Improve Your Python: Metaclasses and Dynamic Classes With Type

http://www.jeffknupp.com/blog/2013/12/28/improve-your-python-metaclasses-and-dynamic-classes-with-type/

by lukashed • 13 years ago

115 13 13 years ago

As he says himself at the end, these aren't used very often because they aren't very useful.

The example at the end is perfectly well written using the class statement and using register as a class decorator, while being more familiar and readable.

It gets tiring to hear people say "oh advanced Python? Like metaclasses, I'll learn that".

Learn useful things instead, like writing readable, testable code.

jkbr• 13 years ago

I think the reason type is used in the last example is to have the name of the class correspond with the table name. Otherwise a regular class would suffice like you say.

ris• 13 years ago

Remember class decorators didn't exist when metaclasses came about.

dsj36• 13 years ago

I think it was David Beazley who characterized metaclasses as "infecting" an inheritance tree. If `class A` has a metaclass, then all subclasses of `A` will inherit the metaclass as well. This behavior is not shared by class decorators, which only affect the decorated class.

So, I think this is a pretty reasonable place to use them -- getting magic behavior from class decorators / metaclasses is bad enough, but getting surprised when losing it upon subclassing is even worse!

jamesdutc• 13 years ago

This "infectious behaviour" (though perhaps unfairly characterised) leads to the clearest purpose of metaclasses.

Metaclasses (and build_class) is a mechanism by which to enforce a constraint from a base type to a derived type.

(Note that, in practice, there is some trickiness around metaclasses on derived classes: http://seriously.dontusethiscode.com/2013/04/18/derived-meta...)

It's trivial to enforce a constraint from a derived type to a base type. e.g.,

    # base.py
    class Base:
        def spam(self): pass

    # derived.py
    class Derived(Base):
        assert hasattr(Base, 'spam') # or abc, &c.
        def ham(self):
            return self.spam

But how can we enforce a constraint in the other direction? (e.g., abc.ABCMeta)

    # base.py
    from functools import wraps
    
    class metaclass(type):
    	def __new__(m, n, b, d):
    		assert 'spam' in d # must implement, not just inherit
    
    		# can even enforce behaviour via wrapping
    		spam = d['spam']
    		@wraps(spam)
    		def wrap(*args, **kwargs):
    			print('wrap({}, {})'.format(args, kwargs))
    			return spam(*args, **kwargs)
    		d['spam'] = wrap
    
    		return type.__new__(m, n, b, d)
    
    class Base(metaclass=metaclass):
    	def spam(self): pass
    
    # derived.py
    class Derived(Base):
    	def spam(self): pass

pdonis• 13 years ago

The example at the end is perfectly well written using the class statement

No, it can't, because the name of the class is dynamically determined at run time. The class statement can't take a variable for the name of the class. That's the point.

reuven• 13 years ago

Whenever I teach a course in advanced Python, the people asking for the training specifically request/demand that I talk about metaclasses. This is almost certainly because they (1) want an advanced class, and (2) metaclasses are advanced. So yeah, I cover them to some degree in my courses, but I give plenty of advance warning that we're looking at this because it's fascinating to see how some of the deeper recesses of Python classes work, not because this will truly be useful.

Sure enough, when I'm done with my explanation, the participants agree that metaclasses are really something that they're not planning to touch or use. And then I repeat my claim (which isn't original with me at all) that decorators are easier to understand and maintain, and should be used instead.

I totally agree that learning to write readable, testable code is a far better use of time. (I include that in my classes, too... but people rarely request that, I'm afraid...)

auvrw• 13 years ago

yeah, making things complicated just for the sake of making things complicated sucks. that said, model forms are one of the most useful parts of django, and they use metaclasses, although for all i know, ModelForm could have been implemented another way.

i only know about that part of the django implementation because i was trying to set up some kind of class inheritance and found that things were behaving strangely. so the conclusion in this case was that the use of metaclass allowed some cleverness in the implementation of ModelForm, but it made it difficult to layer more cleverness on top of ModelForm.

temuze• 13 years ago

Here's another amazing tutorial for metaclasses: http://stackoverflow.com/a/6581949/764463

This is probably my all-time favorite StackOverflow post!

Tehnix• 13 years ago

I actually prefer the StackOverflow post you linked over the OPs blog post (not bashing on the blog post, the SO answer is just so good).

SpacemanSpiff• 13 years ago

Did anyone pick up the OP's book, "Writing Idiomatic Python"? I'm tempted to buy, as someone new to Python. Looks like it may be useful for learning some Python best practices.

div• 13 years ago

As luck would have it, I just bought and read "Writing Idiomatic Python" earlier today.

It's a quick read and seems to be a nice shortcut to avoid a few newbie pitfalls (I'm a Python newbie myself).

dmcg• 13 years ago

I'd like to buy the book, but is it really necessary to enter my email and password, twice each, then give my address, all to give you money through PayPal?

Improve Your Python: Metaclasses and Dynamic Classes With Type

Related Stories