CircArgs commented on PR #4466:
URL: https://github.com/apache/iceberg/pull/4466#issuecomment-1106714989
Hey @rdblue I wanted to circle back on the original without raising the
error. I threw it in without considering the implications based on your
suggestion, but I feel as though we still don't need such a thing. When one
creates an `And` for example, it will call `__new__` and run through the logic
on whether we are dealing with a tautology or a contradiction and yield
`AlwaysTrue` or `AlwaysFalse` respectively. This logic, in conjunction with
taking in `*rest` and reducing to binary `And`s makes us have to return and
that's why I'm using `__new__`. Take this example:
```python
class A:
def __new__(cls):
return B()
def __init__(self):
print('A init')
class B:
def __init__(self):
print('B init')
A()
>>> "B init"
So A's `__new__` was called but the resulting object was a B so its
`__init__` was called. I'm just showing how `__new__` has full control here and
that `And`'s `__init__` will only ever see what `__new__` wants it to see. As
for what happens via possible calls to `super` from child classes, do we have
to worry about that? I believed these classes would not be inherited from and
if they were it is the user's responsibility to understand the implications
especially when dealing with dunder methods.
```
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
