Django

From DUNE
Jump to navigation Jump to search

Installation

There are a few ways to install Django, perhaps the cleanest and easiest is by using pip. With Python 3+ you will need to install pip3 first, like

apt-get install python3-pip

After that, Django is obtained by

pip3 install django==1.10

...and other versions available instead of 1.10 can be specified if needed. An important and popular Django add-on package "tables2" can be added likewise:

pip3 install django-tables2

To check which version of Django you are using at the moment, start interactive Python and use this:

import django
django.VERSION

DB

sqlite

Django comes packaged with a sqlite database. It will work fine for initial development and testing, but it won't handle concurrency in most cases (e.g. multiple clients of same type updating same DB tables). There are a few reasons for that, such as:

  • lack of table/row lock capability
  • presence of the file lock on the whole DB file

If the application requires any level of concurrency a different DB engine will be needed such as PostgreSQL, MySQL, Oracle etc.

Deploying DB for Django

sqlite permissions

Assuming you are using sqlite, the file permissions on the DB file do matter if when you deploy under Apache. So you either need to set wide permissions (may not be a good idea depending on the security situation) or change the owner to "www-data" (on Ubuntu) or "apache" (on CentOS). Other OS may require similar tweaks.

PostgreSQL

An example of the "settings.py" clause:

DATABASES = {
    'default': {
        'ENGINE': 'django.db.backends.postgresql',
        'NAME': 'foo',
        'USER': 'bar',
        'PASSWORD': '***',
        'HOST': '',
        'PORT': '',
    }
}

Transactions

Utilizing transactions may become necessary in scenarios when concurrency needs to be handled. In addition, locking specific rows for update can be achieved transparently (if the DB backend supports it) in the following manner:

entries = Entry.objects.select_for_update().filter(author=request.user)

This should be done within a transaction block which perhaps throws an exception if the row is locked.

The Development Server

Not suitable for real life deployment, as the name suggests it's for development only. In addition to potential security issues, it may be challenging for thread safety unless one disables multithreading. If you want to access the development server on a local network, from a different computer, the command line below can serve as an example of how to start it:

./manage.py runserver 0.0.0.0:8000

If 0.0.0.0 is skipped, the server will work but won't be accessible from a different machine.