RESTful interface in Flask and issues serializing - json

I'm learning Backbone.js and Flask (and Flask-sqlalchemy). I chose Flask because I read that it plays well with Backbone implementing RESTful interfaces. I'm currently following a course that uses (more or less) this model:
class Tasks(db.Model):
id = db.Column(db.Integer, primary_key=True)
title = db.Column(db.String(80), unique=True)
completed = db.Column(db.Boolean, unique=False, default=False)
def __init__(self, title, completed):
self.title = title
self.completed = completed
def json_dump(self):
return dict(title=self.title, completed=self.completed)
def __repr__(self):
return '<Task %r>' % self.title
I had to add a json_dump method in order to send JSON to the browser. Otherwise, I would get errors like object is not JSON serializable, so my first question is:
Is there a better way to do serialization in Flask? It seems that some objects are serializable but others aren't, but in general, it's not as easy as I expected.
After a while, I ended up with the following views to take care of each type of request:
#app.route('/tasks')
def tasks():
tasks = Tasks.query.all()
serialized = json.dumps([c.json_dump() for c in tasks])
return serialized
#app.route('/tasks/<id>', methods=['GET'])
def get_task(id):
tasks = Tasks.query.get(int(id))
serialized = json.dumps(tasks.json_dump())
return serialized
#app.route('/tasks/<id>', methods=['PUT'])
def put_task(id):
task = Tasks.query.get(int(id))
task.title = request.json['title']
task.completed = request.json['completed']
db.session.add(task)
db.session.commit()
serialized = json.dumps(task.json_dump())
return serialized
#app.route('/tasks/<id>', methods=['DELETE'])
def delete_task(id):
task = Tasks.query.get(int(id))
db.session.delete(task)
db.session.commit()
serialized = json.dumps(task.json_dump())
return serialized
#app.route('/tasks', methods=['POST'])
def post_task():
task = Tasks(request.json['title'], request.json['completed'])
db.session.add(task)
db.session.commit()
serialized = json.dumps(task.json_dump())
return serialized
In my opinion, it seems a bit verbose. Again, what is the proper way to implement them? I have seen some extensions that offer RESTful interfaces in Flask but those look quite complex to me.
Thanks

I would use a module to do this, honestly. We've used Flask-Restless for some APIs, you might take a look at that:
https://flask-restless.readthedocs.org/en/latest/
However, if you want build your own, you can use SQLAlchemy's introspection to output your objects as key/value pairs.
http://docs.sqlalchemy.org/en/rel_0_7/core/schema.html#metadata-reflection
Something like this, although I always have to triple-check I got the syntax right, so take this as a guide more than working code.
#app.route('/tasks')
def tasks():
tasks = Tasks.query.all()
output = []
for task in tasks:
row = {}
for field in Tasks.__table__.c:
row[str(field)] = getattr(task, field, None)
output.append(row)
return jsonify(data=output)
I found this question which might help you more. I'm familiar with SQLAlchemy 0.7 and it looks like 0.8 added some nicer introspection techniques:
SQLAlchemy introspection

Flask provides jsonify function to do this. Check out its working here.
Your json_dump method is right though code can be made concise. See this code snippet
#app.route('/tasks')
def tasks():
tasks = Tasks.query.all()
return jsonify(data=[c.json_dump() for c in tasks])

