citadel/server/modules/rssclient/serv_rssclient.c

   1 // Bring external RSS and/or Atom feeds into rooms.  This module implements a
   2 // very loose parser that scrapes both kinds of feeds and is not picky about
   3 // the standards compliance of the source data.
   4 //
   5 // Copyright (c) 2007-2023 by the citadel.org team
   6 //
   7 // This program is open source software.  Use, duplication, or disclosure
   8 // is subject to the terms of the GNU General Public License, version 3.
   9
  10 #include <stdlib.h>
  11 #include <unistd.h>
  12 #include <stdio.h>
  13 #include <time.h>
  14 #include <ctype.h>
  15 #include <string.h>
  16 #include <errno.h>
  17 #include <sys/types.h>
  18 #include <sys/stat.h>
  19 #include <expat.h>
  20 #include <curl/curl.h>
  21 #include <libcitadel.h>
  22 #include "../../citadel_defs.h"
  23 #include "../../server.h"
  24 #include "../../citserver.h"
  25 #include "../../support.h"
  26 #include "../../config.h"
  27 #include "../../threads.h"
  28 #include "../../ctdl_module.h"
  29 #include "../../msgbase.h"
  30 #include "../../parsedate.h"
  31 #include "../../database.h"
  32 #include "../../citadel_dirs.h"
  33 #include "../../context.h"
  34 #include "../../internet_addressing.h"
  35
  36 struct rssfeed {
  37         char url[SIZ];                  // string containing the URL of an RSS or Atom feed
  38         char room[ROOMNAMELEN];         // the name of the room which is pulling this feed
  39 };
  40
  41 struct rssparser {
  42         char url[SIZ];
  43         char room[ROOMNAMELEN];
  44         StrBuf *CData;
  45         struct CtdlMessage *msg;
  46         char *link;
  47         char *description;
  48         char *item_id;
  49 };
  50
  51 time_t last_run = 0L;
  52
  53
  54 // This handler is called whenever an XML tag opens.
  55 void rss_start_element(void *data, const char *el, const char **attribute) {
  56         struct rssparser *r = (struct rssparser *)data;
  57         int i;
  58
  59         if (server_shutting_down) return;                       // shunt the whole operation if we're exiting
  60
  61         if (
  62                 (!strcasecmp(el, "entry"))
  63                 || (!strcasecmp(el, "item"))
  64         ) {
  65                 // this is the start of a new item(rss) or entry(atom)
  66                 if (r->msg != NULL) {
  67                         CM_Free(r->msg);
  68                         r->msg = NULL;
  69                 }
  70                 r->msg = malloc(sizeof(struct CtdlMessage));
  71                 memset(r->msg, 0, sizeof(struct CtdlMessage));
  72                 r->msg->cm_magic = CTDLMESSAGE_MAGIC;
  73                 r->msg->cm_anon_type = MES_NORMAL;
  74                 r->msg->cm_format_type = FMT_RFC822;
  75         }
  76
  77         else if (!strcasecmp(el, "link")) {                     // atom feeds have the link as an attribute
  78                 for(i = 0; attribute[i]; i += 2) {
  79                         if (!strcasecmp(attribute[i], "href")) {
  80                                 if (r->link != NULL) {
  81                                         free(r->link);
  82                                         r->link = NULL;
  83                                 }
  84                                 r->link = strdup(attribute[i+1]);
  85                                 string_trim(r->link);
  86                         }
  87                 }
  88         }
  89 }
  90
  91
  92 // This handler is called whenever an XML tag closes.
