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 #include <time.h>
  21 #include <ctype.h>
  22 #include <string.h>
  23 #include <errno.h>
  24 #include <sys/types.h>
  25 #include <sys/stat.h>
  26 #include <expat.h>
  27 #include <curl/curl.h>
  28 #include <libcitadel.h>
  29 #include "citadel.h"
  30 #include "server.h"
  31 #include "citserver.h"
  32 #include "support.h"
  33 #include "config.h"
  34 #include "threads.h"
  35 #include "ctdl_module.h"
  36 #include "msgbase.h"
  37 #include "parsedate.h"
  38 #include "database.h"
  39 #include "citadel_dirs.h"
  40 #include "context.h"
  41 #include "internet_addressing.h"
  42
  43 struct rssroom {
  44         struct rssroom *next;
  45         char *room;
  46 };
  47
  48 struct rssurl {
  49         struct rssurl *next;
  50         char *url;
  51         struct rssroom *rooms;
  52 };
  53
  54 struct rssparser {
  55         StrBuf *CData;
  56         struct CtdlMessage *msg;
  57         char *link;
  58         char *description;
  59         char *item_id;
  60         struct rssroom *rooms;
  61 };
  62
  63 time_t last_run = 0L;
  64 struct rssurl *rsstodo = NULL;
  65
  66
  67 // This handler is called whenever an XML tag opens.
  68 //
  69 void rss_start_element(void *data, const char *el, const char **attribute)
  70 {
  71         struct rssparser *r = (struct rssparser *)data;
  72         int i;
  73
  74         if (
  75                 (!strcasecmp(el, "entry"))
  76                 || (!strcasecmp(el, "item"))
  77         ) {
  78                 // this is the start of a new item(rss) or entry(atom)
  79                 if (r->msg != NULL) {
  80                         CM_Free(r->msg);
  81                         r->msg = NULL;
  82                 }
  83                 r->msg = malloc(sizeof(struct CtdlMessage));
  84                 memset(r->msg, 0, sizeof(struct CtdlMessage));
  85                 r->msg->cm_magic = CTDLMESSAGE_MAGIC;
  86                 r->msg->cm_anon_type = MES_NORMAL;
  87                 r->msg->cm_format_type = FMT_RFC822;
  88         }
  89
  90         else if (!strcasecmp(el, "link")) {                     // atom feeds have the link as an attribute
  91                 for(i = 0; attribute[i]; i += 2) {
  92                         if (!strcasecmp(attribute[i], "href")) {
  93                                 if (r->link != NULL) {
  94                                         free(r->link);
  95                                         r->link = NULL;
  96                                 }
  97                                 r->link = strdup(attribute[i+1]);
  98                                 striplt(r->link);
  99                         }
 100                 }
 101         }
 102 }
 103
 104
 105 // This handler is called whenever an XML tag closes.
 106 //
 107 void rss_end_element(void *data, const char *el)
 108 {
 109         struct rssparser *r = (struct rssparser *)data;
 110         StrBuf *encoded_field;
 111
 112         if (StrLength(r->CData) > 0) {                          // strip leading/trailing whitespace from field
 113                 StrBufTrim(r->CData);
 114         }
 115
 116         if (                                                    // end of a new item(rss) or entry(atom)
 117                 (!strcasecmp(el, "entry"))
 118                 || (!strcasecmp(el, "item"))
 119         ) {
 120                 if (r->msg != NULL) {                           // Save the message to the rooms
 121
 122                         // use the link as an item id if nothing else is available
 123                         if ((r->item_id == NULL) && (r->link != NULL)) {
 124                                 r->item_id = strdup(r->link);
 125                         }
 126
 127                         // check the use table
 128                         StrBuf *u = NewStrBuf();
 129                         StrBufAppendPrintf(u, "rss/%s", r->item_id);
 130                         int already_seen = CheckIfAlreadySeen(u);
 131                         FreeStrBuf(&u);
 132
 133                         if (already_seen == 0) {
 134
 135                                 // Compose the message text
 136                                 StrBuf *TheMessage = NewStrBuf();
 137                                 StrBufAppendPrintf(TheMessage,
 138                                         "Content-type: text/html\n\n"
 139                                         "\n\n"
