Re: relationship to SRFI-25 Bradley Lucier 28 Sep 2015 21:53 UTC

On 09/28/2015 05:29 PM, Per Bothner wrote:
>
>
> On 09/28/2015 02:04 PM, Bradley Lucier wrote:
>> On 09/27/2015 01:33 AM, Per Bothner wrote:
>>>
>>> I think this has been mentioned before: the name
>>> specialized-array-share!
>>> should not end in a '!'.
>>
>> Perhaps, I don't remember the remark.
>>
>> Let me say what issue I was trying to address by adding "!" and you
>> can offer some advice.
>>
>> I've generally thought that the bang "!" in set-cdr!, vector-set!,
>> etc., was to alert the reader that something strange or nonlocal could
>> happen.
>
> I thank of it as something that causes a side-effecting modification.
> Can you think of counter-examples?
>
>> In our case, if we have
>>
>> (define domain (interval '#(0 0) '#(100 100)))
>> (define a (array->specialized-array (array domain list)))
>>
>> for which
>>
>> ((getter a) 0 0) => (0 0)
>>
>> and then
>>
>> (define b (specialized-array-share! a (interval '#(0) '#(100)) (lambda
>> (i) (values i j))))
>>
>>
>> then calling
>>
>> ((setter b) #t 0)
>>
>> will change subsequent values of
>>
>> ((getter a) 0 0) => #t
>
> How is this different from:
>
> #|kawa:1|# (define lx (list 'x 'a 'b))
> #|kawa:2|# (define ly (cons 'y (cdr lx)))
> #|kawa:3|# (set-car! (cdr ly) 'w)
> #|kawa:4|# lx
> (x w b)
>
> The specialized-array-share! is the equivalent of line 2 - i.e. we
> create a new data structure
> that share with another.

I agree completely.

(1) I don't particularly like my notation; (2) sharing mutable
structures can cause strange situations because of later mutations; (3)
these later mutations using existing primitive procedures usually have a
bang in their names to indicate that they are mutations (I think it's to
alert the code reader that the code is doing something dangerous); (4)
when using this array API there is no point in this whole process where
a bang occurs textually.

So I put a bang in the notation for the sharing step.

This is just me trying to explain my previous notational choice, and
I'll probably remove the bang in the next revision.

Brad