2 * Copyright (C) 2003-2004:
3 * Stefano Zacchiroli <zack@cs.unibo.it>
4 * for the HELM Team http://helm.cs.unibo.it/
6 * This file is part of HELM, an Hypertextual, Electronic
7 * Library of Mathematics, developed at the Computer Science
8 * Department, University of Bologna, Italy.
10 * HELM is free software; you can redistribute it and/or
11 * modify it under the terms of the GNU General Public License
12 * as published by the Free Software Foundation; either version 2
13 * of the License, or (at your option) any later version.
15 * HELM is distributed in the hope that it will be useful,
16 * but WITHOUT ANY WARRANTY; without even the implied warranty of
17 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
18 * GNU General Public License for more details.
20 * You should have received a copy of the GNU General Public License
21 * along with HELM; if not, write to the Free Software
22 * Foundation, Inc., 59 Temple Place - Suite 330, Boston,
25 * For details, see the HELM World-Wide-Web page,
26 * http://helm.cs.unibo.it/
31 open Http_getter_common
33 open Http_getter_types
35 exception Not_implemented of string
36 exception UnexpectedGetterOutput
43 type logger_callback = HelmLogger.html_tag -> unit
45 let stdout_logger tag = print_string (HelmLogger.string_of_html_tag tag)
47 let not_implemented s = raise (Not_implemented ("Http_getter." ^ s))
49 let (index_line_sep_RE, index_sep_RE, trailing_types_RE,
50 heading_cic_RE, heading_theory_RE, heading_nuprl_RE,
51 heading_rdf_cic_RE, heading_rdf_theory_RE) =
52 (Pcre.regexp "[ \t]+", Pcre.regexp "\r\n|\r|\n",
53 Pcre.regexp "\\.types$",
54 Pcre.regexp "^cic:", Pcre.regexp "^theory:", Pcre.regexp "^nuprl:",
55 Pcre.regexp "^helm:rdf.*//cic:", Pcre.regexp "^helm:rdf.*//theory:")
57 (* global maps, shared by all threads *)
60 lazy (new Http_getter_map.map (Lazy.force Http_getter_env.cic_dbm))
62 lazy (new Http_getter_map.map (Lazy.force Http_getter_env.nuprl_dbm))
64 lazy (new Http_getter_map.map (Lazy.force Http_getter_env.rdf_dbm))
66 lazy (new Http_getter_map.map (Lazy.force Http_getter_env.xsl_dbm))
68 let uri_tree = ref None
71 | None -> assert false
73 let is_prefetch_on () =
74 match !uri_tree with None -> false | Some _ -> true
77 let path = Lazy.force Http_getter_env.dump_file in
78 Tree.save_to_disk path (deref_if_some uri_tree);
79 Http_getter_md5.create_hash [
80 (Lazy.force Http_getter_env.cic_dbm_real);
84 if not (Http_getter_md5.check_hash ()) then
87 uri_tree := Some (Tree.load_from_disk
88 (Lazy.force Http_getter_env.dump_file))
90 let sync_with_map () =
91 if not (Http_getter_md5.check_hash ()) then begin
92 let tree = ref (Some Tree.empty_tree) in
93 Http_getter_logger.log "Updating cic map dump...";
94 let t = Unix.time () in
95 (Lazy.force cic_map)#iter
97 tree := Some (Tree.add_uri k (deref_if_some tree)));
99 Http_getter_logger.log
100 (sprintf "done in %.0f sec" (Unix.time () -. t));
103 Http_getter_logger.log "Cic map dump is up to date!";
104 load_tree () (* XXX TASSI: race condition here *)
107 let maps = [ cic_map; nuprl_map; rdf_map; xsl_map ]
108 let close_maps () = List.iter (fun m -> (Lazy.force m) # close) maps
109 let clear_maps () = List.iter (fun m -> (Lazy.force m) # clear) maps
