Creating a production ready API with Python and Django Rest Framework – part 1

The aim if this tutorial is to show how to create a production ready solution for a REST API, using Python and Django Rest Framework. I will show you how to first create a very basic API, how to handle the authentication and permissions and I will cover deployment and hosting of images. The full source code of the tutorial is available at: https://github.com/andreagrandi/drf-tutorial

Summary of the complete tutorial

  1. Create the basic structure for the API
  2. Add Authentication and POST methods
  3. Testing the API
  4. Switching from Sqlite to PostgreSQL
  5. Hosting the API on Heroku
  6. Add an Image field and save images to S3

Create the basic structure for the API

For this tutorial I will assume you have correctly installed at least Python (I will use Python 2.7.x), virtualenv and virtualenvwrapper on your system and I will explain how to create everything else step by step.

Note: at the time of writing, the tutorial has been based on Django 1.10.1 and Django Rest Framework 3.4.7

Creating the main project structure

Data Model

We will create the API for a generic products catalog, using a very simple structure (to keep things simple). Edit the file catalog/models.py adding these lines:

after creating the model, we need to add ‘catalog’ application to INSTALLED_APPS. Edit settings.py and add the app at the end of the list:

at this point the Django application will be recognised by the project and we can create and apply the schema migration:

API Serializer

Serializers are those components used to convert the received data from JSON format to the relative Django model and viceversa. Create the new file catalog/serializers.py and place this code inside:

In this case we are using a ModelSerializer. We need to create a new class from it, and specify the model attribute, that’s it. In this case we also specify the fields we want to return.

API View

The serializer alone is not able to respond to an API request, that’s why we need to implement a view. In this first version of the view (that we will improve soon) we will “manually” transform the data available in the serializer dictionary to a JSON response. In catalog/views.py add this code:

At this point we need to tell our Django app to use this API view when a certain URL is requested. We first need to add this code in catalog/urls.py

and finally we need to add this to drftutorial/urls.py

Testing our work

At this point we should be able to start our Django app:

Let’s install a tool that will help us to test the API:

now we can use it to call our URL:

It works! It’s an empty response of course, because we still don’t have any data to show, but we will see later how to load some example data in our database. If you have been able to follow the tutorial up to this point, that’s good. If not, don’t worry. You can checkout the code at exactly this point of the tutorial doing:

Improving the API View

There is a quicker and more efficient way of implementing the same API view we have seen before. We can use a class based view, in particular the APIView class and also have the JSON response implemented automatically. Change the code inside catalog/views.py with this one:

You will also have to change catalog/urls.py in this way:

You can check the source code for this step of the tutorial with:

There is also another way of writing the same view. Let’s try it with ListAPIView. Edit catalog/views.py again and substitute the code with this one:

With a ListAPIView we are basically creating a read-only API that is supposed to return a list of items. We need to specify a queryset and the serializer_class parameters. Once again, you can get up to this point, checking out the related git tag:

Creating Initial Data

An API that doesn’t return any data is not very useful, right? Also, at the moment we haven’t implemented yet any feature that let us insert any data. That’s why I’ve created a data migration for you that will insert some data for you. You may notice that the example data contains some Italian products… out of the scope of this tutorial, I strongly advise you to google this products and if you ever happen to visit Italy, try them. You won’t regret!

Going back to our data migration, you first have to create an empty one with:

and then open the file that has been created under catalog/migrations/ named 0002_….. (your name will be different from mine, so just edit the one starting with 0002 and you will be fine) and fill it with this code:

to apply the migration we just created, just do:

If you try to test the API again from the command line, you will get these products back:

Again, you can get up to this point with:

One more thing…

No, we are not going to present a new amazing device, I’m sorry, but I want to show you a nice Django Rest Framework feature you can have without much additional work. Edit settings.py and add rest_framework to the list of INSTALLED_APPS:

Now, if you are still running the Django app, try to visit this url from your browser: http://127.0.0.1:8000/products/
That’s very nice, isn’t it? You can have browsable APIs at no cost.

