I have a program which calculates a set of plain interlinked objects (the objects consist of properties which basically are either String, int or link to another object). I would like to have the objects stored in a relational database for easy SQL querying (from another program).
Moreover, the objects (classes) tend to change and evolve. I would like to have a generic solution not requiring any changes in the 'persistence layer' whenever the classes evolve.
Do you see any way to do that?
What about storing the objects in JSON? You could write a function that serialize your object before storing it into the database. If you have a specific identifier for your objects, I would suggest to use it as index so that you can easily retrieve it.
I haven't used it myself, but I this sounds like use case for SQLAlchemy:
It basically let's you define object-properties as table-columns or relations to other classes and takes care of basically all direct db-interaction. There might be other libraries out there doing the same, but SQLAlchemy is the one I once stumbled over and seems to be pretty commonly used.
Column('name', Type) stanza.
sebastian's idea is probably the easiest to use, since it does the object mapping for you, however, if you are already familiar with SQL, and prefer that, and are using python 2.5 or later, you might consider the sqlite3 library. It's a very easy to use library, and I'm pretty sure it doesn't require any work to set it up. You might have to have the sqlite3 libraries installed, though. http://docs.python.org/2/library/sqlite3.html
SQLAlchemy is great. Another leading option is using the ORM from django:
https://docs.djangoproject.com/en/dev/intro/tutorial01/#creating-models
Basically creating classes in python which extend the django model class are mapped to standard relational database tables.
