| 1 |
abate |
70 |
(* Loading XML documents *)
|
| 2 |
|
|
|
| 3 |
|
|
(*TODO: close the file ! *)
|
| 4 |
|
|
|
| 5 |
|
|
open Pxp_yacc
|
| 6 |
|
|
open Pxp_lexer_types
|
| 7 |
|
|
open Pxp_types
|
| 8 |
|
|
open Value
|
| 9 |
abate |
233 |
open Ident
|
| 10 |
abate |
70 |
|
| 11 |
abate |
131 |
let is_ws s =
|
| 12 |
|
|
let rec check i =
|
| 13 |
|
|
(i < 0) ||
|
| 14 |
|
|
(match s.[i] with
|
| 15 |
|
|
| ' ' | '\t' | '\n' | '\r' -> check (i - 1)
|
| 16 |
|
|
| _ -> false) in
|
| 17 |
|
|
check (String.length s - 1)
|
| 18 |
|
|
|
| 19 |
|
|
|
| 20 |
abate |
70 |
let string s q =
|
| 21 |
abate |
131 |
String (0,String.length s,s,q)
|
| 22 |
abate |
70 |
|
| 23 |
abate |
188 |
let attrib att =
|
| 24 |
abate |
233 |
let att = List.map (fun (l,v) -> LabelPool.mk l, string v nil) att in
|
| 25 |
|
|
LabelMap.from_list (fun _ _ -> assert false) att
|
| 26 |
abate |
188 |
|
| 27 |
|
|
let elem tag att child =
|
| 28 |
abate |
222 |
Xml (Atom (Atoms.mk tag), Pair (Record (attrib att), child))
|
| 29 |
abate |
188 |
|
| 30 |
|
|
let load_xml_aux s =
|
| 31 |
abate |
70 |
let config = { default_config with
|
| 32 |
|
|
store_element_positions = false;
|
| 33 |
|
|
drop_ignorable_whitespace = true
|
| 34 |
|
|
}
|
| 35 |
|
|
in
|
| 36 |
|
|
let mgr = create_entity_manager config (from_file s) in
|
| 37 |
|
|
let next_event =
|
| 38 |
|
|
create_pull_parser config (`Entry_document[]) mgr in
|
| 39 |
|
|
let curr = ref E_end_of_stream in
|
| 40 |
|
|
let get () =
|
| 41 |
|
|
match next_event () with
|
| 42 |
abate |
83 |
| Some (E_error exn) -> failwith (Pxp_types.string_of_exn exn)
|
| 43 |
|
|
| Some E_end_of_stream -> failwith "Unexpected end of XML stream"
|
| 44 |
abate |
70 |
| Some x -> curr := x
|
| 45 |
|
|
| None -> () in
|
| 46 |
|
|
|
| 47 |
abate |
172 |
let txt = Buffer.create 1024 in
|
| 48 |
|
|
|
| 49 |
abate |
70 |
let rec parse_elt name att =
|
| 50 |
abate |
188 |
let elt = elem name att (parse_seq ()) in
|
| 51 |
abate |
70 |
(match !curr with
|
| 52 |
|
|
| E_end_tag (_,_) -> get ()
|
| 53 |
|
|
| _ -> failwith "Expect end_tag");
|
| 54 |
|
|
elt
|
| 55 |
|
|
|
| 56 |
abate |
172 |
|
| 57 |
|
|
and dump_txt q =
|
| 58 |
|
|
let data = Buffer.contents txt in
|
| 59 |
|
|
Buffer.clear txt;
|
| 60 |
|
|
if (is_ws data) then q () else string data (q ())
|
| 61 |
abate |
70 |
|
| 62 |
abate |
172 |
and parse_seq () =
|
| 63 |
abate |
70 |
match !curr with
|
| 64 |
abate |
172 |
| E_start_tag (name,att,_) ->
|
| 65 |
abate |
70 |
get ();
|
| 66 |
abate |
172 |
dump_txt (fun () ->
|
| 67 |
|
|
let e1 = parse_elt name att in
|
| 68 |
|
|
let rest = parse_seq () in
|
| 69 |
|
|
Pair (e1, rest)
|
| 70 |
|
|
)
|
| 71 |
abate |
70 |
| E_char_data data ->
|
| 72 |
abate |
172 |
get();
|
| 73 |
|
|
Buffer.add_string txt data;
|
| 74 |
|
|
parse_seq ()
|
| 75 |
abate |
70 |
| E_end_tag (_,_) ->
|
| 76 |
abate |
172 |
dump_txt (fun () -> nil)
|
| 77 |
abate |
70 |
| _ -> failwith "Expect start_tag, char_data, or end_tag"
|
| 78 |
|
|
|
| 79 |
|
|
and parse_doc () =
|
| 80 |
|
|
match !curr with
|
| 81 |
|
|
| E_start_tag (name,att,_) -> get (); parse_elt name att
|
| 82 |
|
|
| _ -> get (); parse_doc ()
|
| 83 |
|
|
in
|
| 84 |
|
|
get ();
|
| 85 |
abate |
73 |
parse_doc ()
|
| 86 |
|
|
|
| 87 |
abate |
91 |
|
| 88 |
abate |
188 |
let load_xml s =
|
| 89 |
abate |
126 |
Location.protect_op "load_xml";
|
| 90 |
abate |
188 |
try load_xml_aux s
|
| 91 |
abate |
91 |
with exn ->
|
| 92 |
|
|
raise
|
| 93 |
|
|
(Location.Generic (Pxp_types.string_of_exn exn))
|
| 94 |
|
|
|
| 95 |
abate |
188 |
|
| 96 |
|
|
let load_html s =
|
| 97 |
|
|
let rec val_of_doc q = function
|
| 98 |
|
|
| Nethtml.Data data ->
|
| 99 |
|
|
if (is_ws data) then q else string data q
|
| 100 |
|
|
| Nethtml.Element (tag, att, child) ->
|
| 101 |
|
|
Pair (elem tag att (val_of_docs child), q)
|
| 102 |
|
|
and val_of_docs = function
|
| 103 |
|
|
| [] -> nil
|
| 104 |
|
|
| h::t -> val_of_doc (val_of_docs t) h
|
| 105 |
|
|
in
|
| 106 |
|
|
|
| 107 |
|
|
Location.protect_op "load_xml";
|
| 108 |
|
|
let ic = open_in s in
|
| 109 |
|
|
let doc = Nethtml.parse_document
|
| 110 |
|
|
~dtd:Nethtml.relaxed_html40_dtd
|
| 111 |
|
|
(Lexing.from_channel ic) in
|
| 112 |
|
|
let doc = Nethtml.decode ~subst:(fun _ -> "???") doc in
|
| 113 |
|
|
close_in ic;
|
| 114 |
|
|
val_of_docs doc
|