | Nil -> Tag.nullt
-
-(*
-let select_next tb tf t s =
- match s.node with
- | Node (below) -> begin
- match t.node with
- | Node( n) ->
- { t with node = norm (tree_select_next t.doc n (Ptset.Int.to_int_vector tb) (Ptset.Int.to_int_vector tf) below) }
- | Text (i,n) when equal_node nil n ->
- let p = tree_parent_doc t.doc i in
- { t with node = norm (tree_select_next t.doc p (Ptset.Int.to_int_vector tb) (Ptset.Int.to_int_vector tf) below) }
- | Text(_,n) ->
- if Ptset.mem (tree_tag_id t.doc n) (Ptset.Int.union tb tf)
- then { t with node=Node(n) }
- else
- let vb = Ptset.Int.to_int_vector tb in
- let vf = Ptset.Int.to_int_vector tf in
- let node =
- let dsc = tree_select_below t.doc n vb vf in
- if equal_node nil dsc
- then tree_select_next t.doc n vb vf below
- else dsc
- in
- { t with node = norm node }
- | _ -> {t with node = Nil }
- end
-
- | _ -> { t with node = Nil }
+let is_left t n = tree_is_first_child t.doc n
-
+let is_below_right t n1 n2 = tree_is_ancestor t.doc (tree_parent t.doc n1) n2
+let parent t n = tree_parent t.doc n
- let select_foll_only tf t s =
- match s.node with
- | Node (below) ->
- begin
- match t.node with
- | Node(n) ->
- { t with node= norm (tree_select_foll_only t.doc n (Ptset.Int.to_int_vector tf) below) }
- | Text(i,n) when equal_node nil n ->
- let p = tree_parent_doc t.doc i in
- { t with node= norm (tree_select_foll_only t.doc p (Ptset.Int.to_int_vector tf) below) }
- | Text(_,n) ->
- if Ptset.mem (tree_tag_id t.doc n) tf
- then { t with node=Node(n) }
- else
- let vf = Ptset.Int.to_int_vector tf in
- let node =
- let dsc = tree_select_desc_only t.doc n vf in
- if tree_is_nil dsc
- then tree_select_foll_only t.doc n vf below
- else dsc
- in
- { t with node = norm node }
- | _ -> { t with node = Nil }
- end
- | _ -> {t with node=Nil }
-
-let select_below tc td t=
- match t.node with
- | Node( n) ->
- let vc = Ptset.Int.to_int_vector tc
- in
- let vd = Ptset.Int.to_int_vector td
- in
- { t with node= norm(tree_select_below t.doc n vc vd) }
- | _ -> { t with node=Nil }
-
-
-let select_desc_only td t =
- match t.node with
- | Node(n) ->
- let vd = Ptset.Int.to_int_vector td
- in
- { t with node = norm(tree_select_desc_only t.doc n vd) }
- | _ -> { t with node = Nil }
+let first_child t = (); fun n -> tree_first_child t.doc n
+let first_element t = (); fun n -> tree_first_element t.doc n
+(* these function will be called in two times: first partial application
+ on the tag, then application of the tag and the tree, then application of
+ the other arguments. We use the trick to let the compiler optimize application
*)
-let tagged_desc tag t =
- match t.node with
- | Node(n) ->
- { t with node = norm(tree_tagged_desc t.doc n tag) }
- | _ -> { t with node = Nil }
-
-
-let tagged_foll_ctx tag t s =
- match s.node with
- | Node (below) ->
- begin
- match t.node with
- | Node(n) ->
- { t with node= norm (tree_tagged_foll_below t.doc n tag below) }
- | Text(i,n) when equal_node nil n ->
- let p = tree_prev_doc t.doc i in
- { t with node= norm (tree_tagged_foll_below t.doc p tag below) }
- | Text(_,n) ->
- if (tree_tag_id t.doc n) == tag
- then { t with node=Node(n) }
- else
- let node =
- let dsc = tree_tagged_desc t.doc n tag in
- if tree_is_nil dsc
- then tree_tagged_foll_below t.doc n tag below
- else dsc
- in
- { t with node = norm node }
- | _ -> { t with node = Nil }
- end
- | _ -> {t with node=Nil }
+let tagged_child t tag = () ; fun n -> tree_tagged_child t.doc n tag
+
+let select_child t = fun ts ->
+ let v = ptset_to_vector ts in ();
+ fun n -> tree_select_child t.doc n v
+
+let next_sibling t = (); fun n -> tree_next_sibling t.doc n
+let next_element t = (); fun n -> tree_next_element t.doc n
+
+let tagged_sibling t tag = (); fun n -> tree_tagged_sibling t.doc n tag
+
+let select_sibling t = fun ts ->
+ let v = (ptset_to_vector ts) in ();
+ fun n -> tree_select_foll_sibling t.doc n v
+
+let next_sibling_ctx t = (); fun n _ -> tree_next_sibling t.doc n
+let next_element_ctx t = (); fun n _ -> tree_next_element t.doc n
