From: Robert Klemme <shortcutter@...>
Date: 2009-03-03T05:53:57+09:00
Subject: Re: Newbie question > Hash workings

On 02.03.2009 19:18, Adam Gardner wrote:
> Robert Klemme wrote:
>> On 01.03.2009 12:35, AD60 wrote:
>>> I see the use with an integer default, I specifically meant the
>>> Hash.new([]) syntax as problematic, but it probably all depends on
>>> knowing what your doing.
>> Ah.  Since you talked about "idiom" I thought you were referring to the
>> general pattern.
>>
>>> Just to detail the difference:
>>>
>>> hav = Hash.new { |h,k| h[k] = []}
>>> hav[0] += [1]
>> Actually this is very inefficient.  With this idiom you should rather do
>>
>> hav[0] << [1]
>>
>> because otherwise you'll end up creating and deleting Array instances
>> all the time without any benefit.
>>
>>> it's the difference between "{0=>[1], 1=>[]}" and "{0=>1}" result that
>>> had me wondering. Anyway thanks again, I understand how to use it and
>>> that's more than enough for now.
>> Good.

> The method I would use is this:
> 
> hav = Hash.new {[]}
> 
> This gives you the benefit of creating a new array when necessary 
> (rather than a single default array which can be destructively modified 
> in troubling ways), but means that you only get permanent effects if you 
> explicitly use assignment, so you don't get this
> 
> p hav               # {0=>[1]}
> p hav[1]            # []
> p hav               # {0=>[1], 1=>[]}
> 
> Of course, it means you can't use the << operator to stick values into 
> non-present keys, but I rarely use the << operator anyway.

You probably should get rid of that habit.  And here's why: your 
approach has no means to efficiently work with Arrays as Hash values. 
Either, you have to use += all the time which is inefficient because it 
will create a lot of garbage - or you have to explicitly test for 
presence of the key, which is less efficient than letting Hash's C 
implementation do it.  It is much more elegant to just always do 
hash[key] << value.

If you just want to make sure that you always get an Array back, you 
could as well use Hash.new([].freeze) to avoid accidentally appending to 
the single default all the time.

But it's a free country so if you like to stick with your habits...

Cheers

	robert