184

My response back from MongoDB after querying an aggregated function on document using Python, It returns valid response and i can print it but can not return it.

Error:

TypeError: ObjectId('51948e86c25f4b1d1c0d303c') is not JSON serializable

Print:

{'result': [{'_id': ObjectId('51948e86c25f4b1d1c0d303c'), 'api_calls_with_key': 4, 'api_calls_per_day': 0.375, 'api_calls_total': 6, 'api_calls_without_key': 2}], 'ok': 1.0}

But When i try to return:

TypeError: ObjectId('51948e86c25f4b1d1c0d303c') is not JSON serializable

It is RESTfull call:

@appv1.route('/v1/analytics')
def get_api_analytics():
    # get handle to collections in MongoDB
    statistics = sldb.statistics

    objectid = ObjectId("51948e86c25f4b1d1c0d303c")

    analytics = statistics.aggregate([
    {'$match': {'owner': objectid}},
    {'$project': {'owner': "$owner",
    'api_calls_with_key': {'$cond': [{'$eq': ["$apikey", None]}, 0, 1]},
    'api_calls_without_key': {'$cond': [{'$ne': ["$apikey", None]}, 0, 1]}
    }},
    {'$group': {'_id': "$owner",
    'api_calls_with_key': {'$sum': "$api_calls_with_key"},
    'api_calls_without_key': {'$sum': "$api_calls_without_key"}
    }},
    {'$project': {'api_calls_with_key': "$api_calls_with_key",
    'api_calls_without_key': "$api_calls_without_key",
    'api_calls_total': {'$add': ["$api_calls_with_key", "$api_calls_without_key"]},
    'api_calls_per_day': {'$divide': [{'$add': ["$api_calls_with_key", "$api_calls_without_key"]}, {'$dayOfMonth': datetime.now()}]},
    }}
    ])


    print(analytics)

    return analytics

db is well connected and collection is there too and I got back valid expected result but when i try to return it gives me Json error. Any idea how to convert the response back into JSON. Thanks

20 Answers 20

242

Bson in PyMongo distribution provides json_util - you can use that one instead to handle BSON types

from bson import json_util

def parse_json(data):
    return json.loads(json_util.dumps(data))
Sign up to request clarification or add additional context in comments.

8 Comments

I agree with @tim, this is correct way to deal with BSON data coming from mongo. api.mongodb.org/python/current/api/bson/json_util.html
Yes, seems to be more of a hassle free if we use this way
An example here would be a little more helpful, as this is the best way but the linked documentation isn't the most user friendly for noobs
from bson import json_util json.loads(json_util.dumps(user_collection)) ^ this worked after installing python-bsonjs with pipenv install python-bsonjs
json_util comes from bson NOT from pymongo from my view
|
168

You should define you own JSONEncoder and using it:

import json
from bson import ObjectId

class JSONEncoder(json.JSONEncoder):
    def default(self, o):
        if isinstance(o, ObjectId):
            return str(o)
        return json.JSONEncoder.default(self, o)

JSONEncoder().encode(analytics)

It's also possible to use it in the following way.

json.encode(analytics, cls=JSONEncoder)

6 Comments

Perfect! It worked for me. I already have a Json encoder class, How can i merge that with yours class?My already Json encode class is: 'class MyJsonEncoder(json.JSONEncoder): def default(self, obj): if isinstance(obj, datetime): return str(obj.strftime("%Y-%m-%d %H:%M:%S")) return json.JSONEncoder.default(self, obj)'
@IrfanDayan, just add if isinstance(o, ObjectId): return str(o) before return in method default.
Could you add from bson import ObjectId, so everybody can copy-paste even faster? Thanks!
@defuz Why not just use str? What's wrong with that approach?
@defuz: When I try to use this, ObjectID is removed, but my json response is broken into single characters. I mean when I print each element from the resulting json in a for loop I get each character as an element. Any idea how to solve this?
|
66

Most users who receive the "not JSON serializable" error simply need to specify default=str when using json.dumps. For example:

json.dumps(my_obj, default=str)

This will force a conversion to str, preventing the error. Of course then look at the generated output to confirm that it is what you need.

3 Comments

This should be accepted answer! It cleanly solves the ObjectId issue. As @Asclepius says make sure it doesn't impact other json entries
I agree - it's both the simplest answer and has no (obvious) side effects.
Easy and clean implentation without the need to use bson, if theres a desire to omit oid, and create a 'flat' dictionary (with just the _id key and remaining attributes) . Even though this returns the exscaped quotes, passing the returned escaped JSON to json.loads converts it back to a Python dictionary
53
>>> from bson import Binary, Code
>>> from bson.json_util import dumps
>>> dumps([{'foo': [1, 2]},
...        {'bar': {'hello': 'world'}},
...        {'code': Code("function x() { return 1; }")},
...        {'bin': Binary("")}])
'[{"foo": [1, 2]}, {"bar": {"hello": "world"}}, {"code": {"$code": "function x() { return 1; }", "$scope": {}}}, {"bin": {"$binary": "AQIDBA==", "$type": "00"}}]'

Actual example from json_util.

