Now that regular dicts are ordered and compact, it makes more sense for the _asdict() method to create a regular dict (as it did in its early days) rather than an OrderedDict. The regular dict is much smaller, much faster, and has a much cleaner looking repr.

Historically we would go through a deprecation period for a possibly breaking change; however, it was considered more benefit to users and less disruptive to make the update directly. See the thread starting at: https://mail.python.org/pipermail/python-dev/2019-January/156150.html

New changeset 0bb4bdf by Raymond Hettinger in branch 'master':
bpo-35864: Replace OrderedDict with regular dict in namedtuple() (bpo-11708)
0bb4bdf

Thanks for doing this, Raymond!

FYI, I found a couple of places where it still refers to OrderedDict:

1. in the doc entry in collections.rst [1]
2. in the docstring [2]

[1] 0bb4bdf#diff-a2f0632ea84b755c7ef5b9bd291c7fdfR890
[2] 0bb4bdf#diff-8a750c700ae5ac1d0a14922de83e99ccR432

Also, there's a potentially misleading detail that you might consider correcting in the text. However, making it correct might make it slightly less clear, so it's a bit of a judgement call on how important it is to be explicit here. (FWIW, I think it's worth it to some extent.)

The language reference doesn't guarantee that dictionaries are order-preserving (yet). [1] Therefore there can be (are?) Python implementations where dict isn't ordered. Users of those implementations may be confused by the docs (and docstring) saying the method returns a dict.

This could be addressed by changing those places to say something like it returns an insertion-ordered mapping. The docs entry would also have an "CPython implementation detail" part saying it's actually a dict in CPython.

Since "insertion-ordered mapping" isn't nearly as clear as "dict" (even if more correct), it may make more sense to simply say "dict". However, in that case I'd recommend at the very least to add a "CPython implementation detail" part to the docs entry which says in CPython it returns a dict, but in other implementations it may be some other order-preserving mapping.

Honestly, after having written that the latter option seems more sensible. :)

[1] https://docs.python.org/3.8/reference/datamodel.html#index-29

If you prefer, I'd be glad to open separate issues for either thing.

FWIW, both PEP-468 (kwargs order) and PEP-520 (class definition order) specify order-preserving mapping rather than dict. The main difference is that they are about language features rather than something out of the stdlib. :)

New changeset 85d83ec by Raymond Hettinger (Amador Pahim) in branch 'master':
bpo-35864: fix namedtuple._asdict() docstring (GH-11720)
85d83ec