citadel/modules/rssclient/serv_rssclient.c

   1 /*
   2  * Bring external RSS and/or Atom feeds into rooms.  This module implements a
   3  * very loose parser that scrapes both kinds of feeds and is not picky about
   4  * the standards compliance of the source data.
   5  *
   6  * Copyright (c) 2007-2018 by the citadel.org team
   7  *
   8  * This program is open source software; you can redistribute it and/or modify
   9  * it under the terms of the GNU General Public License version 3.
  10  *
  11  * This program is distributed in the hope that it will be useful,
  12  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  13  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.See the
  14  * GNU General Public License for more details.
  15  */
  16
  17 #include <stdlib.h>
  18 #include <unistd.h>
  19 #include <stdio.h>
  20
  21 #if TIME_WITH_SYS_TIME
  22 # include <sys/time.h>
  23 # include <time.h>
  24 #else
  25 # if HAVE_SYS_TIME_H
  26 #include <sys/time.h>
  27 # else
  28 #include <time.h>
  29 # endif
  30 #endif
  31
  32 #include <ctype.h>
  33 #include <string.h>
  34 #include <errno.h>
  35 #include <sys/types.h>
  36 #include <sys/stat.h>
  37 #include <expat.h>
  38 #include <curl/curl.h>
  39 #include <libcitadel.h>
  40 #include "citadel.h"
  41 #include "server.h"
  42 #include "citserver.h"
  43 #include "support.h"
  44 #include "config.h"
  45 #include "threads.h"
  46 #include "ctdl_module.h"
  47 #include "msgbase.h"
  48 #include "parsedate.h"
  49 #include "database.h"
  50 #include "citadel_dirs.h"
  51 #include "md5.h"
  52 #include "context.h"
  53 #include "internet_addressing.h"
  54
  55 struct rssroom {
  56         struct rssroom *next;
  57         char *room;
  58 };
  59
  60 struct rssurl {
  61         struct rssurl *next;
  62         char *url;
  63         struct rssroom *rooms;
  64 };
  65
  66 struct rssparser {
  67         StrBuf *CData;
  68         struct CtdlMessage *msg;
  69         char *link;
  70         char *description;
  71         char *item_id;
  72         struct rssroom *rooms;
  73 };
  74
  75 time_t last_run = 0L;
  76 struct rssurl *rsstodo = NULL;
  77
  78
  79 // This handler is called whenever an XML tag opens.
  80 //
  81 void rss_start_element(void *data, const char *el, const char **attribute)
  82 {
  83         struct rssparser *r = (struct rssparser *)data;
  84         int i;
  85
  86         if (
  87                 (!strcasecmp(el, "entry"))
  88                 || (!strcasecmp(el, "item"))
  89         ) {
  90                 // this is the start of a new item(rss) or entry(atom)
  91                 if (r->msg != NULL) {
  92                         CM_Free(r->msg);
  93                         r->msg = NULL;
  94                 }
  95                 r->msg = malloc(sizeof(struct CtdlMessage));
  96                 memset(r->msg, 0, sizeof(struct CtdlMessage));
  97                 r->msg->cm_magic = CTDLMESSAGE_MAGIC;
  98                 r->msg->cm_anon_type = MES_NORMAL;
  99                 r->msg->cm_format_type = FMT_RFC822;
 100         }
 101
 102         else if (!strcasecmp(el, "link")) {                     // atom feeds have the link as an attribute
 103                 for(i = 0; attribute[i]; i += 2) {
 104                         if (!strcasecmp(attribute[i], "href")) {
 105                                 if (r->link != NULL) {
 106                                         free(r->link);
 107                                         r->link = NULL;
 108                                 }
 109                                 r->link = strdup(attribute[i+1]);
 110                                 striplt(r->link);
 111                         }
 112                 }
 113         }
 114 }
 115
 116
 117 // This handler is called whenever an XML tag closes.
