popkey() method for dictionaries?

[Fernando Pérez]

John Hunter wrote:

>>>>>> "Fernando" == Fernando Pérez <fperez528 at yahoo.com> writes:
> 
>     Fernando> Sure. First, on principle: the exception option is still
>     Fernando> there, if no default is provided. What giving a default
>     Fernando> buys you is not having to trap the exception
>     Fernando> yourself. If you want a missing key to generate an
>     Fernando> exception, simply don't give a default and that's
>     Fernando> it.
> 
> But then you break the symmetry with get, which provides None as a
> default if no default is provided.  Aren't you now back in danger of
> violating the principle of least surprise?

Mmmh. I hadn't used get without default in a long time, and I'd forgotten its 
not-found return was a None instead of an exception. I would argue _that_ 
violates the 'explicit is better than implicit' rule of python, but it's a 
bit late to change that one :)

In general, since these are all wrappers around dict[key] access operations, 
and at the most basic level such an access to a non-existent key _will_ 
trigger an exception, I'd argue that the default behavior of all these 
functions (get, pop, popitem) should be to raise an exception. In this way 
they would most naturally mimic the underlying semantics, without magic 
behaviors. On the other hand, I think it is _useful_ syntactic convenience to 
have a way of explicitly overriding the exception triggering, for the reasons 
Alex Martelli and I argued elsewhere in this tread.

Unfortunately these functions (get/pop/popitem) have already been written with 
(it seeems to me) inconsistent interfaces, and we'll just get used to keeping 
track of who does what with post-it notes.

Cheers,

f.