Cursors

Handling Cursor Failures

Cursors exist on a MongoDB server. However, the mongoc_cursor_t structure gives the local process a handle to the cursor. It is possible for errors to occur on the server while iterating a cursor on the client. Even a network partition may occur. This means that applications should be robust in handling cursor failures.

While iterating cursors, you should check to see if an error has occurred. See the following example for how to robustly check for errors.

static void
print_all_documents (mongoc_collection_t *collection)
{
   mongoc_cursor_t *cursor;
   const bson_t *doc;
   bson_error_t error;
   bson_t query = BSON_INITIALIZER;
   char *str;

   cursor = mongoc_collection_find_with_opts (collection, query, NULL, NULL);

   while (mongoc_cursor_next (cursor, &doc)) {
      str = bson_as_canonical_extended_json (doc, NULL);
      printf ("%s\n", str);
      bson_free (str);
   }

   if (mongoc_cursor_error (cursor, &error)) {
      fprintf (stderr, "Failed to iterate all documents: %s\n", error.message);
   }

   mongoc_cursor_destroy (cursor);
}

Destroying Server-Side Cursors

The MongoDB C driver will automatically destroy a server-side cursor when mongoc_cursor_destroy() is called. Failure to call this function when done with a cursor will leak memory client side as well as consume extra memory server side. If the cursor was configured to never timeout, it will become a memory leak on the server.

Tailable Cursors

Tailable cursors are cursors that remain open even after they’ve returned a final result. This way, if more documents are added to a collection (i.e., to the cursor’s result set), then you can continue to call mongoc_cursor_next() to retrieve those additional results.

Here’s a complete test case that demonstrates the use of tailable cursors.

Note

Tailable cursors are for capped collections only.

An example to tail the oplog from a replica set.

mongoc-tail.c
#include <bson.h>
#include <mongoc.h>
#include <stdio.h>
#include <stdlib.h>

#ifdef _WIN32
#define sleep(_n) Sleep ((_n) *1000)
#endif


static void
print_bson (const bson_t *b)
{
   char *str;

   str = bson_as_canonical_extended_json (b, NULL);
   fprintf (stdout, "%s\n", str);
   bson_free (str);
}


static mongoc_cursor_t *
query_collection (mongoc_collection_t *collection, uint32_t last_time)
{
   mongoc_cursor_t *cursor;
   bson_t query;
   bson_t gt;
   bson_t opts;

   BSON_ASSERT (collection);

   bson_init (&query);
   BSON_APPEND_DOCUMENT_BEGIN (&query, "ts", &gt);
   BSON_APPEND_TIMESTAMP (&gt, "$gt", last_time, 0);
   bson_append_document_end (&query, &gt);

   bson_init (&opts);
   BSON_APPEND_BOOL (&opts, "tailable", true);
   BSON_APPEND_BOOL (&opts, "awaitData", true);

   cursor = mongoc_collection_find_with_opts (collection, &query, &opts, NULL);

   bson_destroy (&query);
   bson_destroy (&opts);

   return cursor;
}


static void
tail_collection (mongoc_collection_t *collection)
{
   mongoc_cursor_t *cursor;
   uint32_t last_time;
   const bson_t *doc;
   bson_error_t error;
   bson_iter_t iter;

   BSON_ASSERT (collection);

   last_time = (uint32_t) time (NULL);

   while (true) {
      cursor = query_collection (collection, last_time);
      while (!mongoc_cursor_error (cursor, &error) &&
             mongoc_cursor_more (cursor)) {
         if (mongoc_cursor_next (cursor, &doc)) {
            if (bson_iter_init_find (&iter, doc, "ts") &&
                BSON_ITER_HOLDS_TIMESTAMP (&iter)) {
               bson_iter_timestamp (&iter, &last_time, NULL);
            }
            print_bson (doc);
         }
      }
      if (mongoc_cursor_error (cursor, &error)) {
         if (error.domain == MONGOC_ERROR_SERVER) {
            fprintf (stderr, "%s\n", error.message);
            exit (1);
         }
      }

      mongoc_cursor_destroy (cursor);
      sleep (1);
   }
}


int
main (int argc, char *argv[])
{
   mongoc_collection_t *collection;
   mongoc_client_t *client;

   if (argc != 2) {
      fprintf (stderr, "usage: %s MONGO_URI\n", argv[0]);
      return EXIT_FAILURE;
   }

   mongoc_init ();

   client = mongoc_client_new (argv[1]);
   if (!client) {
      fprintf (stderr, "Invalid URI: \"%s\"\n", argv[1]);
      return EXIT_FAILURE;
   }

   mongoc_client_set_error_api (client, 2);

   collection = mongoc_client_get_collection (client, "local", "oplog.rs");

   tail_collection (collection);

   mongoc_collection_destroy (collection);
   mongoc_client_destroy (client);

   return 0;
}

Let’s compile and run this example against a replica set to see updates as they are made.

$ gcc -Wall -o mongoc-tail mongoc-tail.c $(pkg-config --cflags --libs libmongoc-1.0)
$ ./mongoc-tail mongodb://example.com/?replicaSet=myReplSet
{
    "h" : -8458503739429355503,
    "ns" : "test.test",
    "o" : {
        "_id" : {
            "$oid" : "5372ab0a25164be923d10d50"
        }
    },
    "op" : "i",
    "ts" : {
        "$timestamp" : {
            "i" : 1,
            "t" : 1400023818
        }
    },
    "v" : 2
}

The line of output is a sample from performing db.test.insert({}) from the mongo shell on the replica set.

See also mongoc_cursor_set_max_await_time_ms().