Search code examples
djangomavenproject-managementcontinuous-integrationbuild-management

Project management/build tools for a Django project?


Coming from Java development where build and project management tools abound, I'd like to know what's available for Django. I'd really like to use something like Maven for building things, but is there another preferred way of doing it?

I'm really looking for the following:

  • Command-line building: mvn install is just so easy and cool.
  • Command-line test-runs. I'd like to integrate this app into something like Hudson for continuous integration, since I'm hardcore like that.
  • Deployment of media to local test server (JS, CSS, images, etc.)

Is this currently possible with Maven or another tool? I'm embarking on a pretty big project here and I'd like to have a kicking-rad build/project management system like Maven to help the project be able to grow over time.


Solution

  • Two tools come to mind which are both generic python tools - they need not work with Django specifically:

    • Fabric. We use this; it allows us to write remote commands as if we were SSH'd in, upload code etc. These isn't much you can't do and because it is essentially a bash script written in python, it is very easy to get going. But it is also a bash script written in python, which means you can import parts of your django app, run your tests or do anything python can do in the process of running your deployment.
    • Buildout. I've not used this, but our django frontent developer tells me this is absolutely the tool to use. At a guess, it is either the same idea or a slightly more abstract, more python-orientated equivalent.

    I am sure there are other packages available. The best advice I can give you is to evaluate them all briefly and pick the one that best fits your scenario/team working style. Note that these are mostly deployment tools - builds in python don't really make sense, since you don't compile python code.

    In terms of CI related stuff, there are three commands from django you need to know about:

    • ./manage.py test runs your tests from all tests.py files. These are basically django-specific enhancements to unittest, a python package.
    • ./manage.py collectstatic collects static files to a directory of your choosing; highly useful for pulling all the bits together to exist on a static media server. Have a look at the static files howto for how this works.
    • You need South for this one, ./manage.py schemamigration app --auto && ./manage.py migrate app - this applies model changes to the underlying sql schema. Basically, if you just change a model and don't use south, Django will get very upset because the new field won't map to the old sql schema and it won't be able to persist your model. Schema migrations, which south supports, enable you to upgrade an existing, populated database. Worth doing by hand before automating.
    • I should also point out you can write custom management commands to execute ./manage.py whatever you like. This has a fair amount of potential - I've used management commands as the command to execute in a cron job, for example. You could also build interactive commands for admins of your application and, of course, deployment and testing.

    How these are handled really depends on what setup you have available and what tools you choose.

    On compilation: The closest you get to building with python is freeze, a.k.a. py2exe or cxfreeze which produce binaries. Be aware though that all these do is store python bytecode in an exe and pass it through the interpreter, which is a shared object anyway. Also, you can't cxfreeze a django app, since django itself uses dynamic imports which occur at run time and so cannot be evaluated by cxfreeze, which is essentially a compile time tool. So discount building as a task that needs doing. You might see .pyc files appearing in your directory - python converts your python script to python bytecode if changes have been made to it since the last pyc. If no changes have been made, it loads the pyc file from last time. Python bytecode is what gets executed, so this isn't really a speedup in terms of performance, just load time.