Related

django rest framework + mariaDB: custom JSONField

Django: v2.1.5
DRF: v3.9.1
mariaDB: v10.3
Hi, I am a DRF newbie and I have been struggling with json field.
DRF does not support official json field type working with mariaDB and even though there is a 3rd-party package for mysql(django-mysql) but not compatible with mariaDB.
So I searched and started implementing custom jsonfield and it looks like:
model.py:
class JSONField(models.TextField):
def to_dict(self, value):
""" convert json string to python dictionary """
return json.loads(value)
def to_json(self, value):
""" convert python dictionary to json string """
return json.dumps(value)
def from_db_value(self, value, expression, connection):
""" convert string from db to python dictionary """
if value is None:
return value
return self.to_dict(value)
def to_python(self, value):
""" convert model input value to python dictionary """
if isinstance(value, dict):
return value
if value is None:
return value
return self.to_dict(value)
def get_prep_value(self, value):
""" convert python dictionary to string before writing to db """
return self.to_json(value)
class Project(models.Model):
objects = models.Manager()
project_user = JSONField(null=True)....
serializers.py:
class ProjectSerializer(serializers.ModelSerializer):
class Meta:
model = Project
fields = ('project_code'...)
def create(self, validated_data):
"""
Create and return a new `Project` instance, given the validated data.
"""
return Project.objects.create(**validated_data)
views.py:
class ListCreateProjectView(APIView):
"""
POST admin/_proj_/
: create a project
"""
def post(self, request, format=None):
serializer = ProjectSerializer(data=request.data)
if serializer.is_valid():
serializer.save()
return Response(data=serializer.data)
else:
return Response(data=serializer.errors)
please let me know what I am doing wrong here or a way to use 3rd party package rather than custom jsonfield
Thanks a lot and have a great day guys!
For people who suffer from this kind of problem, I actually solved this problem, but in an informal way. I set JSONfield as a textfield in DB, and handle (str->json),(json->str) in the view. But again, this is an informal way (I think) and you will need another solution than this. If you find one, please share it for me and other people :)

Django REST bulk post / post array of JSON objects

I have started to play around with the Django REST framework. So far I succeeded in creating a serializer for my object, creating the post view, post objects and return objects via Javascript's $.post(). So right now I have a proper conversion between my JSONs and Django model objects.
The problem is that I have an array of objects [A1, A2, ..., An]. Right now when I need to post such an array I do it object by object. Is there any possibility to post the whole array at once, and recover an array of objects inside my Django View? If so, what is the pattern to follow here? I guess I could define a new model which is an array of my current model, create a serializer for it, etc., but this does not seem too elegant.
Below are my view and serializer:
#serializers.py
class SearchRequestSerializer(serializers.ModelSerializer):
def create(self):
return SearchRequest(**self.validated_data)
class Meta:
model = SearchRequest
#views.py
#api_view(['POST'])
def post_calculation(request):
if request.method == 'POST':
#JSON to serializer object
serializer = SearchRequestSerializer(data=request.data, many=False)
if (serializer.is_valid() == False):
return Response(serializer.errors, status=status.HTTP_400_BAD_REQUEST)
#create the Python object
search_request = serializer.create()
#-- I do some processing stuff with the search_request object here ---
#object back to JSON
serializer3 = SearchRequestSerializer(search_request, many=False)
return Response(serializer3.data)
There are two solutions to your problem:
The first solution is to override the .create() method of your view
By default, django rest framework assumes you are passing it a single object. To account for the possibility to pass it a list of objects you might rewrite it as follows:
def create(self, request, pk=None, company_pk=None, project_pk=None):
is_many = isinstance(request.data, list)
serializer = self.get_serializer(data=request.data, many=is_many)
serializer.is_valid(raise_exception=True)
self.perform_create(serializer)
headers = self.get_success_headers(serializer.data)
return Response(serializer.data, status=status.HTTP_201_CREATED, headers=headers)
Now your view will handle both single objects in POST data as well as a list of objects.
The second solution is to use a third party package
django-rest-framework-bulk provides the above functionality plus additional features (e.g. bulk update). You can check it out and decide whether it fits your needs.
Update: The solution for function based views
In order to get it to work with your function based view, the approach is similar:
#api_view(['POST'])
def post_calculation(request):
if request.method == 'POST':
is_many = isinstance(request.data, list)
# JSON to serializer object
serializer = SearchRequestSerializer(data=request.data, many=is_many)
if not serializer.is_valid():
return Response(serializer.errors, status=status.HTTP_400_BAD_REQUEST)
serializer.save()
#-- do some processing stuff here ---
return Response(serializer.data)
Do not call .create() method directly, use .save() instead. Also, when using many=False on serializer, the created instance is available under serializer.instance. I am not sure how to obtain the list of created instances though. You can try the same serializer.instance. If it doesn't work, try to find how to get it.

