[ruby-talk:6611] Re: best way to interleaf arrays?

On Tue, 28 Nov 2000, Dave Thomas wrote:

> David Alan Black <[e-mail]> writes:
> 
> > What about a to_h method for arrays?  I mention this mainly for the
> > sake of symmetry with other to_* methods.  Then again, such a method
> > may violate some principle of design: it does seem slightly wrong for
> > Array to "know" how to generate a hash (whereas not strange for Hash
> > to know how to create itself from an array).
> 
> That was my thinking (if you can call it that ;-) It seems reasonable
> for a more complex class to know about a more primitive one, but not
> the other way around.

Although.... what about the cases where two types have to_* methods
for each other?  (Array#to_s, String#to_a, for example)  Are all such
types considered lateral to each other, on the scale of complexity?
Or maybe the difference is that a-to-s/s-to-a don't introduce new
data, whereas hashing an array against a bunch of 1's adds a bunch
of 1's.

There's another way to look at it, though -- somewhat Devil's
advocate, but anyway.  Arrays and hashes are both indexed collections
of elements, with the one singularity that the indices of arrays are
integers.  From that perspective, a case could be made that changing
an array to a hash really wouldn't involve anything that wasn't part
of the array already.  But this would depend on doing it, not as we've
been discussing so far, but like this:

   class Array 
     def to_h
       h = {}
       each_index { |i| h[i] = at i }
       h
     end
   end

   %w{ list of words }.to_h  # => {0=>"list", 1=>"of", 2=>"words"}

One could then invert the hash to get at least one version of the
original effect -- namely, the original array elements as hash keys
and all the values true.  (Even zero!)

In short: just as a string knows enough to break itself down into an
array, an array could perhaps reasonably know enough to present itself
as a hash, if all that's involved is lining up the indices and the
elements and letting them be called something else (i.e., without
interpolating anything new).

Well, that's about the best case I can make for this.  I'm not as
unconvinced by it as I thought I would be :-)  Comments encouraged.


David

-- 
David Alan Black
home: [e-mail]
work: [e-mail]
Web:  http://pirate.shu.edu/~blackdav