My favorite terrible Python error message

Students in my Python classes occasionally get the following error message:

TypeError: object() takes no parameters

This error message is technically true, as I’ll explain in a moment. But it’s surprising and confusing for people who are new to Python, because it doesn’t point to the source of the actual problem.

Here’s the basic idea: Python methods are attributes, which means that when we invoke methods, Python needs to search for the attribute we’ve named. In other words, if I invoke:

o.m()

then Python will first look for the “m” attribute on the “o” object. If “o” has an attribute named “m” (i.e., if hasattr(o, ‘m’) returns True) then it retrieves the attribute’s value, and tries to call it.

However, Python methods aren’t defined on individual objects. They’re defined on classes. Which means that in almost all cases, if “m” is an actual method that can be invoked on “o”, there won’t be any “m” attribute on “o”.  Instead, we’ll need to look at type(o), the class to which “o” belongs, and look there.

And indeed, that’s how attributes work in Python: First search on the named object. If the attribute isn’t there, then look at the object’s class.  So we look for “m” on o’s class.  If the attribute is there, then it is invoked.  That’s what happens in normal method calls.

But say that the attribute isn’t on the class, either. What then? Python continues its search, looking next at the class from which type(o) inherits — which is located on the attribute type(o).__bases__.  This is a tuple, because Python classes can inherit from more than one parent; let’s ignore that for now.

Most classes inherit from the base object in the Python universe, known as “object”.  In Python 3, if you don’t specify “object” as the base from which you inherit, then it’s done for you automatically. In Python 2, failing to specify that a class inherits from “object” means that you have an “old-style class,” which will operate differently. I continue to specify “object” in my Python 3 classes, partly out of habit, partly because I think it looks nicer, and partly because I want my code to be compatible across versions as much as possible.

What happens if the attribute doesn’t exist on “object”?  Then we get an “attribute error,” with Python telling us that the attribute doesn’t exist.

However, this isn’t what happens in the case of the error message I showed:

TypeError: object() takes no parameters

This error message happens when you try to create a new instance of a class. For example:

class Foo(object):
    pass

If I say

f = Foo()

then I don’t get any error message. But if I say

f = Foo(10)

then I get the TypeError.  Why?

Because Python objects are created in two stages: First, the object is created in the __new__ method. This is a method that we almost never want to write; let Python take care of the allocation and creation of new objects.

However, __new__ doesn’t immediately return the object that it has created. Rather, it first searches for an __init__ method, whose job is to add new attributes to the newly created object. How does it look for (and then invoke) __init__?  It turns to the new object, which I’ll call “o” here, and invokes

o.__init__()

So, what happens now? Python looks for “__init__” on “o”, but doesn’t find it.  It looks for “__init__” on type(o), aka the “Foo” class, and doesn’t find it.  So it keeps searching, and looks on “object” for an “__init__” attribute.

Good news: object.__init__ exists!  Moreover, it’s a method!  So Python tries to invoke it, passing the argument that I handed to Foo (i.e., 10).  But object.__init__ doesn’t take any arguments. And thus we get the error message

TypeError: object() takes no parameters

What’s especially confusing, for me and many of my students, is that Python doesn’t say, “object.__init__()” takes no parameters. So they’re not sure how object figures into this, or where their mistake might be.

After reading this, though, I’m hoping that you can guess what it means: Simply put, this error message says, “You forgot to define an __init__ method on your object.”  This can be out of forgetfulness, but I’ve also seen people forget one or more of the underscores on either side of “__init__”, or even (my favorite) define a method called “__int__”, which is great for converting objects into integers, but not for initializing attributes.

So, is the error message wrong? No, it’s perfectly logical. But as with many “perfectly logical” things, it makes sense after you are steeped in the overall logic of the system, and tends to confuse those who most need the help.