Render server side JSON in Django REST Framework

I'm trying to make bulk data downloads by serializing my entire database as JSON. The drf documentation on serializers has a section that says you can simply do:
from rest_framework.renderers import JSONRenderer
serializer = CommentSerializer(comment)
json = JSONRenderer().render(serializer.data)
Unfortunately, this doesn't work for HyperLinked relationships. When you try to do it with them, you get something like:
AssertionError: HyperlinkedIdentityField requires the request in the serializer context. Add context={'request': request} when instantiating the serializer.
So, I figured out I can add context attribute, like:
r = Request(request=HttpRequest())
context = dict(request=r)
serializer = CommentSerializer(comment, context=context)
json = JSONRenderer().render(serializer.data)
Which then returns the error:
django.core.exceptions.ImproperlyConfigured: Could not resolve URL for hyperlinked relationship using view name "opinioncluster-detail". You may have failed to include the related model in your API, or incorrectly configured the lookup_field attribute on this field.
I know this API works properly when it's called from the browser, but I can't get past this when I call it as above. Something that's automatic from the browser doesn't happen when you render it manually.
Any ideas?
First edit
Here's another strategy that seemed promising because it would add the path to my request object:
r = Request(request=RequestFactory().get(reverse('comment-list', kwargs={'version': 'v3'})))
context = dict(request=r)
serializer = CommentSerializer(comment, context=context)
json = JSONRenderer().render(serializer.data)
That returns the same problem as if I hadn't defined a path to the request:
django.core.exceptions.ImproperlyConfigured: Could not resolve URL for hyperlinked relationship using view name "opinioncluster-detail". You may have failed to include the related model in your API, or incorrectly configured the lookup_field attribute on this field.
Second Edit
As I said in my comments, I'm fairly certain my serializers and views aren't to blame, since they work perfectly fine via the browser. Nevertheless, here they are. If you're truly generous, the full serializers, filters, and codebase is online.
View:
class OpinionClusterViewSet(LoggingMixin, viewsets.ModelViewSet):
queryset = OpinionCluster.objects.all()
serializer_class = OpinionClusterSerializer
filter_class = OpinionClusterFilter
ordering_fields = (
'date_created', 'date_modified', 'date_filed', 'citation_count',
'date_blocked',
)
Serializer:
class OpinionClusterSerializer(DynamicFieldsModelSerializer,
serializers.HyperlinkedModelSerializer):
absolute_url = serializers.CharField(source='get_absolute_url',
read_only=True)
panel = serializers.HyperlinkedRelatedField(
many=True,
view_name='judge-detail',
read_only=True,
)
non_participating_judges = serializers.HyperlinkedRelatedField(
many=True,
view_name='judge-detail',
read_only=True,
)
docket = serializers.HyperlinkedRelatedField(
many=False,
view_name='docket-detail',
read_only=True,
)
sub_opinions = serializers.HyperlinkedRelatedField(
many=True,
view_name='opinion-detail',
read_only=True,
)
class Meta:
model = OpinionCluster
This management command creates a MockRequest to be used in a non-browser environment and should allow you to create your JSON:
from django.core.management.base import BaseCommand
from django.http import HttpRequest
from rest_framework.renderers import JSONRenderer
from rest_framework.request import Request
from comments.models import Comment
from comments.serializers import CommentSerializer
class MockRequest(HttpRequest):
def __init__(self):
super(MockRequest, self).__init__()
self.setup_host()
def setup_host(self):
# Required to give absolute urls in output
self.META['HTTP_HOST'] = 'localhost:8000'
class Command(BaseCommand):
help = 'Export JSON'
def handle(self, *args, **options):
request = MockRequest()
serializer_context = {
'request': Request(request),
}
comment = Comment.objects.first()
serializer = CommentSerializer(comment, context=serializer_context)
json = JSONRenderer().render(serializer.data)
print(json)
To remove the hard coded domain you could use the Sites framework to power the host name:
def setup_host(self):
from django.contrib.sites.models import Site
site = Site.objects.get_current()
self.META['HTTP_HOST'] = site.domain
After digging quite deeply into the code, I finally figured this out:
r = RequestFactory().request()
r.version = 'v3'
r.versioning_scheme = URLPathVersioning()
context = dict(request=r)
renderer = JSONRenderer()
json_str = renderer.render(
serializer(item, context=context).data,
accepted_media_type='application/json; indent=2',
)
This seems to work, but the HyperLinkRelated values in the JSON that is serialized have the server set to testserver. I could get around that by setting:
r.META['SERVER_NAME']
But I also need to set r.scheme to https, which doesn't seem to be possible (I get an error that r.scheme cannot be set).
I'm pretty close though, so this is going to have to serve as an answer for now.