  93 void rss_end_element(void *data, const char *el) {
  94         struct rssparser *r = (struct rssparser *)data;
  95         StrBuf *encoded_field;
  96         long msgnum;
  97
  98         if (server_shutting_down) return;                                       // shunt the whole operation if we're exiting
  99
 100         if (StrLength(r->CData) > 0) {                                          // strip leading/trailing whitespace from field
 101                 StrBufTrim(r->CData);
 102         }
 103
 104         if ((!strcasecmp(el, "entry")) || (!strcasecmp(el, "item"))) {          // end of a new item(rss) or entry(atom)
 105                 if (r->msg != NULL) {                                           // Save the message to the room
 106
 107                         // use the link as an item id if nothing else is available
 108                         if ((r->item_id == NULL) && (r->link != NULL)) {
 109                                 r->item_id = strdup(r->link);
 110                         }
 111
 112                         // check the use table
 113                         StrBuf *u = NewStrBuf();
 114                         StrBufAppendPrintf(u, "rss/%s", r->item_id);
 115                         int already_seen = CheckIfAlreadySeen(u);
 116                         FreeStrBuf(&u);
 117
 118                         if (already_seen == 0) {
 119
 120                                 // Compose the message text
 121
 122                                 StrBuf *TheMessage = NewStrBuf();
 123                                 StrBufAppendPrintf(TheMessage, "<html><head></head><body>");
 124
 125                                 if (r->description != NULL) {
 126                                         StrBufAppendPrintf(TheMessage, "%s<br><br>\r\n", r->description);
 127                                         free(r->description);
 128                                         r->description = NULL;
 129                                 }
 130
 131                                 if (r->link != NULL) {
 132                                         StrBufAppendPrintf(TheMessage, "<a href=\"%s\">%s</a>\r\n", r->link, r->link);
 133                                         free(r->link);
 134                                         r->link = NULL;
 135                                 }
 136
 137                                 StrBufAppendPrintf(TheMessage, "</body></html>\r\n");
 138
 139                                 // Quoted-Printable encode the HTML message, because RSS and Atom make no guarantee of line length limits.
 140                                 StrBuf *TheMessage_Encoded = StrBufQuotedPrintableEncode(TheMessage);
 141
 142                                 // Now we reuse TheMessage -- this time it will contain the MIME headers concatenated with the encoded message.
 143                                 FlushStrBuf(TheMessage);
 144                                 StrBufAppendBufPlain(TheMessage, HKEY(
 145                                         "Content-type: text/html; charset=UTF-8\r\n"
 146                                         "Content-Transfer-Encoding: quoted-printable\r\n"
 147                                         "\r\n"
 148                                         ), 0
 149                                 );
 150                                 StrBufAppendBuf(TheMessage, TheMessage_Encoded, 0);
 151                                 FreeStrBuf(&TheMessage_Encoded);
 152
 153                                 CM_SetField(r->msg, eMessageText, ChrPtr(TheMessage));
 154                                 FreeStrBuf(&TheMessage);
 155
 156                                 if (CM_IsEmpty(r->msg, eAuthor)) {
 157                                         CM_SetField(r->msg, eAuthor, "rss");
 158                                 }
 159
 160                                 if (CM_IsEmpty(r->msg, eTimestamp)) {
 161                                         CM_SetFieldLONG(r->msg, eTimestamp, time(NULL));
 162                                 }
 163
 164                                 CtdlSubmitMsg(r->msg, NULL, r->room);
 165                         }
 166                         else {
 167                                 syslog(LOG_DEBUG, "rssclient: already seen %s", r->item_id);
 168                         }
 169
 170                         CM_Free(r->msg);
 171                         r->msg = NULL;
 172                 }
 173
 174                 if (r->item_id != NULL) {
 175                         free(r->item_id);
 176                         r->item_id = NULL;
 177                 }
 178         }
 179
 180         else if (!strcasecmp(el, "title")) {                    // item subject (rss and atom)