+let tagged_sibling_ctx t tag = (); fun n _ -> tree_tagged_sibling t.doc n tag
+
+let select_sibling_ctx t = fun ts ->
+ let v = (ptset_to_vector ts) in ();
+ fun n _ -> tree_select_foll_sibling t.doc n v
+
+let id t n = tree_node_xml_id t.doc n
+
+let tag t n = if n == nil then Tag.nullt else tree_tag_id t.doc n
+
+let tagged_desc t tag = (); fun n -> tree_tagged_desc t.doc n tag
+
+let select_desc t = fun ts ->
+ let v = (ptset_to_vector ts) in ();
+ fun n -> tree_select_desc t.doc n v
+let tagged_foll_ctx t tag = (); fun n ctx -> tree_tagged_foll_below t.doc n tag ctx
+let select_foll_ctx t = fun ts ->
+ let v = (ptset_to_vector ts) in ();
+ fun n ctx -> tree_select_foll_below t.doc n v ctx
let last_idx = ref 0
let array_find a i j =
else loop !last_idx i j
-
-let text_below t =
- let l = Array.length !contains_array in
- match t.node with
- | Node(n) ->
- let i,j = tree_doc_ids t.doc n in
- 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)) }
- | _ -> (*Printf.printf "Here\n%!"; *)
- { t with node = Nil }
-
-let text_next t root =
- let l = Array.length !contains_array in
- let inf = match t.node with
- | 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 = 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
- { t with node = Text(id,tree_next_sibling t.doc (tree_prev_doc t.doc id)) }
- | _ -> { t with node = Nil}
-
-
-(*
- let subtree_tags t tag =
- match t with
- { doc = d; node = Node(NC n) } ->
- subtree_tags d n tag
- | _ -> 0
-
- let select_desc_array = ref [| |]
- let idx = ref 0
-
- let init_tagged_next t tagid =
- let l = subtree_tags (root t) tagid
- in
- tagged_desc_array := Array.create l { t with node= Nil };
- let i = ref 0 in
- let rec collect t =
- if is_node t then begin
- if tag t == tagid then
- begin
- !tagged_desc_array.(!i) <- t;
- incr i;
- end;
- collect (first_child t);
- collect (next_sibling t)
- end;
- in
- collect t;
- idx := 0
-
- let print_id ppf v =
- let pr x= Format.fprintf ppf x in
- match v with
- { node=Nil } -> pr "NULLT: -1"
- | { node=String(i) } | { node=Node(SC(i,_)) } -> pr "DocID: %i" (int_of_node i)
- | { node=Node(NC(i)) } -> pr "Node: %i" (int_of_node i)
-
-
-
-(* let tagged_next t tag =
- if !idx >= Array.length !tagged_desc_array
- then {t with node=Nil}
- else
- let r = !tagged_desc_array.(!idx)
- in
- incr idx; r
-*)
-
-
- let has_tagged_foll t tag = is_node (tagged_foll t tag)
- let has_tagged_desc t tag = is_node (tagged_desc t tag)
-
- let contains t s =
- Array.fold_left (fun a i -> DocIdSet.add i a) DocIdSet.empty (Text.contains t.doc s)
-
-
- let contains_old t s =
- let regexp = Str.regexp_string s in
- let matching arg =
- try
- let _ = Str.search_forward regexp arg 0;
- in true
- with _ -> false
- in
- let rec find t acc = match t.node with
- | Nil -> acc
- | String i ->
- if matching (string t) then DocIdSet.add i acc else acc
- | Node(_) -> (find (left t )) ((find (right t)) acc)
- in
- find t DocIdSet.empty
-
-
- let contains_iter t s =
- let regexp = Str.regexp_string s in
- let matching arg =
- try
- let _ = Str.search_forward regexp arg 0;
- in true
- with _ -> false
- in
- let size = Text.size t.doc in
- let rec find acc n =
- if n == size then acc
- else
- find
- (if matching (Text.get_cached_text t.doc (Obj.magic n)) then
- DocIdSet.add (Obj.magic n) acc
- else acc) (n+1)
- in
- find DocIdSet.empty 0
-
-
-
-
- let count_contains t s = Text.count_contains t.doc s
-*)
let count t s = text_count t.doc s
-(*
- let is_left t =
- if is_root t then false
- else
- if tag (parent t) == Tag.pcdata then false
- else
- let u = left (parent t) in
- (id t) == (id u)
-*)
- let print_xml_fast outc t =
+
+ let print_xml_fast outc tree t =
let rec loop ?(print_right=true) t =
- match t.node with
- | Nil -> ()
- | Text(i,n) -> output_string outc (get_cached_text t.doc i);
- if print_right
- then loop (right t)
- | Node (n) ->
- let tg = Tag.to_string (tag t) in
- let l = left t
- and r = right t
- in
- output_char outc '<';
- output_string outc tg;
- ( match l.node with
- Nil -> output_string outc "/>"
- | Node(_) when Tag.equal (tag l) Tag.attribute ->
