CODEWITHSUNDEEP
jsonpython-2.7google-app-enginegoogle-cloud-datastoreapp-engine-ndb
jvaldenor
jvaldenor

Reputation: 67

App engine datastore to_dict alternative to serialize ndb.Model to JSON

I have a ndb.Model that I want to convert to JSON.

class Users(ndb.Model):
    username = ndb.StringProperty(indexed=True)
    password= ndb.StringProperty(indexed=True)
    created_at = ndb.DateTimeProperty(auto_now_add=True)

user = Users.query(Users.username==username).get()
rv = json.dumps(user.to_dict()) 
print(rv)

It throws this error: 

 TypeError: datetime.datetime(2013, 11, 24, 3, 40, 15) is not JSON serializable

Most of the solutions here are for db.Model and are fairly outdated.

sdk version 1.9.10

Upvotes: 3

Views: 2622

Answers (3)

Maxim Mercuriev
Maxim Mercuriev

Reputation: 456

You can extend Property classes to handle special cases. It is applicable for any property type.

from google.appengine.ext import ndb

class DateTimeProperty(ndb.DateTimeProperty):
    # Override to allow JSON serialization
    def _get_for_dict(self, entity):
        value = super(DateTimeProperty, self)._get_for_dict(entity);
        return value.isoformat()

Then use it in your model:

class Users(ndb.Model):
    username = ndb.StringProperty(indexed=True)
    password= ndb.StringProperty(indexed=True)
    created_at = DateTimeProperty(auto_now_add=True)

And to_dict() as you would normally do:

user = Users.query(Users.username==username).get()
user.to_dict()

Upvotes: 10

turdus-merula
turdus-merula

Reputation: 8854

Another possible solution - inspired from this answer - is to override to_dict() in your User class:

class User(ndb.Model):
    username = ndb.StringProperty(indexed=True)
    password = ndb.StringProperty(indexed=False)
    created_at = DateTimeProperty(auto_now_add=True)

    def to_dict(self):
        result = super(User, self).to_dict()
        result['created_at'] = self.created_at.isoformat()
        return result

Upvotes: 1

Mihail Russu
Mihail Russu

Reputation: 2536

You would need a custom "to JSON" converter that handles formats not natively supported by JSON.

I am using something like the following code that handles most situations for me.

def to_json(self, o):
    if isinstance(o, list):
        return [self.to_json(l) for l in o]
    if isinstance(o, dict):
        x = {}
        for l in o:
            x[l] = self.to_json(o[l])
        return x
    if isinstance(o, datetime.datetime):
        return o.isoformat()
    if isinstance(o, ndb.GeoPt):
        return {'lat': o.lat, 'lon': o.lon}
    if isinstance(o, ndb.Key):
        return o.urlsafe()
    if isinstance(o, ndb.Model):
        dct = o.to_dict()
        dct['id'] = o.key.id()
        return self.to_json(dct)
    return o

So in my case I am also taking care of some other things like GeoPt, and adding an ID field to all ndb.Models but for your case all you'd need would be:

    if isinstance(o, datetime.datetime):
        return o.isoformat()

but I am guessing (not really sure) you would then get a key error as well so you'd also need

    if isinstance(o, ndb.Key):
        return o.urlsafe()

In case if you didn't need the created_at field, you could simply exclude it like

rv = json.dumps(user.to_dict(exclude=['created_at']))

Upvotes: 6

Related Questions