From: Meino Christian Cramer <Meino.Cramer@...>
Date: 2004-08-17T17:51:11+09:00
Subject: Re: Q: Shifting a hash anf and array

From: "Robert Klemme" <bob.news@gmx.net>
Subject: Re: Q: Shifting a hash anf and array
Date: Tue, 17 Aug 2004 17:21:01 +0900

Hi Martin, Hi Robert,

 thank you very much for your sources !
 Great! :)

 I have an additional question below.....vvvvvv

> 
> "Martin DeMello" <martindemello@yahoo.com> schrieb im Newsbeitrag
> news:u_gUc.122944$M95.4319@pd7tw1no...
> > Meino Christian Cramer <Meino.Cramer@gmx.de> wrote:
> > >
> > >  Contents of the array before the trick:
> > >
> > >  1 2 3 4 5 6 7 8 9 0        index
> > >  A B C D E F G H I J            contents
> > >
> > >  (execute *trick*)
> > >
> > >  Contents of the array after the trick:
> > >
> > >  1 2 3 4 5 6 7 8 9 0        index
> > >  J A B C D E F G H I            contents
> >
> > array.unshift(array.pop)
> 
> or array.push(array.shift) for the other direction.
> 
> > >  And now the shift of an hash:
> > >
> > >  Contents of the hash before the trick:
> > >
> > >  A B C D E F G H I J            key
> > >  1 2 3 4 5 6 7 8 9 0        value
> > >
> > >  (execute *trick*)
> > >
> > >  Contents of the hash after the trick:
> > >
> > >  A B C D E F G H I J            key
> > >  0 1 2 3 4 5 6 7 8 9        value
> > >
> > >  Both are very similiar.
> >
> > Not really :) Hashes are inherently unordered, so the concept of a
> > rotation is a bit trickier. This works on the assumption that
> > Hash#values returns values in positions corresponding to the keys in
> > Hash#keys...
> >
> > class Array
> >   def rot1
> >     unshift(pop)
> >   end
> > end
> >
> > Hash[*(hash.keys.rot1.zip(hash.values.rot1).flatten)]
> 
> Hm, IMHO there is one "rot1" too much: you're rotating keys *and* values
> while you should only rotate one of them.  Otherwise I figure the result
> will be the same as the original hash... :-)
> 
> The other drawback of your solution is that it relys on #keys and #values
> returning elements in matching order, i.e., if you zip both you get the same
> hash.  Although that might work it makes me feel a bit wary because AFAIK
> it's not a guaranteed property of Hash (and I wouldn't want to rely on it).
> This might even change over time if the Hash implementation was changed for
> any reasons.
> 
> How about these one liners, that avoids the problem:
> 
> h = {0=>"a", 1=>"b", 2=>"c"}
> # put rotated in h2
> (h2={}).send(:[]=, *h.sort.inject {|(k1,v1),(k2,v2)| h2[k1]=v2; [k2, v1]})

  I tried to understand this...

  "h = {0=>"a", 1=>"b", 2=>"c"}" generates an hash...ok

  "(h2={}).send(:[]=" generates h2, another  hash and sends to its []=
  (element assignment) the....

  "*h.sort.inject {|(k1,v1),(k2,v2)| h2[k1]=v2; [k2, v1]})"

  ...?... I stumbled over the "*" at *h.sort.....What does it mean?


> # rotate inplace
> h.send(:[]=, *h.sort.inject {|(k1,v1),(k2,v2)| h[k1]=v2; [k2, v1]})


  ...same here...

> :-)
> 
> Note: you can omit the sort in the first example, but IMHO an order should
> be imposed on keys.  Otherwise the rotation does not make much sense IMHO.

  Absolutely correct. I was searching for something like "an array
  with indices not being integers but ordered".

> You can omit the sort in the second example, too, which seems to work
> because the set of keys of the hash doesn't change.  But this seems a bit
> hackish to me, too.  You can substitute "sort" by "dup" though.
> 
> Kind regards
> 
>     robert
> 
> 
> PS: Enumerable#inject is cool. ;-)

  PS: Undertstanding and writing such code is much cooler... !

  Ruby.use!
  Meino