Noob Q: subclassing or wrapping file class

[kj]

I'm trying to get the hang of Python's OO model, so I set up this
conceptually simple problem of creating a new file-like class to
read a certain type of file. The data in this type of file consists
of multiline "chunks" separated by lines consisting of a single
".".

My first crack at it looks like this:

class MyFile():
def __init__(self, f):
if hasattr(f, 'next'):
self.fh = f
else:
self.fh = file(f, 'r')

def __iter__(self):
return self

def next(self):
buf = []
for line in self.fh:
if line == '.\n':
break
buf.append(line)
if len(buf) == 0:
raise StopIteration
return buf

For the app I have in mind, I'd like to be able to pass an open
read handle as the argument to the constructor. Hence the form of
__init__.

This works correctly, as far as it goes. There are still many
other file methods that would make sense to implement for this
class and I'm wondering if there's a simple way to do it, without
having to actually write a whole bunch of methods of the form

def foo(self, x, y, z):
return self.fh.foo(x, y, z)

I thought at first that I could achieve this by overriding __getattr__:

def __getattr__(self, attribute):
return self.fh.__getattr__(attribute)

But to my surprise this did not work too well. For example, if I
use a GzipFile object as the argument to MyFile, the iterator of
the resulting object works as expected, but if I attempt to access
its closed attribute I get the error

AttributeError: GzipFile instance has no attribute '__getattr__'

And it doesn't have __getattribute__ either. (And here I'd been
thinking that __getattr__ and __getattribute__ were pretty much
universal... Clearly not!)

Is there a standard idiom that I'm missing for doing this sort of
thing?

Thanks!

kynn

 

[Steven D'Aprano]

On Wed, 16 Sep 2009 21:56:09 +0000, kj wrote:

....

I thought at first that I could achieve this by overriding __getattr__:

def __getattr__(self, attribute):
return self.fh.__getattr__(attribute)

But to my surprise this did not work too well. For example, if I use a
GzipFile object as the argument to MyFile, the iterator of the resulting
object works as expected, but if I attempt to access its closed
attribute I get the error

AttributeError: GzipFile instance has no attribute '__getattr__'

And it doesn't have __getattribute__ either. (And here I'd been
thinking that __getattr__ and __getattribute__ were pretty much
universal... Clearly not!)

Is there a standard idiom that I'm missing for doing this sort of thing?

Instead of:

x.__getattr__('name')

write this:


getattr(x, 'name')



Also, do a search for "automatic delegation python", or just look at Alex
Martelli's cookbook recipe:

http://code.activestate.com/recipes/52295/

 

[kj]

Instead of:

write this:

getattr(x, 'name')

This did the trick.

Also, do a search for "automatic delegation python", or just look at Alex
Martelli's cookbook recipe:

http://code.activestate.com/recipes/52295/

That was instructive. Thanks!

kj

 

[Terry Reedy]

I'm trying to get the hang of Python's OO model, so I set up this
conceptually simple problem of creating a new file-like class to
read a certain type of file. The data in this type of file consists
of multiline "chunks" separated by lines consisting of a single
".".

My first crack at it looks like this:

class MyFile():
def __init__(self, f):
if hasattr(f, 'next'):
self.fh = f
else:
self.fh = file(f, 'r')

I believe open(f, 'r') does the same thing. In 3.x, 'file' is gone and
you must use 'open', so you might want to start using it now.

def __iter__(self):
return self

def next(self):
buf = []
for line in self.fh:
if line == '.\n':
break
buf.append(line)
if len(buf) == 0:
raise StopIteration
return buf

FYI, what you have written to this point is an iterator class that could
be rewritten as a generator function. You might fine that an instructive
exercise.

Terry Jan Reedy

 

[Steven D'Aprano]

This did the trick.

For the record, it's fairly unusual to call double-underscore special
methods directly. Any time you think you need to, it's worth a rethink.