- (loop_attributes (left l);
- match (right l).node with
- | Nil -> output_string outc "/>"
- | _ ->
- output_char outc '>';
- loop (right l);
- output_string outc "</";
- output_string outc tg;
- output_char outc '>' )
- | _ ->
+ if t != nil
+ then
+ let tagid = tree_tag_id tree.doc t in
+ if tagid==Tag.pcdata
+ then
+ begin
+ let tid = tree_my_text tree.doc t in
+ let _ = Printf.eprintf "my_text %i returned %i\n%!" (inode t) (inode tid)
+ in
+ output_string outc (text_get_cached_text tree.doc tid);
+ if print_right
+ then loop (next_sibling tree t);
+ end
+ else
+ let tagstr = Tag.to_string tagid in
+ let l = first_child tree t
+ and r = next_sibling tree t
+ in
+ output_char outc '<';
+ output_string outc tagstr;
+ if l == nil then output_string outc "/>"
+ else
+ if (tag tree l) == Tag.attribute then
+ begin
+ loop_attributes (first_child tree l);
+ if (next_sibling tree l) == nil then output_string outc "/>"
+ else
+ begin
+ output_char outc '>';
+ loop (next_sibling tree l);
+ output_string outc "</";
+ output_string outc tagstr;
+ output_char outc '>';
+ end;
+ end
+ else
+ begin
output_char outc '>';
loop l;
output_string outc "</";
- output_string outc tg;
- output_char outc '>'
- );if print_right then loop r
- and loop_attributes a =
- match a.node with
- | Node(_) ->
- let value =
- match (left a).node with
- | Text(i,_) -> (get_cached_text a.doc i)
- | _ -> assert false
- in
- output_char outc ' ';
- output_string outc (Tag.to_string (tag a));
- output_string outc "=\"";
- output_string outc value;
- output_char outc '"';
- loop_attributes (right a)
- | _ -> ()
+ output_string outc tagstr;
+ output_char outc '>';
+ end;
+ if print_right then loop r
+ and loop_attributes a =
+ if a != nil
+ then
+ let s = (Tag.to_string (tag tree a)) in
+ let attname = String.sub s 3 ((String.length s) -3) in
+ let fsa = first_child tree a in
+ let tid = tree_my_text tree.doc fsa in
+ let _ = Printf.eprintf "my_text %i returned %i\n%!" (inode fsa) (inode tid)
+ in
+ output_char outc ' ';
+ output_string outc attname;
+ output_string outc "=\"";
+ output_string outc (text_get_cached_text tree.doc tid);
+ output_char outc '"';
+ loop_attributes (next_sibling tree a)
in
loop ~print_right:false t
- let print_xml_fast outc t =
- if Tag.to_string (tag t) = "" then
- print_xml_fast outc (first_child t)
- else print_xml_fast outc t
+ let print_xml_fast outc tree t =
+ if (tag tree t) = Tag.document_node then
+ print_xml_fast outc tree (first_child tree t)
+ else print_xml_fast outc tree t
-
+let tags_children t tag =
+ let a,_,_,_ = Hashtbl.find t.ttable tag in a
let tags_below t tag =
- fst(Hashtbl.find t.ttable tag)
-
+ let _,a,_,_ = Hashtbl.find t.ttable tag in a
+let tags_siblings t tag =
+ let _,_,a,_ = Hashtbl.find t.ttable tag in a
let tags_after t tag =
- snd(Hashtbl.find t.ttable tag)
+ let _,_,_,a = Hashtbl.find t.ttable tag in a
+
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
+
+let binary_parent t n =
+ if tree_is_first_child t.doc n
+ then tree_parent t.doc n
+ else tree_prev_sibling t.doc n
+
+let doc_ids t n = tree_doc_ids t.doc n
+
+let subtree_tags t tag = ();
+ fun n -> if n == nil then 0 else
+ tree_subtree_tags t.doc n tag
+
+let get_text t n =
+ let tid = tree_my_text t.doc n in
+ if tid == nulldoc then "" else
+ text_get_cached_text t.doc tid
+
+
+let dump_tree fmt tree =
+ let rec loop t n =
+ if t != nil then
+ let tag = (tree_tag_id tree.doc t ) in
+ let tagstr = Tag.to_string tag in
+ let tab = String.make n ' ' in
+
+ if tag == Tag.pcdata || tag == Tag.attribute_data
+ then
+ Format.fprintf fmt "%s<%s>%s</%s>\n"
+ tab tagstr (text_get_cached_text tree.doc (tree_my_text tree.doc t)) tagstr
+ else begin
+ Format.fprintf fmt "%s<%s>\n" tab tagstr;
+ loop (tree_first_child tree.doc t) (n+2);
+ Format.fprintf fmt "%s</%s>\n%!" tab tagstr;
+ end;
+ loop (tree_next_sibling tree.doc t) n
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
-
-let get_text t = match t.node with
- | Text(i,_) -> get_cached_text t.doc i
- | _ -> ""
+ loop root 0
+;;
+
+