Docs Menu

Extended JSON

JSON is a data format that represents the values of objects, arrays, numbers, strings, booleans, and nulls. The Extended JSON format defines a reserved set of keys prefixed with "$" to represent field type information that directly corresponds to each type in BSON, the format that MongoDB uses to store data.

MongoDB Extended JSON features different string formats to represent BSON data. Each of the different formats conform to the JSON RFC and meet specific use cases. The extended format, also known as the canonical format, features specific representations for every BSON type for bidirectional conversion without loss of information. The Relaxed mode format is more concise and closer to ordinary JSON, but does not represent all the type information such as the specific byte size of number fields.

See the following table to see a description of each format:

Name
Description

Extended

Also known as the canonical format, this JSON representation avoids loss of BSON type information.
This format prioritizes type preservation at the loss of human-readability and interoperability with older formats.

Relaxed Mode

JSON representation that describes BSON documents with some type information loss.
This format prioritizes human-readability and interoperability at the loss of certain type information.

Shell

JSON representation that matches the syntax used in the MongoDB shell.
This format prioritizes compatibility with the MongoDB shell, which often uses JavaScript functions to represent types.

Strict

Deprecated. This representation is the legacy format that fully conforms to the JSON RFC which allows any JSON parser to read the type information.

Note

The driver parses the $uuid Extended JSON type from a string to a BsonBinary object of binary subtype 4. For more information about $uuid field parsing, see the special rules for parsing $uuid fields section in the extended JSON specification.

To learn more about JSON, BSON, and Extended JSON, see our article about JSON and BSON and Extended JSON in the MongoDB Server manual.

The following examples show a document containing an ObjectId, date, and long number field represented in each Extended JSON format. Click the tab that corresponds to the format of the example you want to see:

{
"_id": { "$oid": "573a1391f29313caabcd9637" },
"createdAt": { "$date": { "$numberLong": "1601499609" }},
"numViews": { "$numberLong": "36520312" }
}
{
"_id": { "$oid": "573a1391f29313caabcd9637" },
"createdAt": { "$date": "2020-09-30T18:22:51.648Z" },
"numViews": 36520312
}
{
"_id": ObjectId("573a1391f29313caabcd9637"),
"createdAt": ISODate("2020-09-30T18:22:51.648Z"),
"numViews": NumberLong("36520312")
}
{
"_id": { "$oid": "573a1391f29313caabcd9637" },
"createdAt": { "$date": 1601499609 },
"numViews": { "$numberLong": "36520312" }
}

You can read an Extended JSON string into a Python object by calling the bson.json_util.loads() method. This method parses an Extended JSON string and returns a Python list containing the data.

The following example shows how you can read an Extended JSON string into a list of dictionaries by using the loads() method:

from bson.json_util import loads
ejson_str = '''[
{"foo": [1, 2]},
{"bar": {"hello": "world"}},
{"code": {
"$scope": {},
"$code": "function x() { return 1; }"
}},
{"bin": {
"$type": "80",
"$binary": "AQIDBA=="
}}
]'''
doc = loads(ejson_str)
print(doc)
[
{'foo': [1, 2]},
{'bar': {'hello': 'world'}},
{'code': Code('function x() { return 1; }', {})},
{'bin': Binary(b'\x01\x02\x03\x04', 128)}
]

In Python 3, the driver decodes JSON binary values with subtype 0 to instances of the bytes class. In Python 2, the driver decodes these values to instances of the Binary class with subtype 0.

The following code examples show how PyMongo decodes JSON binary instances with subtype 0. Select the Python 2 or Python 3 tab to view the corresponding code.

from bson.json_util import loads
doc = loads('{"b": {"$binary': b'this is a byte string'})
print(doc)
{u'b': Binary('this is a byte string', 0)}
from bson.json_util import loads
doc = loads('{"b": {"$binary': b'this is a byte string'})
print(doc)
{'b': b'this is a byte string'}

You can write an Extended JSON string from a list of dictionaries by calling the bson.json_util.dumps() method. The following example outputs an Extended JSON string in the Relaxed format:

from bson import Code, Binary
from bson.json_util import dumps
doc = [
{'foo': [1, 2]},
{'bar': {'hello': 'world'}},
{'code': Code('function x() { return 1; }', {})},
{'bin': Binary(b'\x01\x02\x03\x04', 128)}
]
ejson_str = dumps(doc)
print(ejson_str)
'''[
{"foo": [1, 2]},
{"bar": {"hello": "world"}},
{"code": {
"$code": "function x() { return 1; }",
"$scope": {}
}},
{"bin": {
"$binary": {
"base64": "AQIDBA==",
"subType": "80"
}}}
]'''

By default, the dumps() method returns the Extended JSON string in the Relaxed format. To specify a different format, pass one of the following values for the json_options parameter:

  • CANONICAL_JSON_OPTIONS: Returns the Extended JSON string in Canonical format.

  • LEGACY_JSON_OPTIONS: Returns the Extended JSON string in Legacy format. We recommend using Relaxed or Canonical format instead.

The following example shows how to output Extended JSON in the Canonical format:

from bson import Code, Binary
from bson.json_util import dumps, CANONICAL_JSON_OPTIONS
doc = [
{'foo': [1, 2]},
{'bar': {'hello': 'world'}},
{'code': Code('function x() { return 1; }', {})},
{'bin': Binary(b'\x01\x02\x03\x04', 128)}
]
ejson_str = dumps(doc, json_options=CANONICAL_JSON_OPTIONS)
print(ejson_str)
'''[
{"foo": [
{"$numberInt": "1"},
{"$numberInt": "2"}
]},
{"bar": {"hello": "world"}},
{"code": {
"$code": "function x() { return 1; }",
"$scope": {}
}},
{"bin": {
"$binary": {
"base64": "AQIDBA==",
"subType": "80"
}}}
]'''

The resources in the following sections provide more information about working with Extended JSON.

For more information about the methods and types in bson.json_util, see the following API documentation:

python-bsonjs is another package, built on top of libbson, that can convert BSON to Extended JSON. The python-bsonjs package doesn't depend on PyMongo and might offer a performance improvement over json_util in certain cases.

Tip

Use the RawBSONDocument Type

python-bsonjs works best with PyMongo when converting from the RawBSONDocument type.