子类化namedtuple，__ new__和__init__

Question

I am very new to Python and am currently using Python 3.4

Right now, I am apparently banging my head to figure out how best to create a customized immutable object by subclassing 'namedtuple'.

Here are my code that works without errors but I am getting way too many warning messages from my good friend PyCharm, so it was set upon me to find out if I am doing this correctly.

from collections import namedtuple

class Aloe(namedtuple('Plant', [
    'name',
    'color',
    'loveliness'
])):

    def __new__(cls, params):
        return super(Aloe, cls).__new__(
            cls,
            params.get('name'),
            params.get('color'),
            params.get('loveliness')
        )

    def __init__(self, params):
        print(self.name)
        super(Aloe, self).__init__()

When I call it, I get this.

>>> aloe = Aloe(params)
>>> print(aloe)
Aloe(name='aloey', color='green', loveliness='extreme')

---

The first of warning come from __new__ block.

All my params.get('FIELD_NAME') are blamed for ' Unexpected argument '.

Funny thing is, if my remove the arguments, I got,

TypeError: __new__() missing 1 required positional argument:

What does this thing really want me to do?!

---

The second of warning come from __init__ block.

super(Aloe, self).__init__() is blamed for ' Parameter 'typename' unfilled '

So I gave it a good 'Plant' as the desired 'typename'. Then I got,

TypeError: object.__init__() takes no parameters

(ノಠ益ಠ)ノ彡┻━┻

---

Back to the point,

1. My code is working, but I get loads of warnings from PyCharm.

2. There seems absolutely no way I can satisfy PyCharm.

So, I am concerned that I might be neglecting the Pythonic way of handling things. If I am, I wish I could use some tips.

Answer 1

PyCharm is being overly protective here, and incorrect to boot. However, if all you wanted to do is provide some default values, I'd use a factory method:

class Aloe(namedtuple('Plant', 'name color loveliness')):
    @classmethod
    def from_row(cls, name=None, color=None, loveliness=None, **ignored):
        return cls(name, color, loveliness)

and use **params when creating one:

Aloe.from_row(**params)

This applies the key-value pairs from params as separate keyword arguments. Missing keys will set to None , extra keys are captured in **ignored and ignored.

If your database rows will always produce exactly the right 3 keys in a row dictionary, don't even bother with a factory method, just directly use a namedtuple and the above **params syntax:

Aloe(**params)