Wrapping Up

I’ve mentioned at the beginning that this is just the first part of my API tutorial. In the next part I will show you how to let the consumer of your API add some products and leave reviews (we will introduce a new model for this). Also, we will see how to set proper permissions to these new API methods so that only admin users will be able to add products while normal users will be able to add reviews.

References

Some parts of this tutorial and a few examples have been taken directly from the original Django Rest Framework tutorial.

Using Python ipdb from Jupyter

If we try to use the usual ipdb commands from a Jupyter (IPython notebook)

we will get a similar error:

The solution is to use Tracer instead:

Source: http://stackoverflow.com/questions/35613249/using-ipdb-to-debug-python-code-in-one-cell-jupyter-or-ipython

How to publish a Python package to PyPI

PyPI is the Python Package Index, that archive that let you install a package using pip, for example: pip install Flask

In the past days I started writing a Python API client for Toshl expense manager and I decided to publish the library on PyPI. You can have a look at my library here https://github.com/andreagrandi/toshl-python (please note: it’s still in development and Toshl API is not even public yet) in case you are not sure how to structure it.

I found a nice guide but it wasn’t complete (for example it didn’t say how to sign packages) so I decided to rewrite it adding more information.

Create PyPI accounts

To publish packages on PyPI you need to create two accounts: one for the production server and another one for the test server. When you register, please specify (if you have one, but I really hope you do) the PGP id of your public key. Once the accounts are created, you need to create a file named .pypirc in your $HOME folder containing the following configuration:

Please substitute your_username and your_password with the details you sent during the registration.

Preparing the package

I assume you have structured your library in the proper way and have included a setup.py with all the configuration (it’s not something specific to PyPI so you should have done it already). If you haven’t I remember you can give a look at my library here https://github.com/andreagrandi/toshl-python in particular to the setup.py:

Upload the package to PyPI Test server

The first time you upload the package you will need to register it:

and then you will need to build the package and upload it (please note I’m using the –sign to sign the package with PGP):

Upload the package to PyPI production server

Once you have verified that you are able to build and upload the package to the test server (without getting any errors), you should upload it to the production server:

This is everything you need to do if you want to publish a Python package on PyPI. Happy coding!

Using a light sensor with BBC micro:bit and MicroPython

A light sensor is a small component with a particular characteristic: it is basically a resistor and its resistance decreases if the light is more intense. To use it with micro:bit we need to use one of the analogic ports. To build this circuit you will need a breadboard, 3 jumper wires, a 10k resistance and possibly a Kitronik breadboard kit.

The project

I wanted to realise a simple project where, depending on the light intensity captured by the light sensor, the micro:bit shows an image of the Sun if the light is intense and an image of the Moon if the light is less intense.

Here is the complete circuit scheme:

microbit_breadboard_schema_light

“Image Copyright © Kitronik”

 

and here is a picture of the finished project I created:

microbit_breadboard_example_2

The source code I needed is available here:

and as a demo I realised this small video:

Prototyping BBC micro:bit projects with Kitronik breadboard kit

BBC micro:bit has a few IO pins that can be used to interact with external devices. The problem with the board is that it’s not easy to connect the classic jumper wires (those that we normally connect to a breadboard) to the micro:bit, unless using a crocodile clip and being limited to just 3 pins.

Kitronik breadboard kit solves this problem, offering an interface where the micro:bit can be plugged and all the pins are easily connectable to the breadboard using normal male/female jumper wires.

I’ve built a very simple circuit following an example you can find on this manual https://www.kitronik.co.uk/pdf/5603_inventors_kit_for_the_bbc_microbit_tutorial_book.pdf

microbit_breadboard_example_1

To build the circuit you also need 4 male/female jumper wires and two buttons. All this circuit does is to connect the buttons to the micro:bit pins that relate to those buttons. Basically pressing those buttons is the same as pressing button A or button B on the micro:bit board. Here you can see the schema in detail:

Screenshot 2016-02-07 14.30.36

“Image Copyright © Kitronik”

I’ve also made a short video so that you can see it in action:

And of course the source code is available too: