Dr. Dobb's | Storm: An ORM for Python

Storm: An ORM for Python

Scaling issues is just one reason why another Python ORM comes on the scene

July 14, 2007
URL:http://www.drdobbs.com/storm-an-orm-for-python/201000460

Gustavo Niemeyer is lead developer on Canonical's Storm project, a generic open-source object-relational mapper for Python. Canonical, you might recall, is probably best known as developer of the Ubuntu Linux distribution.

DDJ: Gustavo, Storm is a object-relational mapper (ORM) for Python. But there are a number similar tools available today. What's unique about Storm that we need another ORM for Python?

GN: It's true that there are a few good ORMs for Python, and we're very much against reinventing existing technology whenever possible. However due to the enterprise class of the projects being developed by Canonical the same kind of data may be partitioned on multiple databases, for example, as well as other scaling issues not necessarily faced by other ORM tools. We've tried to use existing Python ORMs in that model for some time, and even tried to collaborate with upstream projects, but the result wasn't satisfactory, considering the timeframe available. This lead to Storm being started as a prototype, and when it eventually proved to be a feasible solution for the problem, it replaced the ORM in use up to that point.

DDJ: There are other open source languages -- Ruby comes to mind -- that could use an ORM. How difficult would it be to use what you've done to provide an ORM for another langauge?

GN: It's hard to estimate how much work would be involved for it to be done. Most of the system should be fairly simple to port to any other high-level object-oriented language such as Ruby. A few parts of it may be a bit trickier though, since they explore Python-specific behavior in areas like attribute access. But then that is why we put the code out there. We would be very excited to see it ported to other languages or built against other libraries or into other offerings. That's the beauty of open source after all.

DDJ: Is this the case that you needed a tool like Storm internally, so you created it and found it so good that you decided to share it with other?

GN: Yes, this describes very well what happened. Canonical is very keen on publishing valuable work under open source licenses whenever possible, and also on making available infrastructure for these projects to be developed on. Launchpad and Bazaar are good examples of infrastructures for open source projects to be developed on, even though they are not open source themselves.

DDJ: If readers want to learn more about Storm, is there a web site they can go to?

GN: Yes, at the Storm website we have a tutorial available at the moment, and we are working on more detailed documentation. In the meantime, any uncovered topics are welcome on the mailing list.