 181                 if ((r->msg != NULL) && (CM_IsEmpty(r->msg, eMsgSubject))) {
 182                         encoded_field = NewStrBuf();
 183                         StrBufRFC2047encode(&encoded_field, r->CData);
 184                         CM_SetAsFieldSB(r->msg, eMsgSubject, &encoded_field);
 185                 }
 186         }
 187
 188         else if (!strcasecmp(el, "creator")) {                  // <creator> can be used if <author> is not present
 189                 if ((r->msg != NULL) && (CM_IsEmpty(r->msg, eAuthor))) {
 190                         encoded_field = NewStrBuf();
 191                         StrBufRFC2047encode(&encoded_field, r->CData);
 192                         CM_SetAsFieldSB(r->msg, eAuthor, &encoded_field);
 193                 }
 194         }
 195
 196         else if (!strcasecmp(el, "author")) {                   // <author> supercedes <creator> if both are present
 197                 if (r->msg != NULL) {
 198                         encoded_field = NewStrBuf();
 199                         StrBufRFC2047encode(&encoded_field, r->CData);
 200                         CM_SetAsFieldSB(r->msg, eAuthor, &encoded_field);
 201                 }
 202         }
 203
 204         else if (!strcasecmp(el, "pubdate")) {                  // date/time stamp (rss) Sat, 25 Feb 2017 14:28:01 EST
 205                 if ((r->msg)&&(r->msg->cm_fields[eTimestamp]==NULL)) {
 206                         CM_SetFieldLONG(r->msg, eTimestamp, parsedate(ChrPtr(r->CData)));
 207                 }
 208         }
 209
 210         else if (!strcasecmp(el, "updated")) {                  // date/time stamp (atom) 2003-12-13T18:30:02Z
 211                 if ((r->msg)&&(r->msg->cm_fields[eTimestamp]==NULL)) {
 212                         struct tm t;
 213                         char zulu;
 214                         memset(&t, 0, sizeof t);
 215                         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);
 216                         t.tm_year -= 1900;
 217                         t.tm_mon -= 1;
 218                         CM_SetFieldLONG(r->msg, eTimestamp, mktime(&t));
 219                 }
 220         }
 221
 222         else if (!strcasecmp(el, "link")) {                     // link to story (rss)
 223                 if (r->link != NULL) {
 224                         free(r->link);
 225                         r->link = NULL;
 226                 }
 227                 r->link = strdup(ChrPtr(r->CData));
 228         }
 229
 230         else if (
 231                 (!strcasecmp(el, "guid"))                       // unique item id (rss)
 232                 || (!strcasecmp(el, "id"))                      // unique item id (atom)
 233         ) {
 234                 if (r->item_id != NULL) {
 235                         free(r->item_id);
 236                         r->item_id = NULL;
 237                 }
 238                 r->item_id = strdup(ChrPtr(r->CData));
 239         }
 240
 241         else if (
 242                 (!strcasecmp(el, "description"))                // message text (rss)
 243                 || (!strcasecmp(el, "summary"))                 // message text (atom)
 244                 || (!strcasecmp(el, "content"))                 // message text (atom)
 245         ) {
 246                 if (r->description != NULL) {
 247                         free(r->description);
 248                         r->description = NULL;
 249                 }
 250                 r->description = strdup(ChrPtr(r->CData));
 251         }
 252
 253         if (r->CData != NULL) {
 254                 FreeStrBuf(&r->CData);
 255                 r->CData = NULL;
 256         }
 257 }
 258
 259
 260 // This handler is called whenever data appears between opening and closing tags.
 261 void rss_handle_data(void *data, const char *content, int length) {
 262         struct rssparser *r = (struct rssparser *)data;
 263
 264         if (r->CData == NULL) {
 265                 r->CData = NewStrBuf();
 266         }
 267
 268         StrBufAppendBufPlain(r->CData, content, length, 0);
 269 }
 270
 271
 272 // Feed has been downloaded, now parse it.
 273 // `Feed` is the actual RSS downloaded from the site.
 274 // `url` is a string containing the feed URL
 275 void rss_parse_feed(StrBuf *Feed, char *url, char *room) {
 276         struct rssparser r;
 277
 278         memset(&r, 0, sizeof r);
 279         strcpy(r.url, url);
 280         strcpy(r.room, room);
 281         XML_Parser p = XML_ParserCreate("UTF-8");
 282         XML_SetElementHandler(p, rss_start_element, rss_end_element);
 283         XML_SetCharacterDataHandler(p, rss_handle_data);
 284         XML_SetUserData(p, (void *)&r);
