Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > comp.lang.python > #64326 > unrolled thread

Documentation of dict views change request

Back to article view | Back to comp.lang.python

Contents

  Documentation of dict views change request Charles Hixson <charleshixsn@earthlink.net> - 2014-01-19 13:26 -0800
    Re: Documentation of dict views change request Roy Smith <roy@panix.com> - 2014-01-20 03:10 +0530
      Re: Documentation of dict views change request Chris Angelico <rosuav@gmail.com> - 2014-01-20 08:56 +1100

#64326 — Documentation of dict views change request

Could it please be clearly documented that keys(), values(), and items() 
are not writeable.  I agree that this is how they should be, but it 
would be still better if they were clearly documented as such.  The 
labeling of them as dynamic, while true, was a bit confusing here.  
(I.e., it was talking about changing their value through other means of 
access rather than directly through the returned values.)

P.S.:  Is it reasonable to return the items() of a dict in order to pass 
a read only copy of the values?

-- 
Charles Hixson

[toc] | [next] | [standalone]

#64327

In article <mailman.5728.1390166846.18130.python-list@python.org>,
 Charles Hixson <charleshixsn@earthlink.net> wrote:

> Could it please be clearly documented that keys(), values(), and items() 
> are not writeable.

We'll, technically, they are.

>>> d = {'foo': 1, 'bar':2}
>>> k = d.keys()
>>> k
['foo', 'bar']
>>> k[0] = "some other key"
>>> k
['some other key', 'bar']

Of course, this only changes the list that keys() returns, it doesn't 
affect the dictionary itself (which, I assume, is what you were really 
talking about).

Think this one through.  How *could* altering what keys() returns 
possibly affect the dict?  If it did, that means you could do something 
like:

some_dict.keys().append("some other key")

what would that mean?  You've added a key, but what's the corresponding 
value?  I will admit, the picture becomes a bit fuzzier if you consider:

some_dict.items().append(("some other key", 42))

which you could at least imagine would affect the underlying dict.

[toc] | [prev] | [next] | [standalone]

#64329

On Mon, Jan 20, 2014 at 8:40 AM, Roy Smith <roy@panix.com> wrote:
> In article <mailman.5728.1390166846.18130.python-list@python.org>,
>  Charles Hixson <charleshixsn@earthlink.net> wrote:
>
>> Could it please be clearly documented that keys(), values(), and items()
>> are not writeable.
>
> We'll, technically, they are.
>
> Of course, this only changes the list that keys() returns, it doesn't
> affect the dictionary itself (which, I assume, is what you were really
> talking about).

In Python 3, they return views, not lists. So they really are read-only.

On Mon, Jan 20, 2014 at 8:26 AM, Charles Hixson
<charleshixsn@earthlink.net> wrote:
> P.S.:  Is it reasonable to return the items() of a dict in order to pass a
> read only copy of the values?

No, because it isn't a copy. At least, not in Python 3; if you're
using Python 2, then the wording of your subject line is inaccurate,
see above.

>>> d = {"a":1,"b":2,"c":3}
>>> i = d.items()
>>> del d["b"]
>>> list(i)
[('a', 1), ('c', 3)]

If you returned i from a function and expected it to be a copy, you'd
be in for a nasty shock. Try the .copy() method for a shallow copy, or
look up deepcopy if you need a recursive copy.

ChrisA

[toc] | [prev] | [standalone]

Back to top | Article view | comp.lang.python

csiph-web