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) 2007-2020 by the citadel.org team
   7  *
   8  * This program is open source software; you can redistribute it and/or
   9  * modify 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 "context.h"
  52 #include "internet_addressing.h"
  53
  54 struct rssroom {
  55         struct rssroom *next;
  56         char *room;
  57 };
  58
  59 struct rssurl {
  60         struct rssurl *next;
  61         char *url;
  62         struct rssroom *rooms;
  63 };
  64
  65 struct rssparser {
  66         StrBuf *CData;
  67         struct CtdlMessage *msg;
  68         char *link;
  69         char *description;
  70         char *item_id;
  71         struct rssroom *rooms;
  72 };
  73
  74 time_t last_run = 0L;
  75 struct rssurl *rsstodo = NULL;
  76
  77
  78 // This handler is called whenever an XML tag opens.
  79 //
  80 void rss_start_element(void *data, const char *el, const char **attribute)
  81 {
  82         struct rssparser *r = (struct rssparser *)data;
  83         int i;
  84
  85         if (
  86                 (!strcasecmp(el, "entry"))
  87                 || (!strcasecmp(el, "item"))
  88         ) {
  89                 // this is the start of a new item(rss) or entry(atom)
  90                 if (r->msg != NULL) {
  91                         CM_Free(r->msg);
  92                         r->msg = NULL;
  93                 }
  94                 r->msg = malloc(sizeof(struct CtdlMessage));
  95                 memset(r->msg, 0, sizeof(struct CtdlMessage));
  96                 r->msg->cm_magic = CTDLMESSAGE_MAGIC;
  97                 r->msg->cm_anon_type = MES_NORMAL;
  98                 r->msg->cm_format_type = FMT_RFC822;
  99         }
 100
 101         else if (!strcasecmp(el, "link")) {                     // atom feeds have the link as an attribute
 102                 for(i = 0; attribute[i]; i += 2) {
 103                         if (!strcasecmp(attribute[i], "href")) {
 104                                 if (r->link != NULL) {
 105                                         free(r->link);
 106                                         r->link = NULL;
 107                                 }
 108                                 r->link = strdup(attribute[i+1]);
 109                                 striplt(r->link);
 110                         }
 111                 }
 112         }
 113 }
 114
 115
 116 // This handler is called whenever an XML tag closes.
 117 //
 118 void rss_end_element(void *data, const char *el)
 119 {
 120         struct rssparser *r = (struct rssparser *)data;
 121         StrBuf *encoded_field;
 122
 123         if (StrLength(r->CData) > 0) {                          // strip leading/trailing whitespace from field
 124                 StrBufTrim(r->CData);
 125         }
 126
 127         if (                                                    // end of a new item(rss) or entry(atom)
 128                 (!strcasecmp(el, "entry"))
 129                 || (!strcasecmp(el, "item"))
 130         ) {
 131                 if (r->msg != NULL) {                           // Save the message to the rooms
 132
 133                         // use the link as an item id if nothing else is available
 134                         if ((r->item_id == NULL) && (r->link != NULL)) {
 135                                 r->item_id = strdup(r->link);
 136                         }
 137
 138                         // check the use table
 139                         StrBuf *u = NewStrBuf();
 140                         StrBufAppendPrintf(u, "rss/%s", r->item_id);
 141                         int already_seen = CheckIfAlreadySeen(u);
 142                         FreeStrBuf(&u);
 143
 144                         if (already_seen == 0) {
 145
 146                                 // Compose the message text
 147                                 StrBuf *TheMessage = NewStrBuf();
 148                                 StrBufAppendPrintf(TheMessage,
 149                                         "Content-type: text/html\n\n"
 150                                         "\n\n"
 151                                         "<html><head></head><body>"
 152                                 );
 153
 154                                 if (r->description != NULL) {
 155                                         StrBufAppendPrintf(TheMessage, "%s<br><br>\r\n", r->description);