 140                                         "<html><head></head><body>"
 141                                 );
 142
 143                                 if (r->description != NULL) {
 144                                         StrBufAppendPrintf(TheMessage, "%s<br><br>\r\n", r->description);
 145                                         free(r->description);
 146                                         r->description = NULL;
 147                                 }
 148
 149                                 if (r->link != NULL) {
 150                                         StrBufAppendPrintf(TheMessage, "<a href=\"%s\">%s</a>\r\n", r->link, r->link);
 151                                         free(r->link);
 152                                         r->link = NULL;
 153                                 }
 154
 155                                 StrBufAppendPrintf(TheMessage, "</body></html>\r\n");
 156                                 CM_SetField(r->msg, eMesageText, ChrPtr(TheMessage), StrLength(TheMessage));
 157                                 FreeStrBuf(&TheMessage);
 158
 159                                 if (CM_IsEmpty(r->msg, eAuthor)) {
 160                                         CM_SetField(r->msg, eAuthor, HKEY("rss"));
 161                                 }
 162
 163                                 if (CM_IsEmpty(r->msg, eTimestamp)) {
 164                                         CM_SetFieldLONG(r->msg, eTimestamp, time(NULL));
 165                                 }
 166
 167                                 // Save it to the room(s)
 168                                 struct rssroom *rr = NULL;
 169                                 long msgnum = (-1);
 170                                 for (rr=r->rooms; rr!=NULL; rr=rr->next) {
 171                                         if (rr == r->rooms) {
 172                                                 msgnum = CtdlSubmitMsg(r->msg, NULL, rr->room);         // in first room, save msg
 173                                         }
 174                                         else {
 175                                                 CtdlSaveMsgPointerInRoom(rr->room, msgnum, 0, NULL);    // elsewhere, save a pointer
 176                                         }
 177                                         syslog(LOG_DEBUG, "rssclient: saved message %ld to %s", msgnum, rr->room);
 178                                 }
 179                         }
 180                         else {
 181                                 syslog(LOG_DEBUG, "rssclient: already seen %s", r->item_id);
 182                         }
 183
 184                         CM_Free(r->msg);
 185                         r->msg = NULL;
 186                 }
 187
 188                 if (r->item_id != NULL) {
 189                         free(r->item_id);
 190                         r->item_id = NULL;
 191                 }
 192         }
 193
 194         else if (!strcasecmp(el, "title")) {                    // item subject (rss and atom)
 195                 if ((r->msg != NULL) && (CM_IsEmpty(r->msg, eMsgSubject))) {
 196                         encoded_field = NewStrBuf();
 197                         StrBufRFC2047encode(&encoded_field, r->CData);
 198                         CM_SetAsFieldSB(r->msg, eMsgSubject, &encoded_field);
 199                 }
 200         }
 201
 202         else if (!strcasecmp(el, "creator")) {                  // <creator> can be used if <author> is not present
 203                 if ((r->msg != NULL) && (CM_IsEmpty(r->msg, eAuthor))) {
 204                         encoded_field = NewStrBuf();
 205                         StrBufRFC2047encode(&encoded_field, r->CData);
 206                         CM_SetAsFieldSB(r->msg, eAuthor, &encoded_field);
 207                 }
 208         }
 209
 210         else if (!strcasecmp(el, "author")) {                   // <author> supercedes <creator> if both are present
 211                 if (r->msg != NULL) {
 212                         encoded_field = NewStrBuf();
 213                         StrBufRFC2047encode(&encoded_field, r->CData);
 214                         CM_SetAsFieldSB(r->msg, eAuthor, &encoded_field);
 215                 }
 216         }
 217
 218         else if (!strcasecmp(el, "pubdate")) {                  // date/time stamp (rss) Sat, 25 Feb 2017 14:28:01 EST
 219                 if ((r->msg)&&(r->msg->cm_fields[eTimestamp]==NULL)) {