Unlike Flask's jsonify, "dumps" will return a string, so it cannot be used as a 1:1 replacement of Flask's jsonify.

But this question shows that we can serialize using json_util.dumps(), convert back to dict using json.loads() and finally call Flask's jsonify on it.

Example (derived from previous question's answer):

from bson import json_util, ObjectId
import json

#Lets create some dummy document to prove it will work
page = {'foo': ObjectId(), 'bar': [ObjectId(), ObjectId()]}

#Dump loaded BSON to valid JSON string and reload it as dict
page_sanitized = json.loads(json_util.dumps(page))
return page_sanitized

This solution will convert ObjectId and others (ie Binary, Code, etc) to a string equivalent such as "$oid."

JSON output would look like this:

{
  "_id": {
    "$oid": "abc123"
  }
}

5 Comments

Just to clarify, no need to call 'jsonify' directly from a Flask request handler - just return the sanitized result.
You're absolutely right. A Python dict (which json.loads returns) should automatically be jsonified by Flask.
Isn't a dict object not callable?
@rick112358 how does a dict not being callable relate to this Q&A?
you can also use json_util.loads() to get the exact same dictionary back (instead of one with '$oid' key).
25
from bson import json_util
import json

@app.route('/')
def index():
    for _ in "collection_name".find():
        return json.dumps(i, indent=4, default=json_util.default)

This is the sample example for converting BSON into JSON object. You can try this.

Comments

17

As a quick replacement, you can change {'owner': objectid} to {'owner': str(objectid)}.

But defining your own JSONEncoder is a better solution, it depends on your requirements.

Comments

16

Posting here as I think it may be useful for people using Flask with pymongo. This is my current "best practice" setup for allowing flask to marshall pymongo bson data types.

mongoflask.py

from datetime import datetime, date

import isodate as iso
from bson import ObjectId
from flask.json import JSONEncoder
from werkzeug.routing import BaseConverter


class MongoJSONEncoder(JSONEncoder):
    def default(self, o):
        if isinstance(o, (datetime, date)):
            return iso.datetime_isoformat(o)
        if isinstance(o, ObjectId):
            return str(o)
        else:
            return super().default(o)


class ObjectIdConverter(BaseConverter):
    def to_python(self, value):
        return ObjectId(value)

    def to_url(self, value):
        return str(value)

app.py

from .mongoflask import MongoJSONEncoder, ObjectIdConverter

def create_app():
    app = Flask(__name__)
    app.json_encoder = MongoJSONEncoder
    app.url_map.converters['objectid'] = ObjectIdConverter

    # Client sends their string, we interpret it as an ObjectId
    @app.route('/users/<objectid:user_id>')
    def show_user(user_id):
        # setup not shown, pretend this gets us a pymongo db object
        db = get_db()

        # user_id is a bson.ObjectId ready to use with pymongo!
        result = db.users.find_one({'_id': user_id})

        # And jsonify returns normal looking json!
        # {"_id": "5b6b6959828619572d48a9da",
        #  "name": "Will",
        #  "birthday": "1990-03-17T00:00:00Z"}
        return jsonify(result)


    return app

Why do this instead of serving BSON or mongod extended JSON?

I think serving mongo special JSON puts a burden on client applications. Most client apps will not care using mongo objects in any complex way. If I serve extended json, now I have to use it server side, and the client side. ObjectId and Timestamp are easier to work with as strings and this keeps all this mongo marshalling madness quarantined to the server.

{
  "_id": "5b6b6959828619572d48a9da",
  "created_at": "2018-08-08T22:06:17Z"
}

I think this is less onerous to work with for most applications than.