111 List.iter (fun m -> (Lazy.force m) # sync) maps;
114 let map_of_uri = function
115 | uri when is_cic_uri uri -> Lazy.force cic_map
116 | uri when is_nuprl_uri uri -> Lazy.force nuprl_map
117 | uri when is_rdf_uri uri -> Lazy.force rdf_map
118 | uri when is_xsl_uri uri -> Lazy.force xsl_map
119 | uri -> raise (Unresolvable_URI uri)
121 let update_from_server logger server_url = (* use global maps *)
122 Http_getter_logger.log ("Updating information from " ^ server_url);
123 let xml_url_of_uri = function
124 (* TODO missing sanity checks on server_url, e.g. it can contains $1 *)
125 | uri when (Pcre.pmatch ~rex:heading_cic_RE uri) ->
126 Pcre.replace ~rex:heading_cic_RE ~templ:server_url uri
127 | uri when (Pcre.pmatch ~rex:heading_theory_RE uri) ->
128 Pcre.replace ~rex:heading_theory_RE ~templ:server_url uri
129 | uri when (Pcre.pmatch ~rex:heading_nuprl_RE uri) ->
130 Pcre.replace ~rex:heading_nuprl_RE ~templ:server_url uri
131 | uri -> raise (Invalid_URI uri)
133 let rdf_url_of_uri = function (* TODO as above *)
134 | uri when (Pcre.pmatch ~rex:heading_rdf_cic_RE uri) ->
135 Pcre.replace ~rex:heading_rdf_cic_RE ~templ:server_url uri
136 | uri when (Pcre.pmatch ~rex:heading_rdf_theory_RE uri) ->
137 Pcre.replace ~rex:heading_rdf_theory_RE ~templ:server_url uri
138 | uri -> raise (Invalid_URI uri)
140 logger (`T ("Processing server: " ^ server_url));
142 let (xml_index, rdf_index, xsl_index) =
143 (* TODO keeps index in memory, is better to keep them on temp files? *)
144 (http_get (server_url ^ "/" ^ (Lazy.force Http_getter_env.xml_index)),
145 http_get (server_url ^ "/" ^ (Lazy.force Http_getter_env.rdf_index)),
146 http_get (server_url ^ "/" ^ (Lazy.force Http_getter_env.xsl_index)))
148 if (xml_index = None && rdf_index = None && xsl_index = None) then
149 Http_getter_logger.log (sprintf "Warning: useless server %s" server_url);
150 (match xml_index with
152 logger (`T "- Updating XML db ...");
156 | l when is_blank_line l -> () (* skip blank and commented lines *)
159 (match Pcre.split ~rex:index_line_sep_RE l with
161 assert (is_cic_uri uri || is_nuprl_uri uri) ;
162 (map_of_uri uri)#replace
163 uri ((xml_url_of_uri uri) ^ ".xml.gz")
165 assert (is_cic_uri uri || is_nuprl_uri uri) ;
166 (map_of_uri uri)#replace
167 uri ((xml_url_of_uri uri) ^ ".xml")
169 logger (`T ("Ignoring invalid line: '" ^ l));
171 with Invalid_URI uri ->
172 logger (`T ("Ignoring invalid XML URI: '" ^ l));
174 (Pcre.split ~rex:index_sep_RE xml_index); (* xml_index lines *)
175 logger (`T "All done");
178 (match rdf_index with
180 logger (`T "- Updating RDF db ...");
185 (match Pcre.split ~rex:index_line_sep_RE l with
187 (Lazy.force rdf_map) # replace uri
188 ((rdf_url_of_uri uri) ^ ".xml.gz")
190 (Lazy.force rdf_map) # replace uri
191 ((rdf_url_of_uri uri) ^ ".xml")
193 logger (`T ("Ignoring invalid line: '" ^ l));
195 with Invalid_URI uri ->
196 logger (`T ("Ignoring invalid RDF URI: '" ^ l));
198 (Pcre.split ~rex:index_sep_RE rdf_index); (* rdf_index lines *)
199 logger (`T "All done");
202 (match xsl_index with
204 logger (`T "- Updating XSLT db ...");
207 (fun l -> (Lazy.force xsl_map) # replace l (server_url ^ "/" ^ l))
208 (Pcre.split ~rex:index_sep_RE xsl_index);