 156                                         free(r->description);
 157                                         r->description = NULL;
 158                                 }
 159
 160                                 if (r->link != NULL) {
 161                                         StrBufAppendPrintf(TheMessage, "<a href=\"%s\">%s</a>\r\n", r->link, r->link);
 162                                         free(r->link);
 163                                         r->link = NULL;
 164                                 }
 165
 166                                 StrBufAppendPrintf(TheMessage, "</body></html>\r\n");
 167                                 CM_SetField(r->msg, eMesageText, ChrPtr(TheMessage), StrLength(TheMessage));
 168                                 FreeStrBuf(&TheMessage);
 169
 170                                 if (CM_IsEmpty(r->msg, eAuthor)) {
 171                                         CM_SetField(r->msg, eAuthor, HKEY("rss"));
 172                                 }
 173
 174                                 if (CM_IsEmpty(r->msg, eTimestamp)) {
 175                                         CM_SetFieldLONG(r->msg, eTimestamp, time(NULL));
 176                                 }
 177
 178                                 // Save it to the room(s)
 179                                 struct rssroom *rr = NULL;
 180                                 long msgnum = (-1);
 181                                 for (rr=r->rooms; rr!=NULL; rr=rr->next) {
 182                                         if (rr == r->rooms) {
 183                                                 msgnum = CtdlSubmitMsg(r->msg, NULL, rr->room);         // in first room, save msg
 184                                         }
 185                                         else {
 186                                                 CtdlSaveMsgPointerInRoom(rr->room, msgnum, 0, NULL);    // elsewhere, save a pointer
 187                                         }
 188                                         syslog(LOG_DEBUG, "rssclient: saved message %ld to %s", msgnum, rr->room);
 189                                 }
 190                         }
 191                         else {
 192                                 syslog(LOG_DEBUG, "rssclient: already seen %s", r->item_id);
 193                         }
 194
 195                         CM_Free(r->msg);
 196                         r->msg = NULL;
 197                 }
 198
 199                 if (r->item_id != NULL) {
 200                         free(r->item_id);
 201                         r->item_id = NULL;
 202                 }
 203         }
 204
 205         else if (!strcasecmp(el, "title")) {                    // item subject (rss and atom)
 206                 if ((r->msg != NULL) && (CM_IsEmpty(r->msg, eMsgSubject))) {
 207                         encoded_field = NewStrBuf();
 208                         StrBufRFC2047encode(&encoded_field, r->CData);
 209                         CM_SetAsFieldSB(r->msg, eMsgSubject, &encoded_field);
 210                 }
 211         }
 212
 213         else if (!strcasecmp(el, "creator")) {                  // <creator> can be used if <author> is not present
 214                 if ((r->msg != NULL) && (CM_IsEmpty(r->msg, eAuthor))) {
 215                         encoded_field = NewStrBuf();
 216                         StrBufRFC2047encode(&encoded_field, r->CData);
 217                         CM_SetAsFieldSB(r->msg, eAuthor, &encoded_field);
 218                 }
 219         }
 220
 221         else if (!strcasecmp(el, "author")) {                   // <author> supercedes <creator> if both are present
 222                 if (r->msg != NULL) {
 223                         encoded_field = NewStrBuf();
 224                         StrBufRFC2047encode(&encoded_field, r->CData);
 225                         CM_SetAsFieldSB(r->msg, eAuthor, &encoded_field);
 226                 }
 227         }
 228
 229         else if (!strcasecmp(el, "pubdate")) {                  // date/time stamp (rss) Sat, 25 Feb 2017 14:28:01 EST
 230                 if ((r->msg)&&(r->msg->cm_fields[eTimestamp]==NULL)) {
 231                         CM_SetFieldLONG(r->msg, eTimestamp, parsedate(ChrPtr(r->CData)));
 232                 }
 233         }
 234
