From: Steven Lumos <steven@...>
Date: 2007-03-06T10:44:44+09:00
Subject: Re: [PATCH] Join with block

"Farrel Lifson" <farrel.lifson@gmail.com> writes:
> On 05/03/07, Steven Lumos <steven@lumos.us> wrote:
>> Why bother passing sep to the block?
>
> If the block is a pre-existing proc object there's no guarantee it
> will have access to whatever is passed to the join method as the
> separator.

Okay, but why does that matter? Hopefully I'm not just using an overly
simple, but why is something like

 [1,3,2].join(' (X) ') do |sep,a,b|
   sep.sub('X', a < b ? 'less' : 'greater or equal')
 end

preferable to

 [1,3,2].join {|a,b| a < b ? ' (less) ' : ' (greater or equal) '}

>> Can your implementation handle [1,2,3].join {...} => "1, 2 and 3"?
>
> If the elements in the array are unique it's pretty simple:
>  ary = [1,2,3]
>  ary.join{|sep,a,b| !ary.last.eql?(b) ? ", " : " and "}
>
> For a case where elements are repeated (but sequences are not):
>  ary=[1,2,1]
>  ary.join{|sep,a,b| !ary.last(2).eql?([a,b]) ? "," : " and "}
>
> For more complex cases where sequences are repeated ([1,2,1,2]) you're
> probably going to need some extra logic:
>  ary=[1,2,1,2];idx = 0
>  ary.join{|sep,a,b| ((idx+=1) < ary.length-1) ? ", " : " and "}
>
> Farrel

After reading the archives, I think there are too many conflicting
opinions about what a block argument to #join means.  

*I* think it should help with certain common cases for English where
currently you have type a lot more code to get something that ought to
(IMO) be simple.

Your implementation on the other hand focuses on handling complex
relationships between elements, and doesn't help me at all.  :-)
I wouldn't prefer your example to something like:

  a = [1,2,3]
  "#{a[0..-2].join(', ')} and #{a[-1]}"  #=> "1, 2 and 3"

But then the previous discussions seem to be more interested in
combining #map with #join, although there definitely seems to be
greater interest in your problem domain than mine.

So what's missing is a convincing argument for which of these cases
deserve special treatment and why.  I consider my case to be a small
special case within the traditional domain of #join.  But I'm also
sympathetic to arguments that however close, it's still outside the
domain.

Steve