 285         XML_Parse(p, ChrPtr(Feed), StrLength(Feed), XML_TRUE);
 286         XML_ParserFree(p);
 287 }
 288
 289
 290 // pull one RSS feed and save it to a room
 291 void rss_pull_one_feed(char *url, char *room) {
 292         CURL *curl;
 293         CURLcode res;
 294         StrBuf *Downloaded = NULL;
 295
 296         curl = curl_easy_init();
 297         if (!curl) {
 298                 return;
 299         }
 300
 301         Downloaded = NewStrBuf();
 302
 303         syslog(LOG_DEBUG, "rssclient: fetching %s", url);
 304         curl_easy_setopt(curl, CURLOPT_URL, url);
 305         curl_easy_setopt(curl, CURLOPT_SSL_VERIFYPEER, 0L);
 306         curl_easy_setopt(curl, CURLOPT_SSL_VERIFYHOST, 0L);
 307         curl_easy_setopt(curl, CURLOPT_FOLLOWLOCATION, 1L);                     // Follow redirects
 308         curl_easy_setopt(curl, CURLOPT_WRITEFUNCTION, CurlFillStrBuf_callback); // What to do with downloaded data
 309         curl_easy_setopt(curl, CURLOPT_WRITEDATA, Downloaded);                  // Give it our StrBuf to work with
 310         curl_easy_setopt(curl, CURLOPT_TIMEOUT, 20L);                           // Time out after 20 seconds
 311         res = curl_easy_perform(curl);                                          // Perform the request
 312         if (res != CURLE_OK) {
 313                 syslog(LOG_WARNING, "rssclient: failed to load feed: %s", curl_easy_strerror(res));
 314         }
 315         curl_easy_cleanup(curl);
 316
 317         rss_parse_feed(Downloaded, url, room);
 318         FreeStrBuf(&Downloaded);                                                // free the downloaded feed data
 319 }
 320
 321
 322 // Scan a room's netconfig looking for RSS feed parsing requests
 323 void rssclient_scan_room(struct ctdlroom *qrbuf, void *data) {
 324         char *serialized_config = NULL;
 325         int num_configs = 0;
 326         char cfgline[SIZ];
 327         struct rssfeed one_feed;
 328         int i = 0;
 329         Array *feeds = (Array *)data;
 330
 331         if (server_shutting_down) return;
 332
 333         serialized_config = LoadRoomNetConfigFile(qrbuf->QRnumber);
 334         if (!serialized_config) {
 335                 return;
 336         }
 337
 338         num_configs = num_tokens(serialized_config, '\n');
 339         for (i=0; i<num_configs; ++i) {
 340                 extract_token(cfgline, serialized_config, i, '\n', sizeof cfgline);
 341                 if (!strncasecmp(cfgline, HKEY("rssclient|"))) {
 342                         strcpy(cfgline, &cfgline[10]);
 343                         char *vbar = strchr(cfgline, '|');
 344                         if (vbar != NULL) {
 345                                 *vbar = 0;
 346                         }
 347                         safestrncpy(one_feed.url, cfgline, SIZ);
 348                         safestrncpy(one_feed.room, qrbuf->QRname, ROOMNAMELEN);
 349                         array_append(feeds, &one_feed);
 350                 }
 351         }
 352
 353         free(serialized_config);
 354 }
 355
 356
 357 // Scan for rooms that have RSS client requests configured
 358 void rssclient_scan(void) {
 359         time_t now = time(NULL);
 360
 361         // Run no more than once every 15 minutes.
 362         if ((now - last_run) < 900) {
 363                 syslog(LOG_DEBUG,
 364                         "rssclient: polling interval not yet reached; last run was %ldm%lds ago",
 365                         ((now - last_run) / 60),
 366                         ((now - last_run) % 60)
 367                 );
 368                 return;
 369         }
 370
 371         syslog(LOG_DEBUG, "rssclient: started");
 372         Array *feeds = array_new(sizeof(struct rssfeed));
 373         if (feeds == NULL) {
 374                 syslog(LOG_DEBUG, "rssclient: cannot allocate memory for feed list");
 375                 return;
 376         }
 377         CtdlForEachRoom(rssclient_scan_room, feeds);
 378
 379         for (int i=0; i<array_len(feeds); ++i) {
 380                 struct rssfeed *r = (struct rssfeed *) array_get_element_at(feeds, i);
 381                 rss_pull_one_feed(r->url, r->room);
 382         }
 383
 384         array_free(feeds);
 385         syslog(LOG_DEBUG, "rssclient: ended");
 386         last_run = time(NULL);
 387         return;
 388 }
 389
 390
 391 // Initialization function, called from modules_init.c
 392 char *ctdl_module_init_rssclient(void) {
 393         if (!threading) {
 394                 syslog(LOG_INFO, "rssclient: using %s", curl_version());
 395                 CtdlRegisterSessionHook(rssclient_scan, EVT_TIMER, PRIO_AGGR + 300);
 396         }
 397         return "rssclient";
 398 }