Getting 'TypeError: ObjectId('') is not JSON serializable' when using Flask 0.10.1

前端 未结 4 1510
时光取名叫无心
时光取名叫无心 2021-02-03 10:43

I forked the Flask example, Minitwit, to work with MongoDB and it was working fine on Flask 0.9, but after upgrading to 0.10.1 I get the error in title when I login when I try t

相关标签:
4条回答
  • 2021-02-03 11:08

    EDIT: Even easier fix. You don't even need to do any JSON encoding/decoding.

    Just save the session['_id'] as a string:

    user = db.minitwit.user.find_one({'username': request.form['username']})
    session['_id'] = str(user['_id'])
    

    And then everywhere you want to do something with the session['_id'] you have to wrap it with ObjectId() so it's passed as a ObjectId object to MongoDB.

    if '_id' in session:
        g.user = db.minitwit.user.find_one({'_id': session['_id']})
    

    to:

    if '_id' in session:
        g.user = db.minitwit.user.find_one({'_id': ObjectId(session['_id'])})
    

    You can see the full diff for the fix on my github repo.

    If anyone cares to know why the 'TypeError: ObjectId('') is not JSON serializable' "issue" appeared in Flask 0.10.1, it's because they changed the way sessions are stored. They are now stored as JSON so since the '_id' object in MongoDB isn't standard JSON, it failed to serialize the session token, thus giving the TypeError. Read about the change here: http://flask.pocoo.org/docs/upgrading/#upgrading-to-010

    0 讨论(0)
  • 2021-02-03 11:08

    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)
    
    0 讨论(0)
  • 2021-02-03 11:18

    toString converts it to the string and that can stored to session:

    session['_id'] = user['_id'].toString()
    

    alternative session['_id'] = str(user['_id'])

    The above fixed the error for me.

    0 讨论(0)
  • 2021-02-03 11:22

    JSON only supports serializing (encoding/decoding) a limited set of objects types by default. You could extend python JSON's decoder/encoder's to handle this situation though.

    In terms of encoding an object which contains on ObjectID, for example, when ObjectIds are created client side, which will be passed along to some awaiting server, try:

    import json
    from bson.objectid import ObjectId
    
    class Encoder(json.JSONEncoder):
        def default(self, obj):
            if isinstance(obj, ObjectId):
                return str(obj)
            else:
                return obj
    

    Then, in your code, before pushing the data client -> server, run:

    json.dumps(obj, cls=Encoder)
    

    Server side, if we know we're dealing with mongo docs, (dictionary object with an '_id' key), we can define a json decoder hook like the following:

    def decoder(dct):
        for k, v in dct.items():
            if '_id' in dct:
                try:
                    dct['_id'] = ObjectId(dct['_id'])
                except:
                    pass
            return dct
    

    And call it using a call like the following:

    doc = json.loads(in_doc, object_hook=decoder)
    

    You'll probably need to adapt this code a bit, but for the simple case of passing

    0 讨论(0)
提交回复
热议问题