Using Django Rest Framework > 2.4 to get children of parent with #detail_route

I had this working great less than DRF 2.4, but the change to 2.4+ uses #detail_route.
When I do a GET to /api/parent/7/children I expect to get all the children that belong to Parent 7.
But I'm getting an empty array.
Here's my code:
class ParentViewSet(viewsets.ModelViewSet):
serializer_class = ParentSerializer
model = models.Parent
#detail_route()
def children(self, request, pk):
parent = self.get_object()
children = parent.children.all()
serializer = ChildrenSerializer(children)
return Response(serializer.data)
def get_queryset(self):
if self.request.user.is_superuser:
return models.Parent.objects.all()
else:
return models.Parent.objects.filter(user=self.request.user)
def pre_save(self, obj):
obj.user = self.request.user
When I go to the endpoint /api/parent/7/children in the API viewer the response I get is:
{
"detail": "Not found"
}
Any suggestions?
I like this better in principle, you can be much more declarative about what is happening. I know I can set methods on children so I don't have to do a #link and #action for the same resource. Just need to get past this hurdle.
Thanks!
I don't know if this is your issue or not, but you have a typo in the example as presented...
children = parent.children.all()
serializer = ChildrenSerializer(Children)
Should be:
children = parent.children.all()
serializer = ChildrenSerializer(children)
You're passing the Children class to be serialized, not the children instance.

Django REST: ContentNotRenderedError when checking JSON object before rendering

I have a Django model, Note, which has a class-based view. It is supposed to return a JSON object upon the appropriate query.
Before returning the object, however, I would like to check that the user field in the note object matches the user currently logged in. (Users should not be able to access Note objects that are not their own.) To do this, I tried rewriting the get() method, calling on self.retrieve() to inspect the object before returning it:
class NoteDetail(generics.RetrieveUpdateDestroyAPIView):
model = Note
serializer_class = NoteSerializer
permission_classes = (permissions.IsAuthenticated,)
renderer_classes = (JSONRenderer,)
def get(self, request, *args, **kwargs):
current_user = User.objects.get(pk=self.request.user.id)
note = self.retrieve(request, *args, **kwargs)
if note.author is current_user:
return note
else:
raise PermissionDenied('Note does not belong to authenticated user.')(author=current_user)
However, this returns a ContentNotRenderedError when run: The response content must be rendered before it can be accessed.
Is there a way for me to check the object before returning it? Must I find a workaround?
One potential workaround is to redefine get_queryset(), rather than get() or get_object(). get_queryset() is the function that returns all objects of the relevant model; get_object() narrows down among these given the argument pk.
By overriding get_queryset(), you restrict the potential objects that get_object() can pick. Thus the set is already filtered at the time get() is called.
def get_queryset(self):
current_user = User.objects.get(pk=self.request.user.id)
# Must filter by author to prevent making everyone's notes public
queryset = Note.objects.filter(author=current_user)
return queryset