{
  "_id": {"$oid": "5b6b6959828619572d48a9da"},
  "created_at": {"$date": 1533837843000}
}

Comments

11

For those who need to return the data thru Jsonify with Flask:

cursor = db.collection.find()
data = []
for doc in cursor:
    doc['_id'] = str(doc['_id']) # This does the trick!
    data.append(doc)
return jsonify(data)

Comments

9

You could try:

objectid = str(ObjectId("51948e86c25f4b1d1c0d303c"))

Comments

6

in my case I needed something like this:

class JsonEncoder():
    def encode(self, o):
        if '_id' in o:
            o['_id'] = str(o['_id'])
        return o

1 Comment

+1 Ha ! Could it have been more simpler 😍 Generally speaking; to avoid all the fuzz with custom encoders and bson importing, cast ObjectID to string: object['_id'] = str(object['_id'])
5

If you want to send it as a JSON response you need to format in two steps

  1. Using json_util.dumps() from bson to covert ObjectId in BSON response to JSON compatible format i.e. "_id": {"$oid": "123456789"}

The above JSON Response obtained from json_util.dumps() will have backslashes and quotes

  1. To remove backslashes and quotes use json.loads() from json
from bson import json_util
import json

bson_data = [{'_id': ObjectId('123456789'), 'field': 'somedata'},{'_id': ObjectId('123456781'), 'field': 'someMoredata'}]

json_data_with_backslashes = json_util.dumps(bson_data)

# output will look like this
# "[{\"_id\": {\"$oid\": \"123456789\"}, \"field\": \"somedata\"},{\"_id\": {\"$oid\": \"123456781\"}, \"field\": \"someMoredata\"}]"

json_data = json.loads(json_data_with_backslashes)

# output will look like this
# [{"_id": {"$oid": "123456789"},"field": "somedata"},{"_id": {"$oid": "123456781"},"field": "someMoredata"}]

Comments

4

This is how I've recently fixed the error

    @app.route('/')
    def home():
        docs = []
        for doc in db.person.find():
            doc.pop('_id') 
            docs.append(doc)
        return jsonify(docs)

1 Comment

in this case you are not passing '_id' attribute , instead just deleted '_id' and passed other attributes of doc
4

I know I'm posting late but thought it would help at least a few folks!

Both the examples mentioned by tim and defuz(which are top voted) works perfectly fine. However, there is a minute difference which could be significant at times.

  1. The following method adds one extra field which is redundant and may not be ideal in all the cases

Pymongo provides json_util - you can use that one instead to handle BSON types

Output: { "_id": { "$oid": "abc123" } }

  1. Where as the JsonEncoder class gives the same output in the string format as we need and we need to use json.loads(output) in addition. But it leads to

Output: { "_id": "abc123" }

Even though, the first method looks simple, both the method need very minimal effort.

1 Comment

this is very useful for the pytest-mongodb plugin when creating fixtures
4

I would like to provide an additional solution that improves the accepted answer. I have previously provided the answers in another thread here.

from flask import Flask
from flask.json import JSONEncoder

from bson import json_util

from . import resources

# define a custom encoder point to the json_util provided by pymongo (or its dependency bson)
class CustomJSONEncoder(JSONEncoder):
    def default(self, obj): return json_util.default(obj)

application = Flask(__name__)
application.json_encoder = CustomJSONEncoder

if __name__ == "__main__":
    application.run()

Comments

3

If you will not be needing the _id of the records I will recommend unsetting it when querying the DB which will enable you to print the returned records directly e.g

To unset the _id when querying and then print data in a loop you write something like this

records = mycollection.find(query, {'_id': 0}) #second argument {'_id':0} unsets the id from the query
for record in records:
    print(record)

Comments

2

Flask's jsonify provides security enhancement as described in JSON Security. If custom encoder is used with Flask, its better to consider the points discussed in the JSON Security

Comments

2

If you don't want _id in response, you can refactor your code something like this:

jsonResponse = getResponse(mock_data)
del jsonResponse['_id'] # removes '_id' from the final response
return jsonResponse

This will remove the TypeError: ObjectId('') is not JSON serializable error.

Comments

1
from bson.objectid import ObjectId
from core.services.db_connection import DbConnectionService

class DbExecutionService:
     def __init__(self):
        self.db = DbConnectionService()

     def list(self, collection, search):
        session = self.db.create_connection(collection)
        return list(map(lambda row: {i: str(row[i]) if isinstance(row[i], ObjectId) else row[i] for i in row}, session.find(search))

Comments

1
    @app.route('/users', methods = ['GET'])
       def get_users():
       online_users = mongo.db.users.find({})
       list_users = list(online_users)
       json_users = json.dumps(list_users, default=json_util.default)
       return Response(json_users, mimetype='application/json')

This is how I made it working

Comments

0

SOLUTION for: mongoengine + marshmallow

If you use mongoengine and marshamallow then this solution might be applicable for you.

Basically, I imported String field from marshmallow, and I overwritten default Schema id to be String encoded.

from marshmallow import Schema
from marshmallow.fields import String

class FrontendUserSchema(Schema):

    id = String()

    class Meta:
        fields = ("id", "email")

Comments

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.