Issue #16812 has been updated by Dan0042 (Daniel DeLorme).


mrkn (Kenta Murata) wrote in #note-7:
> It may be better to change the behavior of `[*0..10][-100..100]`

I somewhat agree with that. When using range slicing most combinations make sense:

```ruby
[*0..10][0..4]   #first elements
[*0..10][-5..-1] #last elements
[*0..10][1..-2]  #middle elements
```

But a negative start with a non-negative end is quite weird. What is that operation even supposed to mean? What is it useful for?

```ruby
[*0..10][-8..8]  #????

8.times{ |i|  p (0..i) => [*0..i][-3..3]  }
{0..0=>nil}
{0..1=>nil}
{0..2=>[0, 1, 2]}
{0..3=>[1, 2, 3]}
{0..4=>[2, 3]}
{0..5=>[3]}
{0..6=>[]}
{0..7=>[]}
```

So even if `[*0..10][-100..100]` remains supported forever (there doesn't seem to be a point in breaking compatibility; see #16822), it could emit a verbose-mode warning.
And ArithmeticSequence slicing should not attempt to be consistent with that case, because it's useless to start with.
So I believe there are two useful/meaningful possibilities for `(0..20).to_a[(-5..5) % 2]`
a) `[16, 18, 20]` ignore trailing non-negative values, like `(-5..) % 2`; I think this makes the most sense
b) `[1, 3, 5]` ignore leading negative values, like python


----------------------------------------
Feature #16812: Allow slicing arrays with ArithmeticSequence
https://bugs.ruby-lang.org/issues/16812#change-86339

* Author: zverok (Victor Shepelev)
* Status: Assigned
* Priority: Normal
* Assignee: matz (Yukihiro Matsumoto)
----------------------------------------
I believe when concepts of ArithmeticSequence and `Range#%` were introduced, one of the main intended usages was array slicing in scientific data processing. So, it seems to make sense to allow this in `Array#[]`:

```ruby
ary[(5..20) % 2] # each second element between 5 and 20
ary[(0..) % 3] # each third element
ary[10.step(by: -1)] # elements 10, 9, 8, 7 ....
```

PR is [here](https://github.com/ruby/ruby/pull/3055).

My reasoning is as follows:
1. As stated above, ArithmeticSequence and `Range#%` seem to have been introduced exactly for this goal
2. Python has its slicing syntax as `begin:end:step` (with a possibility to omit either), and it seems to be well respected and used feature for data processing. So I believe it is useful, and relatively easy to integrate into existing functionality

I expect the usual "it is ugly and unreadable!" backlash. 
I don't have an incentive, nor energy, to "defend" the proposal, so I would not.



-- 
https://bugs.ruby-lang.org/

Unsubscribe: <mailto:ruby-core-request / ruby-lang.org?subject=unsubscribe>
<http://lists.ruby-lang.org/cgi-bin/mailman/options/ruby-core>