mongoc_cursor_t *
mongoc_database_aggregate (mongoc_database_t *database,
                           const bson_t *pipeline,
                           const bson_t *opts,
                           const mongoc_read_prefs_t *read_prefs)


  • database: A mongoc_database_t.

  • pipeline: A bson_t, either a BSON array or a BSON document containing an array field named “pipeline”.

  • opts: A bson_t containing options for the command, or NULL.

  • read_prefs: A mongoc_read_prefs_t or NULL.

opts may be NULL or a BSON document with additional command options:

For a list of all options, see the MongoDB Manual entry on the aggregate command.


This function creates a cursor which sends the aggregate command on the underlying database upon the first call to mongoc_cursor_next(). For more information on building aggregation pipelines, see the MongoDB Manual entry on the aggregate command. Note that the pipeline must start with a compatible stage that does not require an underlying collection (e.g. “$currentOp”, “$listLocalSessions”).

Read preferences, read and write concern, and collation can be overridden by various sources. The highest-priority sources for these options are listed first in the following table. In a transaction, read concern and write concern are prohibited in opts and the read preference must be primary or NULL. Write concern is applied from opts, or if opts has no write concern and the aggregation pipeline includes “$out”, the write concern is applied from database.

Read Preferences

Read Concern

Write Concern












See the example for transactions and for the “distinct” command with opts.

This function is considered a retryable read operation unless the pipeline contains a write stage like $out or $merge. Upon a transient error (a network error, errors due to replica set failover, etc.) the operation is safely retried once. If retryreads is false in the URI (see mongoc_uri_t) the retry behavior does not apply.


This function returns a newly allocated mongoc_cursor_t that should be freed with mongoc_cursor_destroy() when no longer in use. The returned mongoc_cursor_t is never NULL, even on error. The user must call mongoc_cursor_next() on the returned mongoc_cursor_t to execute the initial command.

Cursor errors can be checked with mongoc_cursor_error_document(). It always fills out the bson_error_t if an error occurred, and optionally includes a server reply document if the error occurred server-side.


Failure to handle the result of this function is a programming error.


#include <bson/bson.h>
#include <mongoc/mongoc.h>

static mongoc_cursor_t *
current_op_query (mongoc_client_t *client)
   mongoc_cursor_t *cursor;
   mongoc_database_t *database;
   bson_t *pipeline;

   pipeline = BCON_NEW ("pipeline",

   /* $currentOp must be run on the admin database */
   database = mongoc_client_get_database (client, "admin");

   cursor = mongoc_database_aggregate (
      database, pipeline, NULL, NULL);

   bson_destroy (pipeline);
   mongoc_database_destroy (database);

   return cursor;