 235         else if (!strcasecmp(el, "updated")) {                  // date/time stamp (atom) 2003-12-13T18:30:02Z
 236                 if ((r->msg)&&(r->msg->cm_fields[eTimestamp]==NULL)) {
 237                         struct tm t;
 238                         char zulu;
 239                         memset(&t, 0, sizeof t);
 240                         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);
 241                         t.tm_year -= 1900;
 242                         t.tm_mon -= 1;
 243                         CM_SetFieldLONG(r->msg, eTimestamp, mktime(&t));
 244                 }
 245         }
 246
 247         else if (!strcasecmp(el, "link")) {                     // link to story (rss)
 248                 if (r->link != NULL) {
 249                         free(r->link);
 250                         r->link = NULL;
 251                 }
 252                 r->link = strdup(ChrPtr(r->CData));
 253         }
 254
 255         else if (
 256                 (!strcasecmp(el, "guid"))                       // unique item id (rss)
 257                 || (!strcasecmp(el, "id"))                      // unique item id (atom)
 258         ) {
 259                 if (r->item_id != NULL) {
 260                         free(r->item_id);
 261                         r->item_id = NULL;
 262                 }
 263                 r->item_id = strdup(ChrPtr(r->CData));
 264         }
 265
 266         else if (
 267                 (!strcasecmp(el, "description"))                // message text (rss)
 268                 || (!strcasecmp(el, "summary"))                 // message text (atom)
 269                 || (!strcasecmp(el, "content"))                 // message text (atom)
 270         ) {
 271                 if (r->description != NULL) {
 272                         free(r->description);
 273                         r->description = NULL;
 274                 }
 275                 r->description = strdup(ChrPtr(r->CData));
 276         }
 277
 278         if (r->CData != NULL) {
 279                 FreeStrBuf(&r->CData);
 280                 r->CData = NULL;
 281         }
 282 }
 283
 284
 285 // This handler is called whenever data appears between opening and closing tags.
 286 //
 287 void rss_handle_data(void *data, const char *content, int length)
 288 {
 289         struct rssparser *r = (struct rssparser *)data;
 290
 291         if (r->CData == NULL) {
 292                 r->CData = NewStrBuf();
 293         }
 294
 295         StrBufAppendBufPlain(r->CData, content, length, 0);
 296 }
 297
 298
 299 // Feed has been downloaded, now parse it.
 300 //
 301 void rss_parse_feed(StrBuf *Feed, struct rssroom *rooms)
 302 {
 303         struct rssparser r;
 304
 305         memset(&r, 0, sizeof r);
 306         r.rooms = rooms;
 307         XML_Parser p = XML_ParserCreate("UTF-8");
 308         XML_SetElementHandler(p, rss_start_element, rss_end_element);
 309         XML_SetCharacterDataHandler(p, rss_handle_data);
 310         XML_SetUserData(p, (void *)&r);
 311         XML_Parse(p, ChrPtr(Feed), StrLength(Feed), XML_TRUE);
 312         XML_ParserFree(p);
 313 }
 314
 315
 316 // Add a feed/room pair into the todo list
 317 //
 318 void rssclient_push_todo(char *rssurl, char *roomname)
 319 {
 320         struct rssurl *r = NULL;
 321         struct rssurl *thisone = NULL;
 322         struct rssroom *newroom = NULL;
 323
 324         syslog(LOG_DEBUG, "rssclient: will fetch %s to %s", rssurl, roomname);
 325
 326         for (r=rsstodo; r!=NULL; r=r->next) {
 327                 if (!strcasecmp(r->url, rssurl)) {
 328                         thisone = r;
 329                 }
 330         }
 331
 332         if (thisone == NULL) {
 333                 thisone = malloc(sizeof(struct rssurl));
 334                 thisone->url = strdup(rssurl);
 335                 thisone->rooms = NULL;
 336                 thisone->next = rsstodo;
 337                 rsstodo = thisone;
 338         }
 339
 340         newroom = malloc(sizeof(struct rssroom));
 341         newroom->room = strdup(roomname);
 342         newroom->next = thisone->rooms;
 343         thisone->rooms = newroom;
 344 }
 345
 346
 347 // pull one feed (possibly multiple rooms)
 348 //
 349 void rss_pull_one_feed(struct rssurl *url)
 350 {
 351         CURL *curl;
 352         CURLcode res;
 353         StrBuf *Downloaded = NULL;
 354
