types.SimpleNamespace should preserve attribute ordering (?)

[ericsnowcurrently]

BPO	39075
Nosy	@rhettinger, @ericsnowcurrently, @serhiy-storchaka, @1st1, @ZackerySpytz, @miss-islington, @tirkarthi, @shihai1991
PRs	bpo-39075: types.SimpleNamespace no longer sorts attributes in its repr #19430

^{Note: these values reflect the state of the issue at the time it was migrated and might not reflect the current state.}

Show more details

GitHub fields:

assignee = None
closed_at = <Date 2020-05-16.01:30:01.391>
created_at = <Date 2019-12-17.16:21:21.111>
labels = ['interpreter-core', 'type-feature', '3.9']
title = 'types.SimpleNamespace should preserve attribute ordering (?)'
updated_at = <Date 2020-05-16.01:30:01.390>
user = 'https://github.com/ericsnowcurrently'

bugs.python.org fields:

activity = <Date 2020-05-16.01:30:01.390>
actor = 'eric.snow'
assignee = 'none'
closed = True
closed_date = <Date 2020-05-16.01:30:01.391>
closer = 'eric.snow'
components = ['Interpreter Core']
creation = <Date 2019-12-17.16:21:21.111>
creator = 'eric.snow'
dependencies = []
files = []
hgrepos = []
issue_num = 39075
keywords = ['patch']
message_count = 9.0
messages = ['358553', '358563', '358613', '358785', '358786', '358793', '358913', '359119', '368995']
nosy_count = 8.0
nosy_names = ['rhettinger', 'eric.snow', 'serhiy.storchaka', 'yselivanov', 'ZackerySpytz', 'miss-islington', 'xtreak', 'shihai1991']
pr_nums = ['19430']
priority = 'normal'
resolution = 'fixed'
stage = 'resolved'
status = 'closed'
superseder = None
type = 'enhancement'
url = 'https://bugs.python.org/issue39075'
versions = ['Python 3.9']

[ericsnowcurrently]

types.SimpleNamespace was added in 3.3 (for use in sys.implementation; see PEP-421), which predates the change to preserving insertion order in dict. At the time we chose to sort the attributes in the repr, both for ease of reading and for a consistent output.

The question is, should SimpleNamespace stay as it is (sorted repr) or should it show the order in which attributes were added?

On the one hand, alphabetical order can be useful since it makes it easier for readers to find attributes, especially when there are many. However, for other cases it is helpful for the repr to show the order in which attributes were added.

FWIW, I favor changing the ordering in the repr to insertion-order. Either is relatively trivial to get after the fact (whether "sorted(vars(ns))" or "list(vars(ns))"), so I don't think any folks that benefit from alphabetical order will be seriously impacted.

[serhiy-storchaka]

We usually do not iterate all attributes, and dir() always sort attribute names.

[rhettinger]

+1 for dropping the sort. Also, the repr should be made to round-trip with eval().

[shihai1991]

IMHO, dropping the sort should be a default behavior. If some user need
this feature, maybe we could supply a param to open the sort function or not?

[rhettinger]

@haisai SimpleNamespace can't use keyword arguments because those are already used to pass in data. Also, we want to keep it simple; hence, the name :-)

[shihai1991]

@Raymond OK, copy that.

[ericsnowcurrently]

IMHO, dropping the sort should be a default behavior. If some user need
this feature, maybe we could supply a param to open the sort function or not?

Consider opening a separate issue (or start a thread on python-ideas)
about adding a more sophisticated implementation to the stdlib's
"reprlib" module. If you do so then please draw from the
argparse._AttributeHolder implementation.

[shihai1991]

Consider opening a separate issue (or start a thread on python-ideas)
about adding a more sophisticated implementation to the stdlib's
"reprlib" module. If you do so then please draw from the
argparse._AttributeHolder implementation.

FWIW, i created a new bpo: https://bugs.python.org/issue39173

[miss-islington]

New changeset 6b6092f by Zackery Spytz in branch 'master':
bpo-39075: types.SimpleNamespace no longer sorts attributes in its repr (GH-19430)
6b6092f