This is one of the most enjoyable reads I've had in a long time. A must read for any of you programming history buffs. I found the Java and C# contributions most enjoyable.
A Brief, Incomplete, and Mostly Wrong History of Programming Languages.
Tuesday, May 26, 2009
Thursday, May 14, 2009
Odd Python Errors
Over the past year or so I've started keeping a text file around that details some of the not-so-obvious errors that Python might throw out there. Every now and then I run into one of them and it makes my brain hurt just a little bit.
Here are a couple of my favourites.
Pfft? Wha? Ouch. The offending code is as follows.
In other words, C1 and C2 each define a __metaclass__ attribute, but issubclass(M1, M2) and issubclass(M2, M1) both return false. Python raises a TypeError as the metaclasses are not in the same inheritance chain.
The following, however, does work:
In this situation, issubclass(M2, M1) is true so Python is able to build C3 using M2. What happens if we add another base class to C3?
Inserting 'G' into the mix causes the same error to occur.
The same error, but why? M3 is a child of M1, so according to the above example, this should work, correct? Well, not really. Consider the wording of the above error message. The metaclass of C3 is not a subclass of all of it's bases. The fix?
Now the metaclass of C3 (M3) is a subclass of all of the metaclasses of all of C3's bases.
The offending code is as follows:
This one isn't quite so bad. The '__slots__' attribute is used when allocating heap memory for the new type. In general terms, pointers to the attributes defined in '__slots__' are inserted into the new type's structure. Since C1 and C2 both define '__slots__' and they're not from the same inheritance chain, the type structures differ. Thus, we have an instance lay-out conflict.
The fix?
Note that it is entirely possible to add another class to C3's bases so long as it doesn't define any additional '__slots__' attributes.
Are there any other odd errors out there that trip people up? If you've got any, I'd love to add them to my little text file.
Here are a couple of my favourites.
Metaclass Conflicts
There are a few reasons these might pop up, but one error message in particular seems to stand out. Each time I see it I've got to read through it a few times and mentally break it down into components.
Traceback (most recent call last):
File "t.py", line 44, in
class C3(C1, C2): pass
TypeError: Error when calling the metaclass bases
metaclass conflict: the metaclass of a derived \
class must be a (non-strict) subclass of the metaclasses \
of all its bases
Pfft? Wha? Ouch. The offending code is as follows.
class M1(type): pass
class M2(type): pass
class C1(object):
__metaclass__ = M1
class C2(object):
__metaclass__ = M2
class C3(C1, C2): pass #<--- Boom!
In other words, C1 and C2 each define a __metaclass__ attribute, but issubclass(M1, M2) and issubclass(M2, M1) both return false. Python raises a TypeError as the metaclasses are not in the same inheritance chain.
The following, however, does work:
class M1(type): pass
class M2(M1): pass
class C1(object):
__metaclass__ = M1
class C2(object):
__metaclass__ = M2
class C3(C2, C1): pass #<--- No Boom!
In this situation, issubclass(M2, M1) is true so Python is able to build C3 using M2. What happens if we add another base class to C3?
class M1(type): pass
class M2(M1): pass
class M3(M1): pass
class C1(object):
__metaclass__ = M1
class C2(object):
__metaclass__ = M2
class G(object):
__metaclass__ = M3
class C3(G, C2, C1): pass #<--- Boom!
Inserting 'G' into the mix causes the same error to occur.
Traceback (most recent call last):
File "t7.py", line 16, in
class C3(G, C2, C1): pass #<--- Boom!
TypeError: Error when calling the metaclass bases
metaclass conflict: the metaclass of a derived \
class must be a (non-strict) subclass of the metaclasses \
of all its bases
The same error, but why? M3 is a child of M1, so according to the above example, this should work, correct? Well, not really. Consider the wording of the above error message. The metaclass of C3 is not a subclass of all of it's bases. The fix?
class M1(type): pass
class M2(M1): pass
class M3(M2): pass
class C1(object):
__metaclass__ = M1
class C2(object):
__metaclass__ = M2
class G(object):
__metaclass__ = M3
class C3(G, C2, C1): pass #<--- No Boom!
Now the metaclass of C3 (M3) is a subclass of all of the metaclasses of all of C3's bases.
Layout Conflicts
This is another one of my favourites. I spent quite a while tracking this one down the first time it showed up.
Traceback (most recent call last):
File "t88.py", line 7, in
class C3(C2, C1): pass
TypeError: Error when calling the metaclass bases
multiple bases have instance lay-out conflict
The offending code is as follows:
class C1(object):
__slots__ = ('attr1', 'attr2')
class C2(object):
__slots__ = ('attr3', 'attr4')
class C3(C2, C1): pass #<--- Boom!
This one isn't quite so bad. The '__slots__' attribute is used when allocating heap memory for the new type. In general terms, pointers to the attributes defined in '__slots__' are inserted into the new type's structure. Since C1 and C2 both define '__slots__' and they're not from the same inheritance chain, the type structures differ. Thus, we have an instance lay-out conflict.
The fix?
class C1(object):
__slots__ = ('attr1', 'attr2')
class C2(C1):
__slots__ = ('attr3', 'attr4')
class C3(C2, C1): pass #<-- No Boom!
Note that it is entirely possible to add another class to C3's bases so long as it doesn't define any additional '__slots__' attributes.
Are there any other odd errors out there that trip people up? If you've got any, I'd love to add them to my little text file.
Subscribe to:
Posts (Atom)