django-async-gt 0.6.2

async_image_save

As you cannot pass as image to a Celery task, this app deconstructs the UploadedFile instance, passes it to a Celery task and reconstructs it there. Finally saving the uploaded image.

License

All code is under a BSD-style license, see LICENSE for details.

Source: https://github.com/gterzian/django_async.git

Requirements

• python >= 2.7

• django >= 1.4

• Celery >= 3.0

• PIL >= 1.1.6

Installation

To install run:

pip install django-async-gt

Configuration

You first of all need to have django-celery set up in your project.

'async_image_save'

The Problem: Celery tasks cannot accept images as arguments

Django uploads images and other files in the form of UpLoadedFile objects. UpLoadedFile is the abstract baseclasse, while TemporaryUploadedFile and InMemoryUploadedFile are the built-in concrete subclasses. An UploadedFile object behaves somewhat like a file object and represents some file data that the user submitted with a form.

The uploadedfile is received in the view as part of request.FILES, which you will usually bind to a form. Running the form’s is_valid method will then validate this file, or in the case of an ImageField whether the fiel is an actual image, and return the UpLoadedFile object for you to use, as either a TemporaryUploadedFile or InMemoryUploadedFile.

Once Django form validation has been succesfully run, you can safely assume that you are dealing with an actual image. The normal course of business is to immediatly bind the uploaded file object to a model instance, and saving the instance to the database. This means saving the image to your data store while the client is still waiting for a response from the server, which takes a few seconds, especially if you are using S3.

Wouldn’t it be better to pass the image along to a celery worker to save in the background?

The problem is that Celery needs to be able to pickle objects to pass them along to workers, and it cannot pickle a file like object such as an image.

The Solution: de- and reconstruct the image

The remaining option is to deconstruct the file object, write it’s data into a string and taking all the other info that you need. This data can be pickled and therefore passed on to Celery. You then simply need to reconstruct an actual TemporaryUploadedFile or InMemoryUploadedFile object on the other end, and bind this object to an instance of a model, by passing the id of that instance along with all other raw file ‘data’. You can then finally save the model instance.

Read the Django docs

UploadedFile: https://docs.djangoproject.com/en/1.4/topics/http/file-uploads/#django.core.files.uploadedfile.UploadedFile

Binding data to forms: https://docs.djangoproject.com/en/1.4/ref/forms/api/#binding-uploaded-files

How to use it in your project

The functionalities of this app reside in the save_image function, to be used in your views like the below. The function deconstructs the image and sends the data to the async_save task. Please not that the task clears the cache. This is necessary in my opinion because just adding the image to the instance doesn’t invalidate the cache like creating a new instance would, therefore the saved image is not seen by the user until after the cache time out. Therefore clearing the cache in the task seems necessary.

from async_image_save.utils import save_image

def example_view(request):
    if request.method == "POST":
        form = YourModelForm(request.POST, request.FILES)
        if form.is_valid():
            # assuming your model has a main_photo ImageField
            # save the instance without an image
            instance = form.save(commit=False)
            instance.main_photo = None
            instance.save()
            instance.users.add(request.user.userprofile)
            # send the image to be saved by a worker
            save_image(form.cleaned_data['main_photo'], instance)

            return HttpResponseRedirect(reverse('home'))
    else:
        form = YourModelForm()
    context['form'] = form
    return render_to_response("home.html", context, context_instance=RequestContext(request))