Inconsistent PEP 0448 implementation

[vedgar]

BPO	24569
Nosy	@benjaminp, @vedgar

^{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 2015-07-05.15:39:29.900>
created_at = <Date 2015-07-05.13:05:43.402>
labels = ['type-bug']
title = 'Inconsistent PEP 0448 implementation'
updated_at = <Date 2015-07-05.16:35:23.673>
user = 'https://github.com/vedgar'

bugs.python.org fields:

activity = <Date 2015-07-05.16:35:23.673>
actor = 'veky'
assignee = 'none'
closed = True
closed_date = <Date 2015-07-05.15:39:29.900>
closer = 'python-dev'
components = []
creation = <Date 2015-07-05.13:05:43.402>
creator = 'veky'
dependencies = []
files = []
hgrepos = []
issue_num = 24569
keywords = []
message_count = 5.0
messages = ['246312', '246319', '246322', '246323', '246324']
nosy_count = 3.0
nosy_names = ['benjamin.peterson', 'python-dev', 'veky']
pr_nums = []
priority = 'normal'
resolution = 'fixed'
stage = 'resolved'
status = 'closed'
superseder = None
type = 'behavior'
url = 'https://bugs.python.org/issue24569'
versions = ['Python 3.5']

[vedgar]

It seems the consequences of PEP-0448 weren't really thought through. :-/ (And BTW why isn't it in "What's new in Python 3.5"? I know there is a file with full details, but I guess this really should be notable enough.)

{0:1, **{0:2}, 0:3, 0:4}

Do you know what is the value of that dict? And does it make sense to you? It surely doesn't make sense to me (though I understand the implementation). I know things are really subtle and even Guido gets it wrong (https://bugs.python.org/msg234413), even without PEP-0448, but this particular instance is horrible.

ValueError would be perfect, I'd be ok with 4, I'd shrug on 1, I'd frown on 2, but I _scream_ on 3. Please fix this until it's too late and fictional "backward compatibility" starts to freeze the wrong behaviour.

[python-dev]

New changeset a4df0fe62b46 by Benjamin Peterson in branch '3.5':
set items in dict displays from left to right (closes bpo-24569)
https://hg.python.org/cpython/rev/a4df0fe62b46

New changeset 75852d90c225 by Benjamin Peterson in branch 'default':
merge 3.5 (bpo-24569)
https://hg.python.org/cpython/rev/75852d90c225

[vedgar]

Benjamin, you're my hero. :-)

I'm not really at home in C source... is it possible that you have also changed

{0:1, 0:2}

to be {0:2} (as opposed to {0:1} as it is now)? I'm completely fine with that and find it more logical (and as I said in the previous message, it matches BDFL's mental model, which is the real reference implementation of Python:), but I'm afraid backward compatibility zealots will ruin the fun.

[benjaminp]

Python 3.3.5 (default, Aug 19 2014, 23:45:33) 
[GCC 4.7.3] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> {0:1, 0:2}
{0: 2}

Python 2.7.9 (default, Dec 24 2014, 23:52:11) 
[GCC 4.8.3] on linux2
Type "help", "copyright", "credits" or "license" for more information.
>>> {0:1, 0:2}
{0: 2}

[vedgar]

Ah, so it was broken _only_ on 3.5. That should teach me not to uninstall Py3.x as soon as Py3.x+1 comes out. :-)

Ok, thank you very much.