 220                         CM_SetFieldLONG(r->msg, eTimestamp, parsedate(ChrPtr(r->CData)));
 221                 }
 222         }
 223
 224         else if (!strcasecmp(el, "updated")) {                  // date/time stamp (atom) 2003-12-13T18:30:02Z
 225                 if ((r->msg)&&(r->msg->cm_fields[eTimestamp]==NULL)) {
 226                         struct tm t;
 227                         char zulu;
 228                         memset(&t, 0, sizeof t);
 229                         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);
 230                         t.tm_year -= 1900;
 231                         t.tm_mon -= 1;
 232                         CM_SetFieldLONG(r->msg, eTimestamp, mktime(&t));
 233                 }
 234         }
 235
 236         else if (!strcasecmp(el, "link")) {                     // link to story (rss)
 237                 if (r->link != NULL) {
 238                         free(r->link);
 239                         r->link = NULL;
 240                 }
 241                 r->link = strdup(ChrPtr(r->CData));
 242         }
 243
 244         else if (
 245                 (!strcasecmp(el, "guid"))                       // unique item id (rss)
 246                 || (!strcasecmp(el, "id"))                      // unique item id (atom)
 247         ) {
 248                 if (r->item_id != NULL) {
 249                         free(r->item_id);
 250                         r->item_id = NULL;
 251                 }
 252                 r->item_id = strdup(ChrPtr(r->CData));
 253         }
 254
 255         else if (
 256                 (!strcasecmp(el, "description"))                // message text (rss)
 257                 || (!strcasecmp(el, "summary"))                 // message text (atom)
 258                 || (!strcasecmp(el, "content"))                 // message text (atom)
 259         ) {
 260                 if (r->description != NULL) {
 261                         free(r->description);
 262                         r->description = NULL;
 263                 }
 264                 r->description = strdup(ChrPtr(r->CData));
 265         }
 266
 267         if (r->CData != NULL) {
 268                 FreeStrBuf(&r->CData);
 269                 r->CData = NULL;
 270         }
 271 }
 272
 273
 274 // This handler is called whenever data appears between opening and closing tags.
 275 //
 276 void rss_handle_data(void *data, const char *content, int length)
 277 {
 278         struct rssparser *r = (struct rssparser *)data;
 279
 280         if (r->CData == NULL) {
 281                 r->CData = NewStrBuf();
 282         }
 283
 284         StrBufAppendBufPlain(r->CData, content, length, 0);
 285 }
 286
 287
 288 // Feed has been downloaded, now parse it.
 289 //
 290 void rss_parse_feed(StrBuf *Feed, struct rssroom *rooms)
 291 {
 292         struct rssparser r;
 293
 294         memset(&r, 0, sizeof r);
 295         r.rooms = rooms;
 296         XML_Parser p = XML_ParserCreate("UTF-8");
 297         XML_SetElementHandler(p, rss_start_element, rss_end_element);
 298         XML_SetCharacterDataHandler(p, rss_handle_data);
 299         XML_SetUserData(p, (void *)&r);
 300         XML_Parse(p, ChrPtr(Feed), StrLength(Feed), XML_TRUE);
 301         XML_ParserFree(p);
 302 }
 303
 304
 305 // Add a feed/room pair into the todo list
 306 //
 307 void rssclient_push_todo(char *rssurl, char *roomname)
 308 {
 309         struct rssurl *r = NULL;
 310         struct rssurl *thisone = NULL;
 311         struct rssroom *newroom = NULL;
 312
 313         syslog(LOG_DEBUG, "rssclient: will fetch %s to %s", rssurl, roomname);
 314
 315         for (r=rsstodo; r!=NULL; r=r->next) {
 316                 if (!strcasecmp(r->url, rssurl)) {
 317                         thisone = r;
 318                 }
 319         }
 320
 321         if (thisone == NULL) {
 322                 thisone = malloc(sizeof(struct rssurl));
 323                 thisone->url = strdup(rssurl);
 324                 thisone->rooms = NULL;
 325                 thisone->next = rsstodo;
 326                 rsstodo = thisone;
 327         }
 328
 329         newroom = malloc(sizeof(struct rssroom));
 330         newroom->room = strdup(roomname);
 331         newroom->next = thisone->rooms;
 332         thisone->rooms = newroom;
 333 }
 334
 335
 336 // pull one feed (possibly multiple rooms)
 337 //
 338 void rss_pull_one_feed(struct rssurl *url)
 339 {
 340         CURL *curl;
