Currently our API doesn't have any restrictions on who can edit or delete code snippets. We'd like to have some more advanced behavior in order to make sure that:
We're going to make a couple of changes to our `Snippet` model class.
First, let's add a couple of fields. One of those fields will be used to represent the user who created the code snippet. The other field will be used to store the highlighted HTML representation of the code.
When that's all done we'll need to update our database tables.
Normally we'd create a database migration in order to do that, but for the purposes of this tutorial, let's just delete the database and start again.
rm tmp.db
python ./manage.py syncdb
You might also want to create a few different users, to use for testing the API. The quickest way to do this will be with the `createsuperuser` command.
python ./manage.py createsuperuser
## Adding endpoints for our User models
Now that we've got some users to work with, we'd better add representations of those users to our API. Creating a new serializer is easy:
Because `'snippets'` is a *reverse* relationship on the User model, it will not be included by default when using the `ModelSerializer` class, so we needed to add an explicit field for it.
We'll also add a couple of views. We'd like to just use read-only views for the user representations, so we'll use the `ListAPIView` and `RetrieveAPIView` generic class based views.
Right now, if we created a code snippet, there'd be no way of associating the user that created the snippet, with the snippet instance. The user isn't sent as part of the serialized representation, but is instead a property of the incoming request.
The way we deal with that is by overriding a `.pre_save()` method on our snippet views, that allows us to handle any information that is implicit in the incoming request or requested URL.
Now that snippets are associated with the user that created them, let's update our `SnippetSerializer` to reflect that. Add the following field to the serializer definition:
This field is doing something quite interesting. The `source` argument controls which attribute is used to populate a field, and can point at any attribute on the serialized instance. It can also take the dotted notation shown above, in which case it will traverse the given attributes, in a similar way as it is used with Django's template language.
The field we've added is the untyped `Field` class, in contrast to the other typed fields, such as `CharField`, `BooleanField` etc... The untyped `Field` is always read-only, and will be used for serialized representations, but will not be used for updating model instances when they are deserialized.
Now that code snippets are associated with users, we want to make sure that only authenticated users are able to create, update and delete code snippets.
REST framework includes a number of permission classes that we can use to restrict who can access a given view. In this case the one we're looking for is `IsAuthenticatedOrReadOnly`, which will ensure that authenticated requests get read-write access, and unauthenticated requests get read-only access.
If you open a browser and navigate to the browseable API at the moment, you'll find that you're no longer able to create new code snippets. In order to do so we'd need to be able to login as a user.
The `r'^api-auth/'` part of pattern can actually be whatever URL you want to use. The only restriction is that the included urls must use the `'rest_framework'` namespace.
Now if you open up the browser again and refresh the page you'll see a 'Login' link in the top right of the page. If you log in as one of the users you created earier, you'll be able to create code snippets again.
Once you've created a few code snippets, navigate to the '/users/' endpoint, and notice that the representation includes a list of the snippet pks that are associated with each user, in each user's 'snippets' field.
## Object level permissions
Really we'd like all code snippets to be visible to anyone, but also make sure that only the user that created a code snippet is able update or delete it.
To do that we're going to need to create a custom permission.
In the snippets app, create a new file, `permissions.py`
from rest_framework import permissions
class IsOwnerOrReadOnly(permissions.BasePermission):
"""
Custom permission to only allow owners of an object to edit it.
Make sure to also import the `IsOwnerOrReadOnly` class.
from snippets.permissions import IsOwnerOrReadOnly
Now, if you open a browser again, you find that the 'DELETE' and 'PUT' actions only appear on a snippet instance endpoint if you're logged in as the same user that created the code snippet.
Because we now have a set of permissions on the API, we need to authenticate our requests to it if we want to edit any snippets. We haven't set up any [authentication classes][authentication], so the defaults are currently applied, which are `SessionAuthentication` and `BasicAuthentication`.
When we interact with the API through the web browser, we can login, and the browser session will then provide the required authentication for the requests.
If we're interacting with the API programmatically we need to explicitly provide the authentication credentials on each request.
If we try to create a snippet without authenticating, we'll get an error:
curl -i -X POST http://127.0.0.1:8000/snippets/ -d "code=print 123"
{"detail": "Authentication credentials were not provided."}
We can make a successful request by including the username and password of one of the users we created earlier.
curl -X POST http://127.0.0.1:8000/snippets/ -d "code=print 789" -u tom:password
We've now got a fairly fine-grained set of permissions on our Web API, and end points for users of the system and for the code snippets that they have created.
In [part 5][tut-5] of the tutorial we'll look at how we can tie everything together by creating an HTML endpoint for our highlighted snippets, and improve the cohesion of our API by using hyperlinking for the relationships within the system.