
04.03.19 21:24, Guido van Rossum пише:
* Dicts are not like sets because the ordering operators (<, <=, >, >=) are not defined on dicts, but they implement subset comparisons for sets. I think this is another argument pleading against | as the operator to combine two dicts.
Well, I suppose that the next proposition will be to implement the ordering operators for dicts. Because why not? Lists and numbers support them. /sarcasm/ Jokes aside, dicts have more common with sets than with sequences. Both can not contain duplicated keys/elements. Both have the constant computational complexity of the containment test. For both the size of the merging/unioning can be less than the sum of sizes of original containers. Both have the same restrictions for keys/elements (hashability).
* Regarding how to construct the new set in __add__, I now think this should be done like this:
class dict: <other methods> def __add__(self, other): <checks that other makes sense, else return NotImplemented> new = self.copy() # A subclass may or may not choose to override new.update(other) return new
AFAICT this will give the expected result for defaultdict -- it keeps the default factory from the left operand (i.e., self).
No one builtin type that implements __add__ uses the copy() method. Dict would be the only exception from the general rule. And it would be much less efficient than {**d1, **d2}.
* Regarding how often this is needed, we know that this is proposed and discussed at length every few years, so I think this will fill a real need.
And every time this proposition was rejected. What has been changed since it was rejected the last time? We now have the expression form of dict merging ({**d1, **d2}), this should be decrease the need of the plus operator for dicts.