s[i:j:t] = t stipulation

[Neil Cerutti]

On 2007-11-20, Terry Reedy <tjreedy at udel.edu> wrote:
> "Neil Cerutti" <horpner at yahoo.com> wrote in message 
> news:slrnfk65hp.1mk.horpner at FIAD06.norwich.edu...
>| s[i:j:t] = t (1) t must have the same length as the slice it is 
> replacing.
>
> This is essentially the same rule as requiring a proper length
> of t for
>
> a,b,c = t # for whatever number of targets
>
> And people have made similar suggestions as below for that case
> also.
>
>|  Why?
>
> A mismatch could be intentional or accidental.  In most cases
> of this sort, Python assumes 'accident', especially when intent
> can easily be indicated otherwise.

Thanks. Assignment to slices are a convenient way to insert,
assign, and delete elements, but extended slices are only good
for assignment. Perhaps I was searching for consistency in the
wrong place, though.

>| >>> def foo():
>| ...   while True:
>| ...     yield 'a'
>| ...
>| >>> foo()
>| >>> x = range(10)
>| >>> x[::2] = foo()
>|
>| This is infinite loop due to Python building a sequence out of
>| the iterator to check its length.
>|
>| I think it might be more useful for
>|
>| x[::2] = foo()
>|
>| to result in an x of
>|
>| ['a', 1, 'a', 3, 'a', 5, 'a', 7, 'a', 9]
>|
>| In other words, take (j-i)//k elements from t for abs(k) != 1.
>
> Use the appropriate itertools function to take the proper
> number of elements.

And anyway my math was quite wrong. :(

>| A problem, though, arises when t is too short--the sequence
>| could be corrupted before an exception is thrown if you omit the
>| length check.
>|
>| So you'd also have to define
>|
>| x[::2] = 'aaa'
>|
>| as resulting in
>|
>| ['a', 1, 'a', 2, 'a', 3, 5, 7, 9]
>
> No, it should be defined as resulting in
>
> ['a', 1, 'a', 2, 'a', 3, None, 5, None, 7, None, 9] # ;-)

I thought deletion of elements would be more similar to slice
assignment, e.g.:

x[5:] = [] 
  --> [0, 1, 2, 3, 4]
  -/ /-> [0, 1, 2, 3, 4, None, None, None, None, None]

> Or better yet, require the programmer to specify by modifying
> either the target or source spec, as is done now.

It seems a shame to accept iterators but to build a sequence out
of them, if it can be avoided. But if there's too much confusion
about what it should mean, I guess that kills the idea.

-- 
Neil Cerutti