From: David MacMahon <davidm@...>
Date: 2013-10-23T23:15:02-07:00
Subject: [ruby-core:58012] Re: [ruby-trunk - Feature #9049] Shorthands (a:b,	*) for inclusive indexing


On Oct 23, 2013, at 10:36 PM, boris_stitnicky (Boris Stitnicky) wrote:

> @david_macmahon: I do not think that the wish to make a : b an alias of a .. b
> can fit into the language anymore. "x > 0 ? 1 : 2 : 3" could be ambiguously
> parsed as "x > 0 ? ( 1 : 2 ) : 3" or as "x > 0 ? 1 : ( 2 : 3 )"...

Good point.  How about if this new notation limited only to uses within square brackets: `[a:b]`?  There are two distinct cases (maybe more?).  One case is where this used as a literal like `range = [a:b]` and the other where it is used in #[] like `array[a:b]`.

It may seem strange for `[...]` to create a non-Array literal, but the usage would be pretty straightforward:

```ruby
[a:b]      => a..b
[a:b, c:d] => [a..b, c..d]
[[a:b]]    => [a..b]
```

With this case, your example of `x > 0 ? 1 : 2 : 3` would not be syntactically valid, but it could, for example, be re-written as `x > 0 ? [ 1 : 2 ] : 3` (excessive spaces added for emphasis), which would be equivalent to `x > 0 ? 1 .. 2 : 3`.

The two cases are mutually exclusive (unless there is some parsing constraint I'm unaware of).  The #[] case would be the more useful of the two (by far, IMHO).  The [a:b] case doesn't really offer much over the a..b syntax (except for possible consistency with the #[] case).

Dave