isinstance() bug

[Michal Vitecek]

hello,

please consider the following situation:

under the current directory there's a subdirectory 'package' with two
files: __init__.py and module.py

./package:
__init__.py
module.py

module.py contains:

class A(object):
pass

aModule = A()


now, let's do:

current directory:
[color=blue][color=green][color=darkred]
>>> import package.module
>>> type(package.mo dule.aModule)[/color][/color][/color]
<class 'package.module .A'>[color=blue][color=green][color=darkred]
>>> isinstance(pack age.module.aMod ule, package.module. A) # so far good[/color][/color][/color]
1[color=blue][color=green][color=darkred]
>>> a = package.module. A()
>>> isinstance(a, package.module. A) # so far good
>>> import sys
>>> sys.path.append ('package')
>>> import module
>>> a = module.A()
>>> isinstance(a, package.module. A) # will return 0 !!![/color][/color][/color]
0[color=blue][color=green][color=darkred]
>>> isinstance(pack age.module.aMod ule, module.A) # will return 0 !!![/color][/color][/color]
0[color=blue][color=green][color=darkred]
>>>[/color][/color][/color]

how is it possible that it IS important how you imported a class
definition for isinstance() to work? it's insane!

--
fuf (fuf@mageo.cz)

[Sidharth Kuruvila]

Re: isinstance() bug


"Michal Vitecek" <fuf@mageo.cz > wrote in message
news:mailman.90 7.1075296157.12 720.python-list@python.org ...[color=blue]
> hello,
>
> please consider the following situation:
>
> under the current directory there's a subdirectory 'package' with two
> files: __init__.py and module.py
>
> ./package:
> __init__.py
> module.py
>
> module.py contains:
>
> class A(object):
> pass
>
> aModule = A()
>
>
> now, let's do:
>
> current directory:
>[color=green][color=darkred]
> >>> import package.module
> >>> type(package.mo dule.aModule)[/color][/color]
> <class 'package.module .A'>[color=green][color=darkred]
> >>> isinstance(pack age.module.aMod ule, package.module. A) # so far good[/color][/color]
> 1[color=green][color=darkred]
> >>> a = package.module. A()
> >>> isinstance(a, package.module. A) # so far good
> >>> import sys
> >>> sys.path.append ('package')
> >>> import module
> >>> a = module.A()
> >>> isinstance(a, package.module. A) # will return 0 !!![/color][/color]
> 0[color=green][color=darkred]
> >>> isinstance(pack age.module.aMod ule, module.A) # will return 0 !!![/color][/color]
> 0[color=green][color=darkred]
> >>>[/color][/color]
>
> how is it possible that it IS important how you imported a class
> definition for isinstance() to work? it's insane!
>
> --
> fuf (fuf@mageo.cz)
>[/color]

its got to do with how python imports modules.
both imports are treated as seperate modules because the relative paths are
diferent.
python is a dynamic language it figures out where modules are located at run
time, it would be expensive to check all paths to see if they ended up at
the same file.

[Skip Montanaro]

Re: isinstance() bug


Sidharth> python is a dynamic language it figures out where modules are
Sidharth> located at run time, it would be expensive to check all paths
Sidharth> to see if they ended up at the same file.

Maybe not:

Help on function abspath in module posixpath:

abspath(path)
Return an absolute path.

Considering all the other stuff import has to do, this seems like only a
small extra bit of work.

Skip

[Sidharth Kuruvila]

Re: isinstance() bug

i dont know unix.
how does abs path work with sym-links.

[Sidharth Kuruvila]

Re: isinstance() bug

Stupid post must be ignored at all costs

"Sidharth Kuruvila" <sidharthk@hotm ail.com> wrote in message
news:bv8qhc$ph1 mk$1@ID-223265.news.uni-berlin.de...[color=blue]
> i dont know unix.
> how does abs path work with sym-links.
>
>
>[/color]

[Skip Montanaro]

Re: isinstance() bug


Sidharth> i dont know unix.
Sidharth> how does abs path work with sym-links.

Not well. Seems that os.path.realpat h does the right thing though:

% ls -l /etc/rc2.d
lrwxrwxrwx 1 root root 10 May 21 2002 /etc/rc2.d -> rc.d/rc2.d/
% python
Python 2.2.3 (#1, Jun 21 2003, 08:08:22)
[GCC 3.0.1] on linux2
Type "help", "copyright" , "credits" or "license" for more information.[color=blue][color=green][color=darkred]
>>> import os
>>> os.path.abspath ("/etc/rc2.d")[/color][/color][/color]
'/etc/rc2.d'[color=blue][color=green][color=darkred]
>>> os.path.realpat h("/etc/rc2.d")[/color][/color][/color]
'/etc/rc.d/rc2.d'

As someone else pointed out, the issue of multiple hard links to the same
file is a bit tougher nut to crack.

% cd ~/tmp
% cat > a.py
class A: pass
% ln a.py b.py
% python
Python 2.2.3 (#1, Jun 21 2003, 08:08:22)
[GCC 3.0.1] on linux2
Type "help", "copyright" , "credits" or "license" for more information.[color=blue][color=green][color=darkred]
>>> import a
>>> import b
>>> a.A == b.A[/color][/color][/color]
0

You need to compare inode numbers to see if you have the same or different
files.
[color=blue][color=green][color=darkred]
>>> import stat
>>> os.stat(a.__fil e__)[stat.ST_INO][/color][/color][/color]
37041L[color=blue][color=green][color=darkred]
>>> os.stat(b.__fil e__)[stat.ST_INO][/color][/color][/color]
37041L

Skip

[Jeff Epler]

posixpath.abspa th and symlinks

On Wed, Jan 28, 2004 at 10:37:39PM +0530, Sidharth Kuruvila wrote:[color=blue]
> i dont know unix.
> how does abs path work with sym-links.[/color]


os.abspath does not check whether any components are symlinks. In the
presence of symlinks,
/a/../b
and
/b
can refer to separate files if /a is a symlink to a directory. As far as
I know, posixpath.abspa th normalizes the first into the second anyway.

Similarly, if you have a symlink
/a -> c
then posixpath.abspa th("a") doesn't return /c, it still returns /a

You can use os.readlink() on posix systems to get the thing a symlink
points to:[color=blue][color=green][color=darkred]
>>> os.readlink("/proc/self")[/color][/color][/color]
'14717'[color=blue][color=green][color=darkred]
>>> os.readlink("/")[/color][/color][/color]
OSError: [Errno 22] Invalid argument: '/'

Jeff