 118 //
 119 void rss_end_element(void *data, const char *el)
 120 {
 121         struct rssparser *r = (struct rssparser *)data;
 122
 123         if (                                                    // end of a new item(rss) or entry(atom)
 124                 (!strcasecmp(el, "entry"))
 125                 || (!strcasecmp(el, "item"))
 126         ) {
 127
 128                 if (r->msg != NULL) {                           // Save the message to the rooms
 129
 130                         // use the link as an item id if nothing else is available
 131                         if ((r->item_id == NULL) && (r->link != NULL)) {
 132                                 r->item_id = strdup(r->link);
 133                         }
 134
 135                         // check the use table
 136                         StrBuf *u = NewStrBuf();
 137                         StrBufAppendPrintf(u, "rss/%s", r->item_id);
 138                         int already_seen = CheckIfAlreadySeen(u);
 139                         FreeStrBuf(&u);
 140
 141                         if (already_seen == 0) {
 142
 143                                 // Compose the message text
 144                                 StrBuf *TheMessage = NewStrBuf();
 145                                 StrBufAppendPrintf(TheMessage,
 146                                         "Content-type: text/html\n\n"
 147                                         "\n\n"
 148                                         "<html><head></head><body>"
 149                                 );
 150
 151                                 if (r->description != NULL) {
 152                                         StrBufAppendPrintf(TheMessage, "%s<br><br>\r\n", r->description);
 153                                         free(r->description);
 154                                         r->description = NULL;
 155                                 }
 156
 157                                 if (r->link != NULL) {
 158                                         StrBufAppendPrintf(TheMessage, "<a href=\"%s\">%s</a>\r\n", r->link, r->link);
 159                                         free(r->link);
 160                                         r->link = NULL;
 161                                 }
 162
 163                                 StrBufAppendPrintf(TheMessage, "</body></html>\r\n");
 164                                 CM_SetField(r->msg, eMesageText, ChrPtr(TheMessage), StrLength(TheMessage));
 165                                 FreeStrBuf(&TheMessage);
 166
 167                                 if (CM_IsEmpty(r->msg, eAuthor)) {
 168                                         CM_SetField(r->msg, eAuthor, HKEY("rss"));
 169                                 }
 170
 171                                 if (CM_IsEmpty(r->msg, eTimestamp)) {
 172                                         CM_SetFieldLONG(r->msg, eTimestamp, time(NULL));
 173                                 }
 174
 175                                 // Save it to the room(s)
 176                                 struct rssroom *rr = NULL;
 177                                 long msgnum = (-1);
 178                                 for (rr=r->rooms; rr!=NULL; rr=rr->next) {
 179                                         if (rr == r->rooms) {
 180                                                 msgnum = CtdlSubmitMsg(r->msg, NULL, rr->room, 0);
 181                                         }
 182                                         else {
 183                                                 CtdlSaveMsgPointerInRoom(rr->room, msgnum, 0, NULL);
 184                                         }
 185                                         syslog(LOG_DEBUG, "rssclient: saved message %ld to %s", msgnum, rr->room);
 186                                 }
 187                         }
 188                         else {
 189                                 syslog(LOG_DEBUG, "rssclient: already seen %s", r->item_id);
 190                         }
 191
 192                         CM_Free(r->msg);
 193                         r->msg = NULL;
 194                 }
 195
 196                 if (r->item_id != NULL) {
 197                         free(r->item_id);
 198                         r->item_id = NULL;
 199                 }
 200         }
 201
 202         else if (!strcasecmp(el, "title")) {                    // item subject (rss and atom)
 203                 if ((r->msg != NULL) && (CM_IsEmpty(r->msg, eMsgSubject))) {
 204                         CM_SetField(r->msg, eMsgSubject, ChrPtr(r->CData), StrLength(r->CData));
 205                         striplt(r->msg->cm_fields[eMsgSubject]);
 206                 }
 207         }
 208
 209         else if (!strcasecmp(el, "creator")) {                  // <creator> can be used if <author> is not present
 210                 if ((r->msg != NULL) && (CM_IsEmpty(r->msg, eAuthor))) {
 211                         CM_SetField(r->msg, eAuthor, ChrPtr(r->CData), StrLength(r->CData));