 355         syslog(LOG_DEBUG, "rssclient: fetching %s", url->url);
 356
 357         curl = curl_easy_init();
 358         if (!curl) {
 359                 return;
 360         }
 361
 362         Downloaded = NewStrBuf();
 363
 364         curl_easy_setopt(curl, CURLOPT_URL, url->url);
 365         curl_easy_setopt(curl, CURLOPT_SSL_VERIFYPEER, 0L);
 366         curl_easy_setopt(curl, CURLOPT_SSL_VERIFYHOST, 0L);
 367         curl_easy_setopt(curl, CURLOPT_FOLLOWLOCATION, 1L);                     // Follow redirects
 368         curl_easy_setopt(curl, CURLOPT_WRITEFUNCTION, CurlFillStrBuf_callback); // What to do with downloaded data
 369         curl_easy_setopt(curl, CURLOPT_WRITEDATA, Downloaded);                  // Give it our StrBuf to work with
 370         curl_easy_setopt(curl, CURLOPT_TIMEOUT, 20L);                           // Time out after 20 seconds
 371         res = curl_easy_perform(curl);                                          // Perform the request
 372         if (res != CURLE_OK) {
 373                 syslog(LOG_WARNING, "rssclient: failed to load feed: %s", curl_easy_strerror(res));
 374         }
 375         curl_easy_cleanup(curl);
 376
 377         rss_parse_feed(Downloaded, url->rooms);                                 // parse the feed
 378         FreeStrBuf(&Downloaded);                                                // free the downloaded feed data
 379 }
 380
 381
 382 // We have a list, now download the feeds
 383 //
 384 void rss_pull_feeds(void)
 385 {
 386         struct rssurl *r;
 387         struct rssroom *rr;
 388
 389         while (rsstodo != NULL) {
 390                 rss_pull_one_feed(rsstodo);
 391                 r = rsstodo;
 392                 rsstodo = rsstodo->next;
 393                 while (r->rooms != NULL) {
 394                         rr = r->rooms;
 395                         r->rooms = r->rooms->next;
 396                         free(rr->room);
 397                         free(rr);
 398                 }
 399                 free(r->url);
 400                 free(r);
 401         }
 402 }
 403
 404
 405 // Scan a room's netconfig looking for RSS feed parsing requests
 406 //
 407 void rssclient_scan_room(struct ctdlroom *qrbuf, void *data)
 408 {
 409         char *serialized_config = NULL;
 410         int num_configs = 0;
 411         char cfgline[SIZ];
 412         int i = 0;
 413
 414         serialized_config = LoadRoomNetConfigFile(qrbuf->QRnumber);
 415         if (!serialized_config) {
 416                 return;
 417         }
 418
 419         num_configs = num_tokens(serialized_config, '\n');
 420         for (i=0; i<num_configs; ++i) {
 421                 extract_token(cfgline, serialized_config, i, '\n', sizeof cfgline);
 422                 if (!strncasecmp(cfgline, HKEY("rssclient|"))) {
 423                         strcpy(cfgline, &cfgline[10]);
 424                         char *vbar = strchr(cfgline, '|');
 425                         if (vbar != NULL) {
 426                                 *vbar = 0;
 427                         }
 428                         rssclient_push_todo(cfgline, qrbuf->QRname);
 429                 }
 430         }
 431
 432         free(serialized_config);
 433 }
 434
 435
 436 /*
 437  * Scan for rooms that have RSS client requests configured
 438  */
 439 void rssclient_scan(void) {
 440         time_t now = time(NULL);
 441
 442         /* Run no more than once every 15 minutes. */
 443         if ((now - last_run) < 900) {
 444                 syslog(LOG_DEBUG,
 445                         "rssclient: polling interval not yet reached; last run was %ldm%lds ago",
 446                         ((now - last_run) / 60),
 447                         ((now - last_run) % 60)
 448                 );
 449                 return;
 450         }
 451
 452         syslog(LOG_DEBUG, "rssclient: started");
 453         CtdlForEachRoom(rssclient_scan_room, NULL);
 454         rss_pull_feeds();
 455         syslog(LOG_DEBUG, "rssclient: ended");
 456         last_run = time(NULL);
 457         return;
 458 }
 459
 460
 461 CTDL_MODULE_INIT(rssclient)
 462 {
 463         if (!threading)
 464         {
 465                 syslog(LOG_INFO, "rssclient: using %s", curl_version());
 466                 CtdlRegisterSessionHook(rssclient_scan, EVT_TIMER, PRIO_AGGR + 300);
 467         }
 468         return "rssclient";
 469 }
 470