209 logger (`T "All done");
212 Http_getter_logger.log "done with this server"
214 let update_from_all_servers logger () = (* use global maps *)
217 (update_from_server logger)
218 (* reverse order: 1st server is the most important one *)
219 (List.map snd (List.rev (Http_getter_env.servers ())));
222 let update_from_one_server ?(logger = fun _ -> ()) server_url =
223 update_from_server logger server_url
225 let temp_file_of_uri uri =
226 let flat_string s s' c =
227 let cs = String.copy s in
228 for i = 0 to (String.length s) - 1 do
229 if String.contains s' s.[i] then cs.[i] <- c
233 let user = try Unix.getlogin () with _ -> "" in
234 Filename.open_temp_file (user ^ flat_string uri ".-=:;!?/&" '_') ""
236 (* should we use a remote getter or not *)
239 Helm_registry.get "getter.mode" = "remote"
240 with Helm_registry.Key_not_found _ -> false
241 let getter_url () = Helm_registry.get "getter.url"
243 (* Remote interface: getter methods implemented using a remote getter *)
246 let getxslt_remote ~patch_dtd uri = not_implemented "getxslt_remote"
247 let getdtd_remote ~patch_dtd uri = not_implemented "getdtd_remote"
248 let clean_cache_remote () = not_implemented "clean_cache_remote"
249 let list_servers_remote () = not_implemented "list_servers_remote"
250 let add_server_remote ~logger ~position name =
251 not_implemented "add_server_remote"
252 let remove_server_remote ~logger position =
253 not_implemented "remove_server_remote"
254 let getalluris_remote () = not_implemented "getalluris_remote"
255 let getallrdfuris_remote () = not_implemented "getallrdfuris_remote"
256 let ls_remote lsuri = not_implemented "ls_remote"
259 let resolve_remote uri =
260 (* deliver resolve request to http_getter *)
261 let doc = ClientHTTP.get (sprintf "%sresolve?uri=%s" (getter_url ()) uri) in
262 let res = ref Unknown in
263 let start_element tag attrs =
267 res := Resolved (List.assoc "value" attrs)
268 with Not_found -> ())
269 | "unresolvable" -> res := Exception (Unresolvable_URI uri)
270 | "not_found" -> res := Exception (Key_not_found uri)
274 XmlPushParser.default_callbacks with
275 XmlPushParser.start_element = Some start_element
277 let xml_parser = XmlPushParser.create_parser callbacks in
278 XmlPushParser.parse xml_parser (`String doc);
279 XmlPushParser.final xml_parser;
281 | Unknown -> raise UnexpectedGetterOutput
282 | Exception e -> raise e
283 | Resolved url -> url
285 let register_remote ~uri ~url =
286 ClientHTTP.send (sprintf "%sregister?uri=%s&url=%s" (getter_url ()) uri url)
288 let unregister_remote uri =
289 ClientHTTP.send (sprintf "%sunregister?uri=%s" (getter_url ()) uri)
291 let update_remote logger () =
292 let answer = ClientHTTP.get (getter_url () ^ "update") in
296 let getxml_remote ~format ~patch_dtd uri =
298 sprintf "%sgetxml?uri=%s&format=%s&patch_dtd=%s"
300 (match format with `Normal -> "normal" | `Gzipped -> "gz")
301 (match patch_dtd with true -> "yes" | false -> "no")
303 ClientHTTP.get_and_save_to_tmp uri
307 let help () = Http_getter_const.usage_string (Http_getter_env.env_to_string ())
313 (map_of_uri uri)#resolve uri
315 let register ~uri ~url =
317 register_remote ~uri ~url
320 (map_of_uri uri)#add uri url;
321 if is_prefetch_on () then
