Using setattr in a __init__ to add dynamic attributes is a violation of object oriented methodology?

I do realise that python is a dynamic language and meant to be used in a more dynamic way. But when I have read the documentation and found other examples of classes in python community they do not use this setattr approach. They use the standard approach of encapsulation and defining attributes on self within init. This is a standard principle of object oriented programming. Encapsulating data with setters and getters, behaviour via methods effecting the state of the object. You cannot have dynamic attributes added to an object and still be able to consistently use it. There is no context with which to effect the dynamic state. How can encapsulation be followed if the state varies? How can behaviour be applied to it through methods?

Also this opens up subtle hard to find bugs through typos!

All examples I've seen have used the pattern of creating attributes within the init method. I have not seen this setattr pattern in the examples or the python documentation on classes. That is why I think on a theoretical and practical (best practices) stance it shouldn't be encouraged for new python developers. It encourages a bad practice. The dynamic nature of python aside, classes are there to give it a more deterministic approach to OO. Its not the same as prototyping languages like lua and javascript where classes and OO is more loosely applied.

I'm not convinced with the argument that 'its different with python'. OO principles are the same, and community and official examples don't follow this style, which suggests that the consensus of opinion would be not to use setattr in such a way. And definitely not to be taught/encouraged for beginners in a course.

Yes, it's illogical to think that some other piece of software will care about the arbitrary attribute you gave to your object, whether after the fact or during initialization. As the Zen of Python says "explicit is better than implicit", so you shouldn't expect anything that wasn't explicitly configured to work. Just because I gave my str instance an attribute of reversed=True doesn't mean it should be reversed.

The setattr pattern isn't in the official tutorial for classes but neither are properties, magic methods, or class methods. It's an introductory tutorial, it won't cover most uses or approaches. Also, the Python docs are pretty sparse about __init__, __new__, and setattr. Again, they give the basic uses for them and that's it. Pretty common for Python docs, actually.

Looking for examples, I like this one. Yes, it's from 2009 but setattr and __init__ are both pretty old so I think it's acceptable. Ned Batchelder and Lennart Regebro, two people that do quite a bit of Python development and education both weigh in on the question. Note that the original example uses setattr and, oddly, no one calls it out (except for not doing exactly what they're wanting in the original question).

Reuven Lerner, also a common Python educator, shows an example of using it. In fact, they go even further and use the inspect module to do a bit more. Definitely overkill for a beginner class, though.

And even the amazing attrs project uses it to make setting attributes on a class easier and faster.

I 100% agree that you shouldn't use only setattr to set the attributes on your class instances. It is, however, fine to use for catching extra arguments that someone wants to toss onto an instance. Should you limit __init__ to only arguments you'll explicitly set on the instance? Maybe. I think that's going to come down to how your code is going to be used and by whom. Is it completely irresponsible to use setattr inside of __init__? No, not at all.