odict_items and dict_items' repr's don't match OrderedDict's and dict's

[jab]

Bug report

Bug description:

As of #101446, OrderedDict's repr now looks like dict's, but their ItemsViews were not updated to match.

❯ python3.12
Python 3.12.0 (main, Oct  2 2023, 12:03:24) [Clang 16.0.6 ] on darwin
Type "help", "copyright", "credits" or "license" for more information.
>>> from collections import OrderedDict
>>> OrderedDict({1: -1, 2: -2})  # no longer looks like a list of pairs
OrderedDict({1: -1, 2: -2})
>>> OrderedDict({1: -1, 2: -2}).items()  # still looks like a list of pairs
odict_items([(1, -1), (2, -2)])
>>> {1: -1, 2: -2}.items()  # ditto
dict_items([(1, -1), (2, -2)])

CPython versions tested on:

3.12

Operating systems tested on:

No response

[serhiy-storchaka]

And what is wrong with it? It is consistent with keys view and values view. Iterating an items view emits pairs.

[jab]

Thanks for taking a look. I wasn't sure if this was a deliberate decision, or if maybe just no one thought about it at the time #101446 was introduced (I didn't see any discussion of this there or in the associated PR). If you feel it's worth it, please ping Raymond or anyone else involved in the previous changes to weigh in here.

Regarding your consistency point:

Since ItemsViews contain items (not single values like Keys- or ValuesViews), is it better to be consistent with dict's repr, which also displays items as opposed to single values?

Here's what that would look like in context:

>>> {1: 1, 2: 2}
{1: 1, 2: 2}

>>> {1: 1, 2: 2}.items()
dict_items({1: 1, 2, 2})

>>> OrderedDict({1: 1, 2: 2})
OrderedDict({1: 1, 2: 2})

>>> OrderedDict({1: 1, 2: 2}).items()
odict_items({1: 1, 2: 2})


>>> {1: 1, 2: 2}.keys()
dict_keys([1, 2])

>>> OrderedDict({1: 1, 2: 2}).keys()
odict_keys([1, 2])

This would also provide the nice property that all view repr's would consistently have only 2 levels of nested brackets, instead of 3 in the current case of only the items views.

Independent of any consistency-based rationale, more layers of nested brackets are noisier and not as easy to read compared to fewer layers, so that's another reason to consider this proposed change.

[serhiy-storchaka]

On one hand, the repr will be smaller and with less nesting level.

On other hand, it will be less obvious that dict_items represents a collection of pairs, and it repr will be more distinct from reprs of dict_keys and dict_values. It may be worse for education.

It is a very minor issue and I have no strong opinion. cc @rhettinger

[rhettinger]

It is a very minor issue and I have no strong opinion. cc @rhettinger

I'm think the status quo is reasonable and we should leave it as-is.

It was a perfectly reasonable choice for OrderedDict() and its views to follow what dict() does. Afterall, `OrderedDict is a dict subclass.

In contrast, the ItemsView ABCs can't make any assumptions about the underlying mapping, so it is reasonable to just show it's __repr__ so that a programmer can see what was wrapped. The underlying mapping might be virtual, case-insensitive, or have some other interesting behavior that wouldn't be well expressed by only showing a list of pairs.