322 uri_tree := Some (Tree.add_uri uri (deref_if_some uri_tree))
327 unregister_remote uri
331 (map_of_uri uri)#remove uri;
332 if is_prefetch_on () then
333 uri_tree := Some (Tree.remove_uri uri (deref_if_some uri_tree))
335 with Key_not_found _ -> ()
337 let update ?(logger = fun _ -> ()) () =
339 update_remote logger ()
341 update_from_all_servers logger ()
343 let getxml ?(format = `Gzipped) ?(patch_dtd = false) uri =
345 getxml_remote ~format ~patch_dtd uri
347 Http_getter_logger.log ~level:2 ("getxml: " ^ uri);
348 let url = resolve uri in
349 Http_getter_logger.log ~level:2 ("resolved_uri: " ^ url) ;
350 let (fname, outchan) = temp_file_of_uri uri in
351 Http_getter_cache.respond_xml ~via_http:false ~enc:format ~patch:patch_dtd
357 let getxslt ?(patch_dtd = true) uri =
359 getxslt_remote ~patch_dtd uri
362 let url = resolve uri in
363 let (fname, outchan) = temp_file_of_uri uri in
364 Http_getter_cache.respond_xsl ~via_http:false ~url ~patch:patch_dtd outchan;
369 let getdtd ?(patch_dtd = true) uri =
371 getdtd_remote ~patch_dtd uri
373 let url = Lazy.force Http_getter_env.dtd_dir ^ "/" ^ uri in
374 let (fname, outchan) = temp_file_of_uri uri in
375 Http_getter_cache.respond_dtd ~via_http:false ~url ~patch:patch_dtd outchan;
382 clean_cache_remote ()
384 Http_getter_cache.clean ()
386 let list_servers () =
388 list_servers_remote ()
390 Http_getter_env.servers ()
392 let add_server ?(logger = fun _ -> ()) ?(position = 0) name =
394 add_server_remote ~logger ~position name
396 if position = 0 then begin
397 Http_getter_env.add_server ~position:0 name;
398 update_from_one_server ~logger name (* quick update (new server only) *)
399 end else if position > 0 then begin
400 Http_getter_env.add_server ~position name;
402 end else (* already checked by parse_position *)
406 let has_server position = List.mem_assoc position (Http_getter_env.servers ())
408 let remove_server ?(logger = fun _ -> ()) position =
410 remove_server_remote ~logger ()
414 List.assoc position (Http_getter_env.servers ())
416 raise (Invalid_argument (sprintf "no server with position %d" position))
418 Http_getter_env.remove_server position;
422 let return_uris map filter =
424 map#iter (fun uri _ -> if filter uri then uris := uri :: !uris);
432 (Pcre.pmatch ~rex:heading_cic_RE uri)
433 (* && not (Pcre.pmatch ~rex:trailing_types_RE uri) *)
435 return_uris (Lazy.force cic_map) filter
437 let getallrdfuris classs =
439 getallrdfuris_remote ()
442 let base = "^helm:rdf:www\\.cs\\.unibo\\.it/helm/rdf/" in
444 | `Forward -> (fun uri -> Pcre.pmatch ~pat:(base ^ "forward") uri)
445 | `Backward -> (fun uri -> Pcre.pmatch ~pat:(base ^ "backward") uri)
447 return_uris (Lazy.force rdf_map) filter
450 let (++) (oldann, oldtypes, oldbody, oldtree)
451 (newann, newtypes, newbody, newtree) =
452 ((if newann > oldann then newann else oldann),
453 (if newtypes > oldtypes then newtypes else oldtypes),
454 (if newbody > oldbody then newbody else oldbody),
455 (if newtree > oldtree then newtree else oldtree))
457 let (types_RE, types_ann_RE, body_RE, body_ann_RE,
458 proof_tree_RE, proof_tree_ann_RE, trailing_slash_RE, theory_RE) =
459 (Pcre.regexp "\\.types$", Pcre.regexp "\\.types\\.ann$",
