bson_tutorial - Man Page


Using Libbson in Your C Program

Include bson.h

All libbson's functions and types are available in one header file. Simply include bson.h:


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

main (int argc, const char **argv)
   bson_t *b;
   char *j;

   b = BCON_NEW ("hello", BCON_UTF8 ("bson!"));
   j = bson_as_canonical_extended_json (b, NULL);
   printf ("%s\n", j);

   bson_free (j);
   bson_destroy (b);

   return 0;


The libbson installation includes a CMake config-file package, so you can use CMake's find_package command to import libbson's CMake target and link to libbson (as a shared library):


# Specify the minimum version you require.
find_package (bson-1.0 1.7 REQUIRED)

# The "hello_bson.c" sample program is shared among four tests.
add_executable (hello_bson ../../hello_bson.c)
target_link_libraries (hello_bson PRIVATE mongo::bson_shared)

You can also use libbson as a static library instead: Use the mongo::bson_static CMake target:

# Specify the minimum version you require.
find_package (bson-1.0 1.7 REQUIRED)

# The "hello_bson.c" sample program is shared among four tests.
add_executable (hello_bson ../../hello_bson.c)
target_link_libraries (hello_bson PRIVATE mongo::bson_static)


If you're not using CMake, use pkg-config on the command line to set header and library paths:

gcc -o hello_bson hello_bson.c $(pkg-config --libs --cflags libbson-1.0)

Or to statically link to libbson:

gcc -o hello_bson hello_bson.c $(pkg-config --libs --cflags libbson-static-1.0)

Creating a BSON Document

The bson_t structure

BSON documents are created using the bson_t structure. This structure encapsulates the necessary logic for encoding using the BSON Specification. At the core, bson_t is a buffer manager and set of encoding routines.


BSON documents can live on the stack or the heap based on the performance needs or preference of the consumer.

Let's start by creating a new BSON document on the stack. Whenever using libbson, make sure you #include <bson/bson.h>.

bson_t b;

bson_init (&b);

This creates an empty document. In JSON, this would be the same as {}.

We can now proceed to adding items to the BSON document. A variety of functions prefixed with bson_append_ can be used based on the type of field you want to append. Let's append a Utf-8 encoded string.

bson_append_utf8 (&b, "key", -1, "value", -1);

Notice the two -1 parameters. The first indicates that the length of key in bytes should be determined with strlen(). Alternatively, we could have passed the number 3. The same goes for the second -1, but for value.

Libbson provides macros to make this less tedious when using string literals. The following two appends are identical.

bson_append_utf8 (&b, "key", -1, "value", -1);
BSON_APPEND_UTF8 (&b, "key", "value");

Now let's take a look at an example that adds a few different field types to a BSON document.


BSON_APPEND_INT32 (&b, "a", 1);
BSON_APPEND_UTF8 (&b, "hello", "world");
BSON_APPEND_BOOL (&b, "bool", true);

Notice that we omitted the call to bson_init(). By specifying BSON_INITIALIZER we can remove the need to initialize the structure to a base state.

Sub-Documents and Sub-Arrays

To simplify the creation of sub-documents and arrays, bson_append_document_begin() and bson_append_array_begin() exist. These can be used to build a sub-document using the parent documents memory region as the destination buffer.

bson_t parent;
bson_t child;
char *str;

bson_init (&parent);
bson_append_document_begin (&parent, "foo", 3, &child);
bson_append_int32 (&child, "baz", 3, 1);
bson_append_document_end (&parent, &child);

str = bson_as_canonical_extended_json (&parent, NULL);
printf ("%s\n", str);
bson_free (str);

bson_destroy (&parent);
{ "foo" : { "baz" : 1 } }

Simplified BSON C Object Notation

Creating BSON documents by hand can be tedious and time consuming. BCON, or BSON C Object Notation, was added to allow for the creation of BSON documents in a format that looks closer to the destination format.

The following example shows the use of BCON. Notice that values for fields are wrapped in the BCON_* macros. These are required for the variadic processor to determine the parameter type.

bson_t *doc;

doc = BCON_NEW ("foo",
                BCON_INT32 (1),
                BCON_INT32 (100),
                BCON_UTF8 ("value"),

Creates the following document

{ "foo" : { "int" : 1, "array" : [ 100, { "sub" : "value" } ] } }

Handling Errors


Many libbson functions report errors by returning NULL or -1 and filling out a bson_error_t structure with an error domain, error code, and message.

  • error.domain names the subsystem that generated the error.
  • error.code is a domain-specific error type.
  • error.message describes the error.

Some error codes overlap with others; always check both the domain and code to determine the type of error.

BSON_ERROR_JSONBSON_JSON_ERROR_READ_CORRUPT_JS BSON_JSON_ERROR_READ_INVALID_PARAM BSON_JSON_ERROR_READ_CB_FAILUREbson_json_reader_t tried to parse invalid MongoDB Extended JSON. Tried to parse a valid JSON document that is invalid as MongoDBExtended JSON. An internal callback failure during JSON parsing.
BSON_ERROR_READERBSON_ERROR_READER_BADFDbson_json_reader_new_from_file() could not open the file.


Libbson provides a simple way to generate ObjectIDs. It can be used in a single-threaded or multi-threaded manner depending on your requirements.

The bson_oid_t structure represents an ObjectID in MongoDB. It is a 96-bit identifier.


  • 4 bytes : The UNIX timestamp in big-endian format.
  • 5 bytes : A random number.
  • 3 bytes : A 24-bit monotonic counter incrementing from rand() in big-endian.

