Concrete classes -- stylistic question

Andrew Koenig ark at research.att.com
Thu Oct 10 11:54:44 EDT 2002


On a few occasions I've wanted to define ``concrete classes'' --
classes that are so simple that their structure is their interface.
For example, I might want a little class to hold (x, y) pairs.

Of course, I can use a tuple, but then I have to remember the meaning
of each element.  Moreover, if I want to have (x, y) pairs and (r, theta)
pairs, I don't have an easy way of checking which one is which in case
I use the wrong one by accident.

Of course, I could define little classes like this:

        class xy(object):
                def __init__(self, x, y):
                        self.x, self.y = x, y

        class rtheta(object):
                def __init__(self, r, theta):
                        self.r, self.theta = r, theta

Then, if I execute

        foo = xy(3, 4)
        foo.r

I will learn quickly that I used the wrong class.

If I want a number of such classes, it's something of a pain to define
each one separately.  So I came up with an idea:

    class Record(object):
        def __init__(self, **args):
            for i in args:
                setattr(self, i, args[i])
        def __repr__(self):
            names = filter(
                lambda x: not (len(x) >= 4 and x[:2] == x[-2:] == "__"),
                dir(self))
            names.sort()
            return "Record(" + ", ".join(
                      map(lambda x: x + '=' + `getattr(self, x)`, names)) + ")"

Here's an example of how one might use it:

    >>> foo = Record(x=3, y=4)
    >>> foo
    Record(x=3, y=4)
    >>> foo.x
    3
    >>> foo.y
    4
    >>> foo.r
    AttributeError: 'Record' object has no attribute 'r'

My questions:

        1) Am I reinventing the wheel here?  If so, where is the original?

        2) Is this approach Pythonic?  If not, why not, and what would you
           recommend as an alternative?

-- 
Andrew Koenig, ark at research.att.com, http://www.research.att.com/info/ark



More information about the Python-list mailing list