On 10/23/2015 11:29 AM, Bradley Lucier wrote:
> On 10/21/2015 04:20 PM, Arthur A. Gleckler wrote:
>> Hi, Bradley. I'm sorry to hear that you're considering withdrawing SRFI
>> 122, but I completely understand. It's certainly up to you.
>>
>> I should point out that the 60- and 90-day deadlines haven't been
>> enforced for a while. I've been using them less as deadlines than as
>> guidelines, hoping that they would prod discussion and help push toward
>> reaching consensus. If you think that more time will help, please do
>> take it.
>
> Thank you, I'll take advantage of your indulgence to polish and reorganize the material a bit more. (Honestly, I hadn't realized how unpolished and disorganized the original proposal was until after trying to accommodate some of the comments.)
FWIW, next month I plan to spend some time on Kawa's array
feature. The lower-level (Java) design is mostly done, but
there are still some issues; the Scheme API is even more open.
My hope is that SRFI-122 functions can be implemented as a
simple set of wrappers over the Kawa data structure. So I
would welcome going slow SRFI-122
I expect Kawa programs will mostly use function call notation for
array indexing, as well as using "range" objects. However, I
will also support more portable interfaces, including SRFI-25,
and presumably SRFI-122.
(a i j k) ; apply indexes i j k to array a
(set! (a i j k) new-value)
[i <: j] ; integer range from i to j exclusive
[i <=: j] ; integer range from i to j inclusive
(a [i <: j]) ; a slice
(a [i1 i2 i3 i3] j [k1 <: k2]) -> an array ax such that
(ax x y) == (a ([i1 i2 i3 i3] x) (+ k1 y))
I.e. APL where an index can be integer arrays, vectors,
ranges, or plain integers.
Of course SRFI-122 shouldn't specify "function call notation"
for array indexing, but it might be worth considering having
an array-ref convenience function that allows arrays as indexes.
--
--Per Bothner
xxxxxx@bothner.com http://per.bothner.com/