Python refactoring question and create dynamic attributes

[Arup Rakshit]

> On 23-Jun-2019, at 2:31 PM, Cameron Simpson <cs at cskk.id.au> wrote:
> 
> On 23Jun2019 13:26, Arup Rakshit <ar at zeit.io> wrote:
>> In the below code:
>> 
>>   @classmethod
>>   def find(self, id):
>>       if isinstance(id, list):
>>           rows = self.__table__().get_all(*id).run(self.__db__().conn)
>>           result = []
>>           for row in rows:
>>               acategory = Category()
>>               acategory.__dict__.update(row)
>>               result.append(acategory)
>>           return result
>>       else:
>>           adict = self.__table__().get(id).run(self.__db__().conn)
>>           acategory = Category()
>>           acategory.__dict__.update(adict)
>>           return acategory
>> 
>> I have 2 questions:
>> 
>> 1. Is there any better way to create attributes in an object without using __dict__().update() or this is a correct approach?
> 
> setattr() is the usual approach, but that sets a single attribute at a time. If you have many then __dict__.update may be reasonable.
> 
> You should bear in mind that not all objects have a __dict__. It is uncommon, but if a class is defined with a __slots__ attribute then its instances have fixed attribute names and there is no __dict__. Also some builtin types have not __dict__. However, you likely know that the objects you are using have a __dict__, so you're probably good.
> 
> Also, __dict__ bypasses properties and descriptors. That might be important.
> 
>> 2. Can we get the same result what for row in rows: block is producing without killing the readability ?
> 
> Not obviously. It looks pretty direct to me.
> 
> Unless the Category class can be made to accept an attribute map in its __int__ method, then you might do some variable on:
> 
> result = [ Category(row) for row in rows ]
> 
> which is pretty readable.
> 
> BTW, @classmethods receive the class as the first argument, not an instance. So you'd normally write:
> 
> @classmethod
> def find(cls, id):
>   …
> 

What I know, is that first argument is reserved for the instance upon which it is called. It can be any name, so continued to use self. Yes these methods are class method intentionally. I am not aware of so far the naming rules of the first argument of a class or instance method.

> and you would not have a self to use. is __table__ a class atribute or an instance attribute?

Yes __table__ class attribute defined in the BaseModel. 

> 
>> To see the context, here is my source code https://gitlab.com/aruprakshit/flask_awesome_recipes/blob/master/app/models/category.py
> 
> Ah, so Category inherits from BaseModel. That means you can write your own __init__. If it accepted an optional mapping (i.e. a dict or row) you could put the .update inside __init__, supporting the list comprehension I suggest above.

Nice idea. I’ll try this.

> 
> It looks like you've marks almost all the methods as @classmethods. Are you sure about that? They all seem written to use self, and thus be instance methods.
> 
> Cheers,
> Cameron Simpson <cs at cskk.id.au>



Thanks,

Arup Rakshit
ar at zeit.io