From: "boris_stitnicky (Boris Stitnicky)" <boris@...>
Date: 2013-10-26T17:11:37+09:00
Subject: [ruby-core:58047] [ruby-trunk - Feature #9049] Shorthands (a:b,	*) for inclusive indexing


Issue #9049 has been updated by boris_stitnicky (Boris Stitnicky).


boris_stitnicky (Boris Stitnicky) wrote:
> @mohawkjoh: Tilde is bad, too. From basic ASCII (I looked), everything is taken, except
> for ^, and ^ should be avoided because it means power in many languages. Asking for
> a new syntax feature in a maturing language is a big deal. I'd stay with Matlab/Octave's
> colon, if .. and ... are not permissible, but I'd do it outside normal Ruby syntax. Sadly,
> there is no way to overload : in regular Ruby anymore. Your options are:
> 
> 1. Settle for strings: "*, 2:4", or "*|2:4" etc. (do we hate quotes).
> 
> 2. Ask for novel %(some character) literals. But m[ %X(*|2:5) ] is still too much clutter,
> when one dreams of m[ *, 2:5 ]. (On custom literals, #8807, I voted against, as I thought
> it would burden noobs. Maybe I was wrong.)
> 
> 3. Turn to Unicode, as I do in my Pyper, via #method_missing: m.���������0���2���5��� (insane APLism)
> 
> 4. Use a block instance execced in a special object: m.slice { ��� | (2..5) }
> 
> 5. Ask for the following syntactic feature:
> 
>   m��� something ��� # ������ are angle brackets or some similar novel delimiters
> 
> meaning
> 
>   m( "something" )
> 
> This feature would rid us of the hated quotes. Then
> 
>   m��� * | 0 + 2:4 ���
> 
> would simply mean
> 
>   m( "* | 0 + 2:4" )
> 
> where the single string argument you would parse ad libitum. I personally like this
> last option best, though it is far from problem-free. The delimiters would have to be
> escaped in the statement body. So this is what I think we should ask from Matz.
> 
> Varieties would be:
> 
>   m��� some string ���, m��� some string ���, m�� some string ��, m�� some string ��
>   m ��� some string ���, m �� some string ��, etc.


----------------------------------------
Feature #9049: Shorthands (a:b, *) for inclusive indexing
https://bugs.ruby-lang.org/issues/9049#change-42626

Author: mohawkjohn (John Woods)
Status: Open
Priority: Low
Assignee: 
Category: core
Target version: 


=begin
For NMatrix, we've implemented a range shorthand which relies on Hashes: (({m[1=>3,2=>4]})), for example, which returns rows 1 through 3 inclusive of columns 2 through 4 (also inclusive). The original goal was to be able to do (({m[1:3,2:4]})) using the new hash notation, but the new hash notation requires that the key be a symbol ��� it won't accept an integer.

Whether through the hash interface or not, it'd be lovely if there were a shorthand for slicing matrices (and even Ruby Arrays) using colon. This could just be an alternate syntax for ranges, also ��� which might make more sense.

The other related shorthand we'd love to find a way to implement is the all-inclusive shorthand. It gets to be a pain to type (({n[0...n.shape[0],1...3]})) to get a submatrix (a slice), and it's really difficult to read. As a work-around, we currently use the (({:*})) symbol: (({n[:*,1...3]})). But it'd be simpler if there were a way to use a splat operator without an operand as a function argument. It might be a special case where the (({*})) is treated as a (({:*})) automatically. But this edge case might cause confusion with error messages when users make syntax errors elsewhere.

The colon shorthand is the highest priority for us.
=end



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