 212                         striplt(r->msg->cm_fields[eAuthor]);
 213                 }
 214         }
 215
 216         else if (!strcasecmp(el, "author")) {                   // <author> supercedes <creator> if both are present
 217                 if (r->msg != NULL) {
 218                         CM_SetField(r->msg, eAuthor, ChrPtr(r->CData), StrLength(r->CData));    // CM_SetField will free() the previous value
 219                         striplt(r->msg->cm_fields[eAuthor]);
 220                 }
 221         }
 222
 223         else if (!strcasecmp(el, "pubdate")) {                  // date/time stamp (rss) Sat, 25 Feb 2017 14:28:01 EST
 224                 if ((r->msg)&&(r->msg->cm_fields[eTimestamp]==NULL)) {
 225                         CM_SetFieldLONG(r->msg, eTimestamp, parsedate(ChrPtr(r->CData)));
 226                 }
 227         }
 228
 229         else if (!strcasecmp(el, "updated")) {                  // date/time stamp (atom) 2003-12-13T18:30:02Z
 230                 if ((r->msg)&&(r->msg->cm_fields[eTimestamp]==NULL)) {
 231                         struct tm t;
 232                         char zulu;
 233                         memset(&t, 0, sizeof t);
 234                         sscanf(ChrPtr(r->CData), "%d-%d-%dT%d:%d:%d%c", &t.tm_year, &t.tm_mon, &t.tm_mday, &t.tm_hour, &t.tm_min, &t.tm_sec, &zulu);
 235                         t.tm_year -= 1900;
 236                         t.tm_mon -= 1;
 237                         CM_SetFieldLONG(r->msg, eTimestamp, mktime(&t));
 238                 }
 239         }
 240
 241         else if (!strcasecmp(el, "link")) {                     // link to story (rss)
 242                 if (r->link != NULL) {
 243                         free(r->link);
 244                         r->link = NULL;
 245                 }
 246                 r->link = strdup(ChrPtr(r->CData));
 247                 striplt(r->link);
 248         }
 249
 250         else if (
 251                 (!strcasecmp(el, "guid"))                       // unique item id (rss)
 252                 || (!strcasecmp(el, "id"))                      // unique item id (atom)
 253         ) {
 254                 if (r->item_id != NULL) {
 255                         free(r->item_id);
 256                         r->item_id = NULL;
 257                 }
 258                 r->item_id = strdup(ChrPtr(r->CData));
 259                 striplt(r->item_id);
 260         }
 261
 262         else if (
 263                 (!strcasecmp(el, "description"))                // message text (rss)
 264                 || (!strcasecmp(el, "summary"))                 // message text (atom)
 265                 || (!strcasecmp(el, "content"))                 // message text (atom)
 266         ) {
 267                 if (r->description != NULL) {
 268                         free(r->description);
 269                         r->description = NULL;
 270                 }
 271                 r->description = strdup(ChrPtr(r->CData));
 272                 striplt(r->description);
 273         }
 274
 275         if (r->CData != NULL) {
 276                 FreeStrBuf(&r->CData);
 277                 r->CData = NULL;
 278         }
 279 }
 280
 281
 282 // This handler is called whenever data appears between opening and closing tags.
 283 //
 284 void rss_handle_data(void *data, const char *content, int length)
 285 {
 286         struct rssparser *r = (struct rssparser *)data;
 287
 288         if (r->CData == NULL) {
 289                 r->CData = NewStrBuf();
 290         }
 291
 292         StrBufAppendBufPlain(r->CData, content, length, 0);
 293 }
 294
 295
 296 // Feed has been downloaded, now parse it.
 297 //
 298 void rss_parse_feed(StrBuf *Feed, struct rssroom *rooms)
 299 {
 300         struct rssparser r;
 301
 302         memset(&r, 0, sizeof r);
 303         r.rooms = rooms;
 304         XML_Parser p = XML_ParserCreate("UTF-8");
 305         XML_SetElementHandler(p, rss_start_element, rss_end_element);
 306         XML_SetCharacterDataHandler(p, rss_handle_data);
 307         XML_SetUserData(p, (void *)&r);
 308         XML_Parse(p, ChrPtr(Feed), StrLength(Feed), XML_TRUE);
 309         XML_ParserFree(p);
 310 }
 311
 312
 313 // Add a feed/room pair into the todo list
 314 //
 315 void rssclient_push_todo(char *rssurl, char *roomname)
 316 {
 317         struct rssurl *r = NULL;
 318         struct rssurl *thisone = NULL;
 319         struct rssroom *newroom = NULL;
 320
 321         syslog(LOG_DEBUG, "rssclient: will fetch %s to %s", rssurl, roomname);
 322
 323         for (r=rsstodo; r!=NULL; r=r->next) {
 324                 if (!strcasecmp(r->url, rssurl)) {
 325                         thisone = r;
 326                 }
 327         }
 328         if (thisone == NULL) {
 329                 thisone = malloc(sizeof(struct rssurl));
 330                 thisone->url = strdup(rssurl);