460 Pcre.regexp "\\.body$", Pcre.regexp "\\.body\\.ann$",
461 Pcre.regexp "\\.proof_tree$", Pcre.regexp "\\.proof_tree\\.ann$",
462 Pcre.regexp "/$", Pcre.regexp "\\.theory$")
466 "^([^.]*\\.[^.]*)((\\.body)|(\\.proof_tree)|(\\.types))?(\\.ann)?$"
468 let (slash_RE, til_slash_RE, no_slashes_RE) =
469 (Pcre.regexp "/", Pcre.regexp "^.*/", Pcre.regexp "^[^/]*$")
470 let fix_regexp_RE = Pcre.regexp ("^" ^ (Pcre.quote "(cic|theory)"))
476 let looking_for_dir = Pcre.pmatch ~rex:trailing_slash_RE regexp in
477 let pat = Pcre.replace ~rex:trailing_slash_RE ("^" ^ regexp) in
478 let (dir_RE, dir_local_RE, obj_RE, first_comp_RE) =
479 Pcre.regexp (pat ^ "/"), Pcre.regexp "[^/]+/[^/]*",
480 Pcre.regexp (pat ^ "(\\.|$)"), Pcre.regexp "/.*"
482 let exists_theory regexp =
484 Pcre.replace ~rex:fix_regexp_RE ~templ:"theory" regexp ^ "index.theory"
487 ignore (resolve theory);
489 with Key_not_found _ -> false
491 let toplevel_theory =
492 match List.rev (Pcre.split ~rex:slash_RE pat) with
493 | dir :: _ -> Some (dir ^ ".theory")
496 let dirs = ref StringSet.empty in
497 let objs = Hashtbl.create 17 in
499 dirs := StringSet.add (List.hd (Pcre.split ~rex:slash_RE d)) !dirs
502 let basepart = Pcre.replace ~rex:basepart_RE ~templ:"$1" o in
503 let no_flags = false, No, No, No in
506 Hashtbl.find objs basepart
507 with Not_found -> (* no ann, no types, no body, no proof tree *)
512 | s when Pcre.pmatch ~rex:types_RE s -> (false, Yes, No, No)
513 | s when Pcre.pmatch ~rex:types_ann_RE s -> (true, Ann, No, No)
514 | s when Pcre.pmatch ~rex:body_RE s -> (false, No, Yes, No)
515 | s when Pcre.pmatch ~rex:body_ann_RE s -> (true, No, Ann, No)
516 | s when Pcre.pmatch ~rex:proof_tree_RE s -> (false, No, No, Yes)
517 | s when Pcre.pmatch ~rex:proof_tree_ann_RE s -> (true, No, No, Ann)
520 Hashtbl.replace objs basepart (oldflags ++ newflags)
522 (* Variables used in backward compatibility code to map
523 theory:/path/t.theory into theory:/path/t/index.theory
524 when cic:/path/t/ exists *)
525 let the_candidate_for_remapping =
526 (* CSC: Here I am making a strong assumption: the pattern
527 can be only of the form [^:]*:/path where path is
528 NOT a regular expression *)
529 "theory:" ^ Pcre.replace ~rex:(Pcre.regexp "[^:]*:") pat
531 let index_not_generated_yet = ref true in
532 let valid_candidates = ref [] in
533 let candidates_found = ref [] in
535 (*(Lazy.force cic_map) # iter*)
537 (* depending on prefetch *)
538 let if_prefetch if_is if_not =
539 if is_prefetch_on() then if_is else if_not
542 let iter_on_right_thing = if_prefetch
543 (fun f -> List.iter (fun k -> f k "")
544 (Tree.ls_path regexp (deref_if_some uri_tree)))
545 (fun f -> (Lazy.force cic_map) # iter f)
547 let calculate_localpart = if_prefetch
549 (fun uri -> Pcre.replace ~rex:dir_RE uri)
551 let check_if_x_RE = if_prefetch
552 (fun x_RE uri -> true)
553 (fun x_RE uri -> Pcre.pmatch ~rex:x_RE uri)
555 let check_if_dir_RE = check_if_x_RE dir_RE in
556 let check_if_obj_RE = check_if_x_RE obj_RE in
560 (* we work in two ways:
561 * 1 iteration on the whole map
564 * Since in the first case 'key' is a complete uri, while
565 * in the second case it is only the subtree rooted in the
566 * query regex, we must relay only on the localpath.
