django-s3file

A lightweight file upload input for Django and Amazon S3.

Django-S3File allows you to upload files directly AWS S3 effectively bypassing your application server. This allows you to avoid long running requests from large file uploads. This is particularly helpful for if you run your service on AWS Lambda or Heroku where you have a hard request limit.

Features

• lightweight: less 200 lines
• no JavaScript or Python dependencies (no jQuery)
• easy integration
• works just like the built-in
• extendable JavaScript API

For the Nerds

sequenceDiagram
    autonumber
    actor Browser
    participant S3
    participant Middleware
    Browser->>Django: GET form view
    activate Django
    Django->>Browser: RESPONSE w/ presigned POST URL & signed middleware key
    deactivate Django
    Browser->>S3: POST large file
    activate S3
    S3->>Browser: RESPONSE AWS S3 key
    Browser->>Middleware: POST AWS S3 key (signed)
    activate Middleware
    Middleware->>S3: GET AWS S3 key
    S3->>Middleware: RESPONSE large file promise
    deactivate S3
    Middleware->>Django: request incl. large file promise
    deactivate Middleware
    activate Django
    opt only if files is procssed by Django
        Django-->>S3: GET large file
        activate S3
        S3-->>Django: RESPONSE large file
        deactivate S3
    end
    Django->>Browser: RESPONSE success
    deactivate Django

Loading

In a nutshell, we can bypass Django completely and have AWS handle the upload or any processing. Of course, if you want to do something with your file in Django, you can do so, just like before, with the added advantage, that your file is served from within your datacenter.

Installation

Make sure you have Amazon S3 storage setup correctly.

Just install S3file using pip.

pip install django-s3file
# or
pipenv install django-s3file

Add the S3File app and middleware in your settings:

# settings.py

INSTALLED_APPS = (
    '...',
    's3file',
    '...',
)

MIDDLEWARE = (
    '...',
    's3file.middleware.S3FileMiddleware',
    '...',
)

Usage

S3File automatically replaces Django's ClearableFileInput widget, you do not need to alter your code at all.

The ClearableFileInput widget is only than automatically replaced when the STORAGES["default"] setting is set to django-storages' S3Boto3Storage or the dummy FileSystemStorage is enabled.

Setting up the AWS S3 bucket

Upload folder

S3File uploads to a single folder. Files are later moved by Django when they are saved to the upload_to location.

It is recommended to setup expiration for that folder, to ensure that old and unused file uploads don't add up and produce costs.

The default folder name is: tmp/s3file You can change it by changing the S3FILE_UPLOAD_PATH setting.

CORS policy

You will need to allow POST from all origins. Just add the following to your CORS policy.

[
  {
    "AllowedHeaders": [
        "*"
    ],
    "AllowedMethods": [
        "POST"
    ],
    "AllowedOrigins": [
        "*"
    ],
    "ExposeHeaders": [],
    "MaxAgeSeconds": 3000
  }
]

Using S3File in development

Using S3File in development can be helpful especially if you want to use the progress signals described above. Therefore, S3File comes with a AWS S3 dummy backend. It behaves similar to the real S3 storage backend. It is automatically enabled, if the STORAGES["default"] setting is set to FileSystemStorage.

To prevent users from accidentally using the FileSystemStorage and the insecure S3 dummy backend in production, there is also an additional deployment check that will error if you run Django's deployment check suite:

python manage.py check --deploy

We recommend always running the deployment check suite as part of your deployment pipeline.

Uploading multiple files

Django does have limited support for uploading multiple files. S3File fully supports this feature. The custom middleware ensures that files are accessible via request.FILES, even though they have been uploaded to AWS S3 directly and not to your Django application server.

Using optimized S3Boto3Storage

Since S3Boto3Storage supports storing data from any other fileobj, it uses a generalized _save function. This leads to the frontend uploading the file to S3 and then copying it byte-by-byte to perform a move operation just to rename the uploaded object. For large files this leads to additional loading times for the user.

That's why S3File provides an optimized version of this method at storages_optimized.S3OptimizedUploadStorage. It uses the more efficient copy method from S3, given that we know that we only copy from one S3 location to another.

from s3file.storages_optimized import S3OptimizedUploadStorage

class MyStorage(S3OptimizedUploadStorage):  # Subclass and use like any other storage
    default_acl = 'private'