We will now write the (possibly) simplest GeoDjango app ever. It defines a model with a ``PointField`` and a name. Think of it as a way to record famous landmarks. Of course, GeoDjango also has other types, allowing to use lines, polygons, as well as collections of those basic types. Available fields and their options are described in the documentation at https://docs.djangoproject.com/en/dev/ref/contrib/gis/model-api/. Also, we will allow this app to show up in the Django admin site. GeoDjango provides a special geographic admin, that displays geographic fields on a map, and allows you to edit them, just like you would edit a regular text or numeric field. .. At that point, your first GeoDjango app is ready: push it to DotCloud! At the end of the push, you will see the application URL. Add "/admin" at the end of the URL, login with "admin" and "password", and try to add something in the "Places" table!
To turn Django into GeoDjango, you only need two things: * use a geographic-capable ORM engine; * enable the ``django.contrib.gis`` application. Since our database is PostGIS, we will use the relevant ORM engine. .. As you can see, those two steps are easy as pie!
The first step is to edit ``dotcloud.yml``, and replace our plain PostgreSQL with PostGIS. Technically, PostGIS is just PostgreSQL with additional geographic extensions. It provides new SQL types, functions, and also a bunch of tables that are used by PostGIS to store projection information (allowing you to add your custom projections without recompiling or even editing local files), as well as metadata about your own geographic tables. .. If you're new to PostGIS, we recommend checking out http://postgis.refractions.net/ to get a primer about how it works. It's not strictly necessary to work with GeoDjango, but it can help -- just like knowing SQL is not mandatory to work with Django, but will make you a better Django developer.
We still lack the CSS required to make our admin interface look nice. We need to do three things here. First, we will edit ``settings.py`` to specify ``STATIC_ROOT``, ``STATIC_URL``, ``MEDIA_ROOT``, and ``MEDIA_URL``. Note that we decided to put those files in ``/home/dotcloud/data``. By convention, the ``data`` directory will persist across pushes. This is important: while you could store static assets in ``/home/dotcloud/current`` (or one of its subdirectories), you probably don't want to store media (user uploaded files...) in ``current`` or ``code``, because those directories are wiped out at each push. The next step is to instruct Nginx to map ``/static`` and ``/media`` to those directories in ``/home/dotcloud/data``. This is done through a Nginx configuration snippet. You can do many interesting things with custom Nginx configuration files; http://docs.dotcloud.com/guides/nginx/ gives some details about that. The last step is to add the ``collectstatic`` management command to our ``postinstall`` script. Before calling it, we create the required directories, just in case. .. After pushing this last round of modifications, the CSS for the admin site (and other static assets) will be found correctly, and we have a very basic (but functional) Django project to build on!
Since the ``syncdb`` command was run non-interactively, it did not prompt us to create a superuser, and therefore, we don't have a user to login. To create an admin user automatically, we will write a simple Python script that will use Django's environment, load the authentication models, create a ``User`` object, set a password, and give him superuser privileges. The user login will be ``admin``, and its password will be ``password``. Note that if the user already exists, it won't be touched. However, if it does not exist, it will be re-created. If you don't like this ``admin`` user, you should not delete it (it would be re-added each time you push your code) but just remove its privileges and reset its password, for instance. .. At this point, if we push the code, we will be able to login, but we still lack the CSS that will make the admin site look nicer.
To make sure that the database structure is properly created, we want to call ``manage.py syncdb`` automatically each time we push our code. On the first push, this will create the Django tables; later, it will create new tables that might be required by new models you will define. To make that happen, we create a ``postinstall`` script. It is called automatically at the end of each push operation. .. A few remarks: * this is a shell script (hence the ``#!/bin/sh`` shebang at the beginning), but you can also use a Python script if you like; * by default, ``syncdb`` will interactively prompt you to create a Django superuser in the database, but we cannot interact with the terminal during the push process, so we disable this thanks to ``--noinput``. If you push the code at this point, hitting the ``/admin`` URL will display the login form, but we don't have a valid user yet, and the login form won't have the usual Django CSS since we didn't take care about the static assets yet.
We will now activate the Django administration application. Nothing is specific to DotCloud here: we just uncomment the relevant lines of code in ``settings.py`` and ``urls.py``. .. If we push our application now, we can go to the ``/admin`` URL, but since we didn't call ``syncdb`` yet, the database structure doesn't exist, and Django will refuse to do anything useful for us.
Now, we need to edit ``settings.py`` to specify the host, port, user, and password to connect to our database. When you deploy your application, these parameters are stored in the DotCloud Environment File. This allows you to repeat the deployment of your application (e.g. for staging purposes) without having to manually copy-paste the parameters into your settings each time. If you don't want to use the Environment File, you can retrieve the same information with ``dotcloud info hellodjango.db``. The Environment File is a JSON file holding a lot of information about our stack. It contains (among other things) our database connection parameters. We will load this file, and use those parameters in Django's settings. See http://docs.dotcloud.com/guides/environment/ for more details about the Environment File. .. .. note:: We decided to use the ``template1`` database here. This was made to simplify the configuration process. If you want to use another database, you will have to create it manually, or to add some extra commands to the ``postinstall`` script shown in next sections.
It's time to add a database to our project! The first step is to tell DotCloud that we want a PostgreSQL server to be added to our application. Just edit the Build File again. .. Note that we called our database ``db``, but it could have been anything else, really. If you ``dotcloud push`` again, you will see that the database service will be created (DotCloud will notice that we added a section to the Build File).
The ``wsgi.py`` file will bridge between the ``python`` service and our Django app. We need two things here: * inject the ``DJANGO_SETTINGS_MODULE`` variable into the environment, pointing to our project settings module; * setup the ``application`` callable, since that is what the DotCloud service will be looking for. .. We can now push our application, by running ``dotcloud push djangotutorial`` (you can of course use any application name you like). A Python service will be created, the code will be deployed, and the URL of the service will be shown at the end of the build. If you go to this URL, you will see the plain and boring Django page, typical of the "just started" project.
Let's pretend that our Django project is called ``hellodjango``. We will add the essential Django files to our project. Actually, those files did not come out of nowhere: we just ran ``django-admin.py startproject hellodjango`` to generate them! .. note:: The rest of the tutorial assumes that your project is in the ``hellodjango`` directory. If you're following those instructions to run your existing Django project on DotCloud, just replace ``hellodjango`` with the real name of your project directory, of course.
A lot of Python projects use a ``requirements.txt`` file to list their dependencies. DotCloud detects this file, and if it exists, ``pip`` will be used to install the dependencies. In our case, we just need to add ``Django`` to this file. .. ``pip`` is able to install code from PyPI (just like ``easy_install``); but it can also install code from repositories like Git or Mercurial, as long as they contain a ``setup.py`` file. This is very convenient to install new versions of packages automatically without having to publish them on PyPI at each release. See http://www.pip-installer.org/en/latest/requirement-format.html for details about ``pip`` and the format of ``requirements.txt``.
The DotCloud Build File, ``dotcloud.yml``, describes our stack. We will start with a single "python" service (we will add the database later). This service allows us to expose a WSGI-compliant web application. Django can do WSGI out of the box, so that's perfect for us. .. The role and syntax of the DotCloud Build File is explained in further detail in the documentation, at http://docs.dotcloud.com/guides/build-file/.