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