One big advantage of using namedtuples, aside from the immutability I mentioned ...

mikepurvis · on Aug 30, 2017

Setting default values for the namedtuple requires subclassing it, though, which is its own little bag of fun. See all the wailing and gnashing of teeth on SO: https://stackoverflow.com/search?q=namedtuple+subclass

JelleZijlstra · on Aug 30, 2017

If you're using Python 3.6, it's a lot easier now. You can do:

from typing import NamedTuple

class Point(NamedTuple): x: float y: float = 0.0

and then Point(1.0).y == 0.0.

sevensor · on Aug 30, 2017

That's quite nice!

sevensor · on Aug 30, 2017

When this is an issue, I tend to create a namedtuple with default values for the fields at the same time I'm defining the namedtuple type itself.

    NT = namedtuple("NT", ('a', 'b'))
    default_NT = NT(42, 'foo')

And then later, if I want to initialize only one field:

    my_new_NT = default_NT._replace(a=7)

mikepurvis · on Aug 30, 2017

I guess that works, but it's a pretty big departure from the expected API of a normal namedtuple, or any other object-with-constructor, for that matter.

sevensor · on Aug 30, 2017

If you really want it to look more traditional, you could do:

    NT.new = lambda **kwargs: NT(42, 'foo')._replace(**kwargs)

Now you can make NT objects like so:

    nt1 = NT.new()
    nt2 = NT.new(b='bar')

You can call it something else if calling it "new" makes you feel weird. (I'd call it something else, but naming things is hard.)