Thanks for adding it, I read it now.

And sorry to back track a moment - I love the idea of @dataclass and I can
only imagine how must work it was to implement as I am only a beginner.

I'm looking at this primarily from the narrow view point of a user - not so
much understanding the backdrop of how it can actually work.
But I still think this is worth considering as I think this is common
pattern based on this logic:

When using the default construction method it seems reasonable to pass a
default, such as a List.
Because
def __init__(x=[]):
  self.x = x
Is the cleanest way I can think of to write that.
In this new context, if x: list = [] is the cleanest way to write it.
Therefore it should do that, however it is actually implemented.

Let's imagine for a moment that @dataclass becomes the default way of
constructing classes in python.
Do we really want such a common case to require such verbose language?

Looking at Guido's first comment on this, I think that detection mechanism
is what I would expect to happen. To illustrate verbosely:
https://gist.github.com/swirlingsand/d59d01ef79c5ee93f430ed324199bc65
I clearly misunderstood the way classes get passed by default.
I continue to believe that defining it in the argument should be equal to
defining it after. (if there are no other context items present).

From a beginners perspective here, It would appear that in a sense the
default init is actually "silently failing" in that the expected (because
of the two instances not being equal to each other as shown in L10 above)
isolation is not happening.
In a sense then, @dataclass is turning that silent "failure" into a
ValueError which is actually better.

What about something like this:
https://gist.github.com/swirlingsand/2494bc482902fada2248698f7b8af61e
If the common case is for it to be None, then this doesn't fire so there's
no cost.
If it's a common default then it handles it as expected.
If it's not found it raises a ValueError like before, so there's no loss or
harm.
A handful of defaults here may cover 80% of cases and the 20% can continue
to be handled by ValueError
In a project with 40 classes with 10 defaults per class that's 400 lines of
code that look like:
[]
instead of
field(default_factory=lambda: [])
(or {} etc.)

Issue #3 has many comments around copying, but that's not my concern, I'm
just talking about the defaults where the attribute is not provided at all
(ie is None).

I did the above example in regular python since I don't know enough about
how @dataclass is implemented, but it seems reasonable that if it can work
in normal python there should be a way to say augment operation() on line
21 with field()

On Wed, Nov 13, 2019 at 10:11 AM Vedran Čačić <report@bugs.python.org>
wrote:

>
> Vedran Čačić <vedgar@gmail.com> added the comment:
>
> Have you read https://github.com/ericvsmith/dataclasses/issues/3?
>
> ----------
> nosy: +veky
>
> _______________________________________
> Python tracker <report@bugs.python.org>
> <https://bugs.python.org/issue38758>
> _______________________________________
>