 341         CURLcode res;
 342         StrBuf *Downloaded = NULL;
 343
 344         syslog(LOG_DEBUG, "rssclient: fetching %s", url->url);
 345
 346         curl = curl_easy_init();
 347         if (!curl) {
 348                 return;
 349         }
 350
 351         Downloaded = NewStrBuf();
 352
 353         curl_easy_setopt(curl, CURLOPT_URL, url->url);
 354         curl_easy_setopt(curl, CURLOPT_SSL_VERIFYPEER, 0L);
 355         curl_easy_setopt(curl, CURLOPT_SSL_VERIFYHOST, 0L);
 356         curl_easy_setopt(curl, CURLOPT_FOLLOWLOCATION, 1L);                     // Follow redirects
 357         curl_easy_setopt(curl, CURLOPT_WRITEFUNCTION, CurlFillStrBuf_callback); // What to do with downloaded data
 358         curl_easy_setopt(curl, CURLOPT_WRITEDATA, Downloaded);                  // Give it our StrBuf to work with
 359         curl_easy_setopt(curl, CURLOPT_TIMEOUT, 20L);                           // Time out after 20 seconds
 360         res = curl_easy_perform(curl);                                          // Perform the request
 361         if (res != CURLE_OK) {
 362                 syslog(LOG_WARNING, "rssclient: failed to load feed: %s", curl_easy_strerror(res));
 363         }
 364         curl_easy_cleanup(curl);
 365
 366         rss_parse_feed(Downloaded, url->rooms);                                 // parse the feed
 367         FreeStrBuf(&Downloaded);                                                // free the downloaded feed data
 368 }
 369
 370
 371 // We have a list, now download the feeds
 372 //
 373 void rss_pull_feeds(void)
 374 {
 375         struct rssurl *r;
 376         struct rssroom *rr;
 377
 378         while (rsstodo != NULL) {
 379                 rss_pull_one_feed(rsstodo);
 380                 r = rsstodo;
 381                 rsstodo = rsstodo->next;
 382                 while (r->rooms != NULL) {
 383                         rr = r->rooms;
 384                         r->rooms = r->rooms->next;
 385                         free(rr->room);
 386                         free(rr);
 387                 }
 388                 free(r->url);
 389                 free(r);
 390         }
 391 }
 392
 393
 394 // Scan a room's netconfig looking for RSS feed parsing requests
 395 //
 396 void rssclient_scan_room(struct ctdlroom *qrbuf, void *data)
 397 {
 398         char *serialized_config = NULL;
 399         int num_configs = 0;
 400         char cfgline[SIZ];
 401         int i = 0;
 402
 403         serialized_config = LoadRoomNetConfigFile(qrbuf->QRnumber);
 404         if (!serialized_config) {
 405                 return;
 406         }
 407
 408         num_configs = num_tokens(serialized_config, '\n');
 409         for (i=0; i<num_configs; ++i) {
 410                 extract_token(cfgline, serialized_config, i, '\n', sizeof cfgline);
 411                 if (!strncasecmp(cfgline, HKEY("rssclient|"))) {
 412                         strcpy(cfgline, &cfgline[10]);
 413                         char *vbar = strchr(cfgline, '|');
 414                         if (vbar != NULL) {
 415                                 *vbar = 0;
 416                         }
 417                         rssclient_push_todo(cfgline, qrbuf->QRname);
 418                 }
 419         }
 420
 421         free(serialized_config);
 422 }
 423
 424
 425 /*
 426  * Scan for rooms that have RSS client requests configured
 427  */
 428 void rssclient_scan(void) {
 429         time_t now = time(NULL);
 430
 431         /* Run no more than once every 15 minutes. */
 432         if ((now - last_run) < 900) {
 433                 syslog(LOG_DEBUG,
 434                         "rssclient: polling interval not yet reached; last run was %ldm%lds ago",
 435                         ((now - last_run) / 60),
 436                         ((now - last_run) % 60)
 437                 );
 438                 return;
 439         }
 440
 441         syslog(LOG_DEBUG, "rssclient: started");
 442         CtdlForEachRoom(rssclient_scan_room, NULL);
 443         rss_pull_feeds();
 444         syslog(LOG_DEBUG, "rssclient: ended");
 445         last_run = time(NULL);
 446         return;
 447 }
 448
 449
 450 CTDL_MODULE_INIT(rssclient)
 451 {
 452         if (!threading)
 453         {
 454                 syslog(LOG_INFO, "rssclient: using %s", curl_version());
 455                 CtdlRegisterSessionHook(rssclient_scan, EVT_TIMER, PRIO_AGGR + 300);
 456         }
 457         return "rssclient";
 458 }