Sorting ObjectIDs

The typical way to sort in C is using qsort(). Therefore, Libbson provides a qsort() compatible callback function named bson_oid_compare(). It returns less than 1, greater than 1, or 0 depending on the equality of two bson_oid_t structures.

Comparing Object IDs

If you simply want to compare two bson_oid_t structures for equality, use bson_oid_equal().


To generate a bson_oid_t, you may use the following.

bson_oid_t oid;

bson_oid_init (&oid, NULL);

Parsing ObjectID Strings

You can also parse a string containing a bson_oid_t. The input string MUST be 24 characters or more in length.

bson_oid_t oid;

bson_oid_init_from_string (&oid, "123456789012345678901234");
bson_oid_t oid;

bson_oid_init_from_string_unsafe (&oid, "123456789012345678901234");

Hashing ObjectIDs

If you need to store items in a hashtable, you may want to use the bson_oid_t as the key. Libbson provides a hash function for just this purpose. It is based on DJB hash.

unsigned hash;

hash = bson_oid_hash (oid);

Fetching ObjectID Creation Time

You can easily fetch the time that a bson_oid_t was generated using bson_oid_get_time_t().

time_t t;

t = bson_oid_get_time_t (oid);
printf ("The OID was generated at %u\n", (unsigned) t);

Parsing and Iterating BSON Documents


BSON documents are lazily parsed as necessary. To begin parsing a BSON document, use one of the provided Libbson functions to create a new bson_t from existing data such as bson_new_from_data(). This will make a copy of the data so that additional mutations may occur to the BSON document.


If you only want to parse a BSON document and have no need to mutate it, you may use bson_init_static() to avoid making a copy of the data.

bson_t *b;

b = bson_new_from_data (my_data, my_data_len);
if (!b) {
   fprintf (stderr, "The specified length embedded in <my_data> did not match "

bson_destroy (b);

Only two checks are performed when creating a new bson_t from an existing buffer. First, the document must begin with the buffer length, matching what was expected by the caller. Second, the document must end with the expected trailing \0 byte.

To parse the document further we use a bson_iter_t to iterate the elements within the document. Let's print all of the field names in the document.

bson_t *b;
bson_iter_t iter;

if ((b = bson_new_from_data (my_data, my_data_len))) {
   if (bson_iter_init (&iter, b)) {
      while (bson_iter_next (&iter)) {
         printf ("Found element key: \"%s\"\n", bson_iter_key (&iter));
   bson_destroy (b);

Converting a document to JSON uses a bson_iter_t and bson_visitor_t to iterate all fields of a BSON document recursively and generate a Utf-8 encoded JSON string.

bson_t *b;
char *json;

if ((b = bson_new_from_data (my_data, my_data_len))) {
   if ((json = bson_as_canonical_extended_json (b, NULL))) {
      printf ("%s\n", json);
      bson_free (json);
   bson_destroy (b);

Recursing into Sub-Documents

Libbson provides convenient sub-iterators to dive down into a sub-document or sub-array. Below is an example that will dive into a sub-document named "foo" and print it's field names.

bson_iter_t iter;
bson_iter_t child;
char *json;

if (bson_iter_init_find (&iter, doc, "foo") &&
    BSON_ITER_HOLDS_DOCUMENT (&iter) && bson_iter_recurse (&iter, &child)) {
   while (bson_iter_next (&child)) {
      printf ("Found sub-key of \"foo\" named \"%s\"\n",
              bson_iter_key (&child));

Finding Fields using Dot Notation

Using the bson_iter_recurse() function exemplified above, bson_iter_find_descendant() can find a field for you using the MongoDB style path notation such as "".

Let's create a document like {"foo": {"bar": [{"baz: 1}]}} and locate the "baz" field.

bson_t *b;
bson_iter_t iter;
bson_iter_t baz;

b =
   BCON_NEW ("foo", "{", "bar", "[", "{", "baz", BCON_INT32 (1), "}", "]", "}");

if (bson_iter_init (&iter, b) &&
    bson_iter_find_descendant (&iter, "", &baz) &&
    BSON_ITER_HOLDS_INT32 (&baz)) {
   printf ("baz = %d\n", bson_iter_int32 (&baz));

bson_destroy (b);

Validating a BSON Document

If all you want to do is validate that a BSON document is valid, you can use bson_validate().

size_t err_offset;

if (!bson_validate (doc, BSON_VALIDATE_NONE, &err_offset)) {
   fprintf (stderr,
            "The document failed to validate at offset: %u\n",
            (unsigned) err_offset);

See the bson_validate() documentation for more information and examples.



Libbson expects that you are always working with UTF-8 encoded text. Anything else is invalid API use.

If you should need to walk through UTF-8 sequences, you can use the various UTF-8 helper functions distributed with Libbson.

Validating a UTF-8 Sequence

To validate the string contained in my_string, use the following. You may pass -1 for the string length if you know the string is NULL-terminated.

if (!bson_utf8_validate (my_string, -1, false)) {
   printf ("Validation failed.\n");

If my_string has NULL bytes within the string, you must provide the string length. Use the following format. Notice the true at the end indicating \0 is allowed.

if (!bson_utf8_validate (my_string, my_string_len, true)) {
   printf ("Validation failed.\n");

For more information see the API reference for bson_utf8_validate().


MongoDB, Inc


Jan 19, 2023 1.23.2 libbson