Why do SQLAlchemy classes inheriting from Base not need a constructor?

Question

With SQLAlchemy objects inheriting from the Base class I can pass arguments to a class for variables which aren't defined in a constructor:

from sqlalchemy.ext.declarative import declarative_base

Base = declarative_base()

class User(Base):
    __tablename__ = 'users'
    id = Column(Integer, Sequence('user_id_seq'), primary_key=True)
    name = Column(String(50))
    fullname = Column(String(50))
    password = Column(String(12))

    def __repr__(self):
        return "<User(name='%s', fullname='%s', password='%s')>" % (
                                self.name, self.fullname, self.password)

ed_user = User(name='ed', fullname='Ed Jones', password='edspassword')

of course, if I tried passing arguments like that to another class, to set class variables in the same way I'd get an error:

In [1]: class MyClass(object):
   ...:     i = 2
   ...:     

In [2]: instance = MyClass(i=9)
---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-2-6d3ec8445b00> in <module>()
----> 1 instance = MyClass(i=9)

TypeError: object.__new__() takes no parameters

What sort of trickery is SQLalchemy doing? Why don't they just use a constructor (i.e an __init__ method)?

Answer 1

You actually asked two questions here.

First, you asked "What sort of trickery is SQLAlchemy doing". There is a bit more going on behind the scenes using a concept called Metaclasses to dynamically create the Base class.

But in reality all you need to know is that SQLAlchemy is defining a constructor (albeit in a roundabout way) in the Base class that dynamically sets the elements. Here is actually the implementation of that method (at least as it exists at the time of this answer):

def _declarative_constructor(self, **kwargs):
    """A simple constructor that allows initialization from kwargs.

    Sets attributes on the constructed instance using the names and
    values in ``kwargs``.

    Only keys that are present as
    attributes of the instance's class are allowed. These could be,
    for example, any mapped columns or relationships.
    """
    cls_ = type(self)
    for k in kwargs:
        if not hasattr(cls_, k):
            raise TypeError(
                "%r is an invalid keyword argument for %s" %
                (k, cls_.__name__))
        setattr(self, k, kwargs[k])

Basically, this is dynamically determining the keyword arguments, and setting the attributes on the new object automatically. You can imagine theBase class as looking like the following, although keep in mind it is actually a bit more complex (you can read the code to find out more):

class Base(object):
    def __init__(self, **kwargs):
        cls_ = type(self)
        for k in kwargs:
            if not hasattr(cls_, k):
                raise TypeError(
                    "%r is an invalid keyword argument for %s" %
                    (k, cls_.__name__))
            setattr(self, k, kwargs[k])

If you created the above code, any class that you create that inherits from Base would automatically get the ability to have the attributes of the property auto-filled as long as the attribute was already defined on the class. This is because of the typical Python Object-Oriented inheritance structure: if you don't define a method on your User object, Python looks for the method being defined on a base class (in this case the Base method) and will use that instead. This goes for the __init__ method, just like any other method.

Your second question is "Why don't they just use a constructor (i.e an __init__ method)?" Well, as we've described above, they do! They set the _declarative_constructor method to the __init__ attribute the Base class, effectively setting the default logic for object construction. Of course, this only defines the default; you can always override this if you want to...

class User(Base):
    __tablename__ = 'users'
    id = Column(Integer, Sequence('user_id_seq'), primary_key=True)
    name = Column(String(50))
    fullname = Column(String(50))
    password = Column(String(12))

    def __init__(self, name):
        self.name = name
        self.fullname = name
        self.password = generate_new_password()

# The following will now work...
ed_user = User('ed')
mark_user = User(name='mark')

# ...but this will not...
problem_user = User(name='Error', fullname='Error M. McErrorson', password='w00t')