569 * query::= cic:/aaa/bbb/
571 * answer1 ::= the whole map
573 * aswer2 ::= [ "ccc/"; "c1.something"] where
574 * cic:/aaa/bbb/ccc/ and cic:/aaa/bbb/c1.something
575 * are the (partials) uri that matched query
577 * after the localpath extracion we have more uris in the first case,
578 * but at least the are all rooted in the same node.
580 * the Tree.get_frontier may be changed to return the same stuff as
581 * the map iteration+localpath extraction does, but I hope it is not
585 | uri when looking_for_dir && check_if_dir_RE uri ->
587 let localpart = calculate_localpart uri in
588 if Pcre.pmatch ~rex:no_slashes_RE localpart then
590 (* Backward compatibility code to map
591 theory:/path/t.theory into theory:/path/t/index.theory
592 when cic:/path/t/ exists *)
593 if Pcre.pmatch ~rex:theory_RE localpart then
594 candidates_found := localpart :: !candidates_found
600 store_dir localpart ;
601 if Pcre.pmatch localpart ~rex:dir_local_RE then
604 Pcre.replace ~rex:first_comp_RE localpart ^ ".theory"
606 if not (List.mem valid !valid_candidates) then
607 valid_candidates := valid::!valid_candidates
610 | uri when (not looking_for_dir) && check_if_obj_RE uri ->
612 store_obj (Pcre.replace ~rex:til_slash_RE uri)
616 if !index_not_generated_yet &&
617 Pcre.pmatch ~rex:orig_theory_RE uri
619 (index_not_generated_yet := false ;
620 store_obj "index.theory"));
622 if exists_theory regexp then store_obj "index.theory";
625 if not (List.mem localpart !valid_candidates) then
627 ) !candidates_found ;
628 let ls_items = ref [] in
629 StringSet.iter (fun dir -> ls_items := Ls_section dir :: !ls_items) !dirs;
630 Http_getter_misc.hashtbl_sorted_iter
631 (fun uri (annflag, typesflag, bodyflag, treeflag) ->
634 uri = uri; ann = annflag;
635 types = typesflag; body = bodyflag; proof_tree = treeflag
641 (* Shorthands from now on *)
643 let getxml' uri = getxml (UriManager.string_of_uri uri)
644 let resolve' uri = resolve (UriManager.string_of_uri uri)
645 let register' uri url = register ~uri:(UriManager.string_of_uri uri) ~url
646 let unregister' uri = unregister (UriManager.string_of_uri uri)
648 let sync_dump_file () =
649 if is_prefetch_on () then
653 Http_getter_logger.set_log_level
654 (Helm_registry.get_opt_default Helm_registry.int ~default:1
656 Http_getter_logger.set_log_file
657 (Helm_registry.get_opt Helm_registry.string "getter.log_file");
658 Http_getter_env.reload ();
659 let is_prefetch_set =
660 Helm_registry.get_opt_default Helm_registry.bool ~default:false
663 if is_prefetch_set then
664 (* ignore (Thread.create sync_with_map ()) *)