[Python-projects] Suggestion about E1101 Instance of 'None' has no 'get' member

[Barry Scott]

Sylvain Thénault wrote:
> On Tuesday 12 December à 13:38, Barry Scott wrote:
>   
>> I see a lot of E1101 for classes like this:
>>
>> class Foo:
>>     def __init__( self ):
>>        self.some_object = None
>>
>>     def setSome( self, obj ):
>>        self.some_object = obj
>>
>>     def useSome( self ):
>>        self.some_object.someMethod()
>>
>>
>> pylint will complain about the call to someMethod.
>>     
>
> which pylint/astng version are you using ? I don't get any E1101 With your code 
> sample (the setSome method introduce an ambiguity in the inference and
> in that case pylint don't issue E1011).
>   
You are right the simple example does not break, I've not duplicated the 
essence
correctly from my production code. Try this:

class Foo:
    def __init__( self ):
        self.some_object = None

    def useSome( self ):
        self.some_object.someMethod()

a.py:9: [E1101, Foo.useSome] Instance of 'None' has no 'someMethod' member

In this example it shows that someMethod is check but pylint has no 
knowledge
of the intended class of some_object.

In my original example no diagnostic of produced I guess because pylints 
knows
that it cannot get it right.
>  
>   
>> Is there a way to tell pylint that self.some_object is expected to have
>> a certain type? If not is such a feature possible to implement?
>>
>> I'm thinking about annotation like:
>>        self.some_object = None # pylint: variable_type=Some
>>
>> meaning that some_object is expected to be class Some.
>>     
>
> this is certainly feasible at the astng level, the main difficulty being
> that comments are not in the ast and so a preparsing of the module file
> will be necessary (i think). Anyway we'll have to support this kind of
> thing at some point...
>
>   
Indeed. Adding this support will make pylint significantly more useful.

You already parse the comments to process the pylint: ones. Isn't this 
an extension of
that parsing?

Barry