django-pb-model 0.1.8

Protobuf mixin for Django model

Django-pb-model provides model mixin mapping/converting protobuf message. Automatic model generation from protobuf message definitions is supported. Currently support basic value fields and naive relation conversion, including:

• Integer, String, Float, Boolean

• Choices field

• Datetime

• Foreign Key and Many-to-Many relation

• Custom fields, ex: JSON

You could examine testcases for more details

And PRs are always welcome :))

Compatibility

Currently tested with metrics:

• Python2.7, 3.4, 3.5, 3.6, 3.7

• Django1.11, 2.0, 2.1, 2.2

Install

1. pip install

pip install django-pb-model

2. Add django-pb to django settings.py

INSTALLED_APPS = [
    ....,
    pb_model,
    ...
]

3. Run python/django essential commands:

python manage.py makemigrations
python manage.py migrate
python manage.py collectstatic -l

4. Start hacking or using app.

Usage

Declare your protobuf message file, such as account.proto, and compile it. For example:

message Account {
    int id = 1;
    string email = 2;
    string nickname = 3;
}

Then compile it with:

$ protoc --python_out=. account.proto

You will get account_pb2.py.

Now you can interact with your protobuf model, add ProtoBufMixin to your model like:

from django.db import models
from pb_model.models import ProtoBufMixin
from . import account_pb2

class Account(ProtoBufMixin, models.Model):
    pb_model = account_pb2.Account

    email = models.EmailField(max_length=64)
    nickname = models.CharField(max_length=64)

    def __str__(self):
        return "Username: {a.email}, nickname: {a.nickname}".format(a=self)

By above settings, you can covert between django model and protobuf easily. For example:

>>> account = Account.objects.create(email='[e-mail]', nickname='moonmoon')
>>> account.to_pb()
email: "[e-mail]"
nickname: "moonmoon"

>>> account2 = Account()
>>> account2.from_pb(account.to_pb())
<Account: Username: username@mail, nickname: moonmoon>

Automatic field generation

To automatically generate django model fields based on protobuf field types.

If you don’t want to manually specify fields in your django model, you can list names of desired fields under pb_2_dj_fields attribute to have those generated and added to your model automatically.

class Account(ProtoBufMixin, models.Model):
    pb_model = account_pb2.Account
    pb_2_dj_fields = ['email', 'nickname']

Alternatively if you want all protobuf fields to be mapped you can do pb_2_dj_fields = '__all__'.

Fields listed in pb_2_dj_fields can be overwritten using manuall definition.

class Account(ProtoBufMixin, models.Model):
    pb_model = account_pb2.Account
    pb_2_dj_fields = '__all__'

    email = models.EmailField(max_length=64)

Type of generated field depends on corresponding protobuf field type. If you want to change default field type mappings you can overwrite those using pb_auto_field_type_mapping attribute.

Following protobuf field types are supported:

• uint32, int32, uint64, int64, float, double, bool, Enum

• string, bytes

• google.protobuf.Timestamp

• Messages

• oneof fields

• repeated scalar and Message fields

• map fields with scalar as key and scalar or Message as value

Field details

There are several special field types while converting, read following sections.

Field name mapping

To adapt schema migration, field mapping are expected.

For example, the email field in previous session is altered to username, but we don’t want to break the consistence of protobuf protocol. You may add pb_2_dj_field_map attribute to solve this problem. Such as:

class Account(ProtoBufMixin, models.Model):
    pb_model = account_pb2.Account
    pb_2_dj_field_map = {
        "email": "username",  # protobuf field as key and django field as value
    }

    username = models.CharField(max_length=64)
    nickname = models.CharField(max_length=64)

Foreign Key

Foreign key is a connect to another model in Django. According to this property, the foreign key could and should be converted to nested singular message in Protobuf. For example:

message Relation {
    int32 id = 1;
}

message Main {
    int32 id = 1;
    Relation fk = 2;
}

Django model:

class Relation(ProtoBufMixin, models.Model):
    pb_model = models_pb2.Relation


class Main(ProtoBufMixin, models.Model):
    pb_model = models_pb2.Main

    fk = models.ForeignKey(Relation)

With above settings, pb_model would recursively serialize and de-serialize between Django and ProtoBuf.

>>> m = Main.objects.create(fk=Relation.objects.create())
>>> m.to_pb()
id: 1
fk {
    id: 1
}

>>> m2 = Main()
>>> m2.from_pb(m.to_pb())
>>> m2.fk.id
1

Many-to-Many field

M2M field is a QuerySet Relation in Django. By default, we assume target message field is “repeated” nested message, ex:

message M2M {
    int32 id = 1;
}

message Main {
    int32 id = 1;

    repeated M2M m2m = 2;
}

Django model would be:

class M2M(models.Model):
    pass

class Main(models.Model):

    m2m = models.ManyToManyField(M2M)

Django to Protobuf

If this is not the format you expected, overwrite _m2m_to_protobuf() of Django model by yourself.

Protobuf to Django

Same as previous section, we assume m2m field is repeated value in protobuf. By default, NO operation is performed, which means you may query current relation if your converted django model instance has a valid primary key.

If you want to modify your database while converting on-the-fly, overwrite logics such as:

from django.db import transaction

...

class PBCompatibleModel(ProtoBufMixin, models.Model):

    def _repeated_to_m2m(self, dj_field, _pb_repeated_set):
        with transaction.atomic():
            for item in _pb_repeated_set:
                dj_field.get_or_create(pk=item.pk, defaults={....})

    ...

Also, you should write your converting policy if m2m is not nested repeated message in _repeated_to_m2m method

Datetime Field

Datetime is a special singular value.

We currently convert between datetime.datetime (Python) and google.protobuf.timestamp_pb2.Timestamp (ProboBuf), for example:

ProtoBuf message:

package models;

import "google/protobuf/timestamp.proto";

message WithDatetime {
    int32 id = 1;
    google.protobuf.Timestamp datetime_field = 2;
}

Django Model:

class WithDatetime(ProtoBufMixin, models.Model):
    pb_model = models_pb2.WithDatetime

    datetime_field = models.DatetimeField(default=timezone.now())

>>> WithDatetime.objects.create().to_pb()
datetime_field {
seconds: 1495119614
nanos: 282705000
}

Custom Fields

You can write your own field serializers, to convert between django.contrib.postgres.fields.JSONField (Python) and string (Protobuf) for example:

ProtoBuf message:

package models;

message WithJSONBlob {
    int32 id = 1;
    string json_blob = 2;
}

Django Model:

def json_serializer(pb_obj, pb_field, dj_value):
    setattr(pb_obj, pb_field.name, json.dumps(value))

def json_deserializer(instance, dj_field_name, pb_field, pb_value):
    setattr(instance, dj_field_name, json.loads(pb_value))

class WithJSONField(ProtoBufMixin, models.Model):
    pb_model = models_pb2.WithJSONBlob

    pb_2_dj_field_serializers = {
        'JSONField': (json_serializer, json_deserializer),
    }

    json_field = models.JSONField()

Timezone

Note that if you use USE_TZ in Django settings, all datetime would be converted to UTC timezone while storing in protobuf message. And coverted to default timezone in django according to settings.