dict.fromkeys()

[rhettinger]

BPO	643443
Nosy	@gvanrossum, @theller, @rhettinger
Files	sequp.diff: Patch with code, docs, tests, and news fromseq.diff: Static method version fromseq2.diff: Class method version fromseq3.diff: Class method version with fixed new item fromseq4.diff: Class method version with Just's comments

^{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 = 'https://github.com/rhettinger'
closed_at = <Date 2002-11-27.07:36:15.000>
created_at = <Date 2002-11-25.10:29:38.000>
labels = ['interpreter-core']
title = 'dict.fromkeys()'
updated_at = <Date 2002-11-27.07:36:15.000>
user = 'https://github.com/rhettinger'

bugs.python.org fields:

activity = <Date 2002-11-27.07:36:15.000>
actor = 'rhettinger'
assignee = 'rhettinger'
closed = True
closed_date = None
closer = None
components = ['Interpreter Core']
creation = <Date 2002-11-25.10:29:38.000>
creator = 'rhettinger'
dependencies = []
files = ['4731', '4732', '4733', '4734', '4735']
hgrepos = []
issue_num = 643443
keywords = ['patch']
message_count = 14.0
messages = ['41753', '41754', '41755', '41756', '41757', '41758', '41759', '41760', '41761', '41762', '41763', '41764', '41765', '41766']
nosy_count = 4.0
nosy_names = ['gvanrossum', 'theller', 'rhettinger', 'jvr']
pr_nums = []
priority = 'normal'
resolution = None
stage = None
status = 'closed'
superseder = None
type = None
url = 'https://bugs.python.org/issue643443'
versions = ['Python 2.3']

[rhettinger]

Implements dict.sequpdate() as discussed on python-
dev:

def sequpdate(self, iterable, value=True):
    for k in iterable:
        self[k] = value
    return self

[jvr]

Logged In: YES
user_id=92689

I see no reason why d.sequpdate() should return self.
d.update() doesn't either.

Also, as I suggested on python-dev, a fromseq() constructor
(as a class method) would be nice:

  >>> dict.fromseq('End Quit Stop Abort'.split())
  {'End': True, 'Quit': True, 'Stop': True, 'Abort': True}
  >>>

[jvr]

Logged In: YES
user_id=92689

One more thing: I'd say the default value should be None.
This makes the idiom also useful to initialize dicts which
later get filled with proper values. This doesn't take
anything away from your original use case.

[rhettinger]

Logged In: YES
user_id=80475

Attached revised patch incorporating all of the review
comments. Changed default to None. Made a static
method instead of a classmethod because cls was not
used. Doesn't return self.

Tests, docs, and news item revised accordingly.

Ready for BDFL pronouncement.

[gvanrossum]

Logged In: YES
user_id=6380

Hm, but the convention for "alternate constructors" is that
the cls argument *is* used to determine the type of the new
instance. This way, the constructors are inherited in a
meaningful way by subclasses of dict.

[rhettinger]

Logged In: YES
user_id=80475

Done! Revise patch attached.

It is now a class method and uses cls for the type of the
new instance. Added tests to demonstrate meaningful
inheritance.

[theller]

Logged In: YES
user_id=11105

It seems the code and the docs are out of sync in
fromseq2.diff.

The docs say "fromseq(sequence, value=True)" while the
code does "fromseq(sequence, value=None)"

[rhettinger]

Logged In: YES
user_id=80475

Fixed news item.

[jvr]

Logged In: YES
user_id=92689

• I'm not sure whether PyObject_CallObject() is the right
  idiom here. I would have expected to call cls->tp_new (if
  not NULL), but I'm not sure either way.
• I would get rid of the loop counter: you're not using it.
  for (;;) will do nicely
• you're XDECREF'ing values of which you know they're not NULL
• idem XINCREF. Also: I'm pretty sure you should incref d at
  the end: I think it currently leaks.

[jvr]

Logged In: YES
user_id=92689

dang, that should've read "you should NOT incref d at the end".

[rhettinger]

Logged In: YES
user_id=80475

Okay, fixed decrefs, eliminated i; used for(;;).
Left in PyObject_CallObject because it's clean.
Thanks.

[jvr]

Logged In: YES
user_id=92689

Btw. I was looking for more info on METH_CLASS and came
across this thread on python-dev:

http://mail.python.org/pipermail/python-dev/2002-April/023566.html
which resulted in this bug report:
http://www.python.org/sf/548651
which has not been resolve. I'll add a comment there, too.

[gvanrossum]

Logged In: YES
user_id=6380

I'm ok with this, but I haven't had time to review the code.

[rhettinger]

Logged In: YES
user_id=80475

After more code review, added PyDict_Check.

Also, discussion on python-dev showed a need to rename
the method to indicate that keys were being used rather
than items. The sequence part of the name was dropped
because any iterable will do as an argument.

Checked-in as:
Misc/NEWS 1.548
Objects/dictobject.c 2.132
Lib/test/test_types.py 1.41
Doc/lib/libstdtypes.tex 1.111