 331                 thisone->rooms = NULL;
 332                 thisone->next = rsstodo;
 333                 rsstodo = thisone;
 334         }
 335
 336         newroom = malloc(sizeof(struct rssroom));
 337         newroom->room = strdup(roomname);
 338         newroom->next = thisone->rooms;
 339         thisone->rooms = newroom;
 340 }
 341
 342
 343 // pull one feed (possibly multiple rooms)
 344 //
 345 void rss_pull_one_feed(struct rssurl *url)
 346 {
 347         CURL *curl;
 348         CURLcode res;
 349         StrBuf *Downloaded = NULL;
 350
 351         syslog(LOG_DEBUG, "rssclient: fetching %s", url->url);
 352
 353         curl = curl_easy_init();
 354         if (!curl) {
 355                 return;
 356         }
 357
 358         Downloaded = NewStrBuf();
 359
 360         curl_easy_setopt(curl, CURLOPT_URL, url->url);
 361         curl_easy_setopt(curl, CURLOPT_SSL_VERIFYPEER, 0L);
 362         curl_easy_setopt(curl, CURLOPT_SSL_VERIFYHOST, 0L);
 363         curl_easy_setopt(curl, CURLOPT_FOLLOWLOCATION, 1L);                     // Follow redirects
 364         curl_easy_setopt(curl, CURLOPT_WRITEFUNCTION, CurlFillStrBuf_callback); // What to do with downloaded data
 365         curl_easy_setopt(curl, CURLOPT_WRITEDATA, Downloaded);                  // Give it our StrBuf to work with
 366         curl_easy_setopt(curl, CURLOPT_TIMEOUT, 20L);                           // Time out after 20 seconds
 367         res = curl_easy_perform(curl);                                          // Perform the request
 368         if (res != CURLE_OK) {
 369                 syslog(LOG_WARNING, "rssclient: failed to load feed: %s", curl_easy_strerror(res));
 370         }
 371         curl_easy_cleanup(curl);
 372
 373         rss_parse_feed(Downloaded, url->rooms);                                 // parse the feed
 374         FreeStrBuf(&Downloaded);                                                // free the downloaded feed data
 375 }
 376
 377
 378 // We have a list, now download the feeds
 379 //
 380 void rss_pull_feeds(void)
 381 {
 382         struct rssurl *r;
 383         struct rssroom *rr;
 384
 385         while (rsstodo != NULL) {
 386                 rss_pull_one_feed(rsstodo);
 387                 r = rsstodo;
 388                 rsstodo = rsstodo->next;
 389                 while (r->rooms != NULL) {
 390                         rr = r->rooms;
 391                         r->rooms = r->rooms->next;
 392                         free(rr->room);
 393                         free(rr);
 394                 }
 395                 free(r->url);
 396                 free(r);
 397         }
 398 }
 399
 400
 401 // Scan a room's netconfig looking for RSS feed parsing requests
 402 //
 403 void rssclient_scan_room(struct ctdlroom *qrbuf, void *data)
 404 {
 405         char *serialized_config = NULL;
 406         int num_configs = 0;
 407         char cfgline[SIZ];
 408         int i = 0;
 409
 410         serialized_config = LoadRoomNetConfigFile(qrbuf->QRnumber);
 411         if (!serialized_config) {
 412                 return;
 413         }
 414
 415         num_configs = num_tokens(serialized_config, '\n');
 416         for (i=0; i<num_configs; ++i) {
 417                 extract_token(cfgline, serialized_config, i, '\n', sizeof cfgline);
 418                 if (!strncasecmp(cfgline, HKEY("rssclient|"))) {
 419                         strcpy(cfgline, &cfgline[10]);
 420                         char *vbar = strchr(cfgline, '|');
 421                         if (vbar != NULL) {
 422                                 *vbar = 0;
 423                         }
 424                         rssclient_push_todo(cfgline, qrbuf->QRname);
 425                 }
 426         }
 427
 428         free(serialized_config);
 429 }
 430
 431
 432 /*
 433  * Scan for rooms that have RSS client requests configured
 434  */
 435 void rssclient_scan(void) {
 436         time_t now = time(NULL);
 437
 438         /* Run no more than once every 15 minutes. */
 439         if ((now - last_run) < 900) {
 440                 syslog(LOG_DEBUG,
 441                         "rssclient: polling interval not yet reached; last run was %ldm%lds ago",
 442                         ((now - last_run) / 60),
 443                         ((now - last_run) % 60)
 444                 );
 445                 return;
 446         }
 447
 448         syslog(LOG_DEBUG, "rssclient: started");
 449         CtdlForEachRoom(rssclient_scan_room, NULL);
 450         rss_pull_feeds();
 451         syslog(LOG_DEBUG, "rssclient: ended");
 452         last_run = time(NULL);
 453         return;
 454 }
 455
 456
 457 CTDL_MODULE_INIT(rssclient)
 458 {
 459         if (!threading)
 460         {
 461                 syslog(LOG_INFO, "rssclient: using %s", curl_version());
 462                 CtdlRegisterSessionHook(rssclient_scan, EVT_TIMER, PRIO_AGGR + 300);
 463         }
 464         return "rssclient";
 465 }
 466