Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > comp.lang.python > #34067 > unrolled thread

weird isinstance/issubclass behavior?

Back to article view | Back to comp.lang.python

Contents

  weird isinstance/issubclass behavior? lars van gemerden <lars@rational-it.com> - 2012-11-29 06:59 -0800
    Re: weird isinstance/issubclass behavior? Chris Angelico <rosuav@gmail.com> - 2012-11-30 02:08 +1100
    Re: weird isinstance/issubclass behavior? lars van gemerden <lars@rational-it.com> - 2012-11-29 08:07 -0800
    Re: weird isinstance/issubclass behavior? Terry Reedy <tjreedy@udel.edu> - 2012-11-29 16:00 -0500

#34067 — weird isinstance/issubclass behavior?

Hi,

I have encountered some strange behavior of isinstance(/issubclass): depending on the import path used for classes i get different output, while the classes i compare are in the same file. 

Basically if i import a class as:

    from mod1.mod2 import A

or:

    from mod0.mod1.mod2 import A

which both result in importing the same class, a call to isinstance(inst, A) in another module can have a different output. In this module

    print type(inst), A, isinstance(inst, A), issubclass(type(inst), A)

gives:

    <class 'mod0.mod1.mod2.A'> <class 'mod1.mod2.A'> False False

resp.

    <class 'mod0.mod1.mod2.A'> <class 'mod0.mod1.mod2.A'> True True

which seems somewhat logical, but in my case a strange gotcha. My question is:
Is this intended, inevitable or a bug?

Cheers, Lars

PS: this is somewhat simpler than the actual case i've encountered, and i haven't tested this exact case, but for now i hope this is enough to get some of your insight.

[toc] | [next] | [standalone]

#34068

On Fri, Nov 30, 2012 at 1:59 AM, lars van gemerden <lars@rational-it.com> wrote:
> Basically if i import a class as:
>
>     from mod1.mod2 import A
>
> or:
>
>     from mod0.mod1.mod2 import A
>
> which both result in importing the same class, a call to isinstance(inst, A) in another module can have a different output.

What you may be seeing there is that you've imported the module twice.
There are two entirely separate module objects, taken from the same
file. Something instantiated from one class isn't an instance of the
other class even if they're indistinguishable classes; a little
monkeypatching might show you what's going on (fiddle with one class,
check the other, fiddle hasn't happened).

As a general rule, importing a module in different ways is considered
dangerous. Be consistent, and then this sort of oddity won't occur -
and you also won't have other oddities, eg with other global state.

ChrisA

[toc] | [prev] | [next] | [standalone]

#34070

On Thursday, November 29, 2012 3:59:37 PM UTC+1, lars van gemerden wrote:
> Hi,
> 
> 
> 
> I have encountered some strange behavior of isinstance(/issubclass): depending on the import path used for classes i get different output, while the classes i compare are in the same file. 
> 
> 
> 
> Basically if i import a class as:
> 
> 
> 
>     from mod1.mod2 import A
> 
> 
> 
> or:
> 
> 
> 
>     from mod0.mod1.mod2 import A
> 
> 
> 
> which both result in importing the same class, a call to isinstance(inst, A) in another module can have a different output. In this module
> 
> 
> 
>     print type(inst), A, isinstance(inst, A), issubclass(type(inst), A)
> 
> 
> 
> gives:
> 
> 
> 
>     <class 'mod0.mod1.mod2.A'> <class 'mod1.mod2.A'> False False
> 
> 
> 
> resp.
> 
> 
> 
>     <class 'mod0.mod1.mod2.A'> <class 'mod0.mod1.mod2.A'> True True
> 
> 
> 
> which seems somewhat logical, but in my case a strange gotcha. My question is:
> 
> Is this intended, inevitable or a bug?
> 
> 
> 
> Cheers, Lars
> 
> 
> 
> PS: this is somewhat simpler than the actual case i've encountered, and i haven't tested this exact case, but for now i hope this is enough to get some of your insight.

I know for sure that the imports both import the same file, though if i understand you correctly, it creates 2 different module objects? Are module object only created on an import statement?  

The 2 parameters of isinstance() follow a very different import path. 

Anyway, to not spend too much time us this, i'll just file it under 'dangerous' and use the second import statement. It does solve my problem. 

Thanks, Lars

[toc] | [prev] | [next] | [standalone]

#34078

On 11/29/2012 9:59 AM, lars van gemerden wrote:
> Hi,
>
> I have encountered some strange behavior of isinstance(/issubclass): depending on the import path used for classes i get different output, while the classes i compare are in the same file.
>
> Basically if i import a class as:
>
>      from mod1.mod2 import A
> or:
>      from mod0.mod1.mod2 import A
>
> which both result in importing the same class,

As other said, both import the same abstract class but create two 
different concrete class objects.

 > a call to isinstance(inst, A) in another module can have a different 
output.
> In this module
>      print type(inst), A, isinstance(inst, A), issubclass(type(inst), A)
> gives:
>      <class 'mod0.mod1.mod2.A'> <class 'mod1.mod2.A'> False False

Add print id(type(inst)), id(A) and you will see that they are different 
objects. The isinstance and issubclass calls compare the classes by 
identity (the default meaning of ==) and so False, False are correct.

-- 
Terry Jan Reedy

[toc] | [prev] | [standalone]

Back to top | Article view | comp.lang.python

csiph-web