BSON is a record-like data type with a standard binary representation defined at http://www.bsonspec.org. This implements version 1.0 of that spec. The standard binary form allows for easy data interchange between systems. In particular, MongoDB uses it for exchanging data between the MongoDB server and its clients.
The root BSON data type is bson:document(), a list of name-value pairs, analogous to an associative array, dictionary, or record. In this implementation, for writability and readability, the list of pairs is flattened (i.e. the tuples for each pair are elided), and the list is actually a tuple to distinguish it from list (array) of values. Hence a document is a tuple with alternating name and value elements, where a name is an atom() and a value is a bson:value(), which includes basic types like boolean(), number(), atom(), bson:utf8() (string), and compound types like [bson:value()] and bson:document(). For example,
> Doc = {x,<<"abc">>, y,[1,2,3], z,{a,'Foo', b,4.2}}.
is a document with three fields: {x,<<"abc">>} and {y,[1,2,3]}, and {z,{a,'Foo', b,4.2}}. There is a function bson:fields that converts a document to a list of fields but normally you don't need it. Instead you should use the following operations on documents: bson:lookup, bson:at, bson:include, bson:exclude, bson:update, bson:merge, and bson:append.
> {[1,2,3]} = bson:lookup (y, Doc).
> {} = bson:lookup (w, Doc).
> [1,2,3] = bson:at (y, Doc). % error if missing
> {x,<<"abc">>, y,[1,2,3]} = bson:include ([x, y], Doc).
> {z,{a,'Foo', b,4.2}} = bson:exclude ([x, y], Doc).
> {x,<<"abc">>, y,[1,2,3], z,null} = bson:update (z, null, Doc).
> {x,<<"abc">>, y,[1,2,3], z,null, w,1} = bson:merge ({w,1, z,null}, Doc).
> {w,1, x,<<"abc">>, y,[1,2,3], z,{a,'Foo', b,4.2}} = bson:append ({w,1}, Doc).
For the full list of bson:value() types see the bson module. Notice that an Erlang string() will be interpreted as a list of integers, so remember to alway delimit string literals with binary brackets (eg. <<"abc">>) and convert string variables using bson:utf8. You may be tempted to use atoms instead of strings, but you should only use atoms for enumerated types.
There are some special bson:value() types like bson:javascript() that are tagged tuples, eg. {javascript, {x,1}, <<"function (y) {return y + x}">>}. But embedded documents are also tuples, so how do we distinguish between the two? Tagged tuple bson:value() values intentionally have an odd number of elements, to distinguish them from documents, which always have an even number of elements, as they store key-value pairs.
API Docs - Documentation generated from source code comments.
Erlang Maps support is enabled.
To encode map just use bson_binary:put_document/1:
MapWithMap = #{<<"map">> => true, <<"simple">> => <<"not">>, <<"why">> => #{<<"because">> => <<"with map">>, <<"ok">> => true}},
Encoded3 = bson_binary:put_document(MapWithMap).
Map will be treated as bson document.
To decode binary not as bson document, but as map - use bson_binary:get_map/1:
{GotMap3, <<>>} = bson_binary:get_map(Encoded3).
See bson_tests:maps_get_test/0 for more details.