From: Greg Hauptmann <greg.hauptmann.ruby@...>
Date: 2008-10-27T16:41:01+09:00
Subject: Re: Hash.merge_add! extension - how does this code look?

I just noticed the code suggested might have a problem when the
original hash is empty. That is using the ability to pass a block to
'merge!" is good, but when the source hash is empty it does not seem
to trigger use of the code in the block to merge.

-----code----
  def merge_add!(h)
    self.merge!(h) do |key, existing, new|
      new = new.instance_of?(Array) ? new : [new]
      existing = existing.instance_of?(Array) ? existing : [existing]
      existing + new
    end
  end
-------------
---in console---
?> {}.merge_add!({1 => 100})
=> {1=>100}               <<== DID NOT PUT THE '100' IN AN ARRAY!!

>> {1 => 300}.merge_add!({1 => 100})
=> {1=>[300, 100]}
--------------

regards
Greg

On Mon, Oct 27, 2008 at 5:21 PM, Stefan Rusterholz <apeiros@gmx.net> wrote:
>
> Greg Hauptmann wrote:
> > thanks for highlighting this!  Is the quickest way to normalise to Array
> > via
> > the following?
> >     hashitem.instance_of?(Array) ? hashitem : [hashitem]
>
> If you're building up the hash yourself then I'd use the following
> idiom:
> collector = Hash.new { |h,k| h[k] = [] }
> collector['key1'] << 'value1'
> collector['key2'] << 'value1'
> collector['key1'] << 'value2'
> p collector
>
> Simplifies adding to and reading from the hash. But be aware that with
> the hash having a different default value than nil, you can't use 'if
> collector[key] then' to test for existence of a key anymore, you have to
> use has_key?.
>
> If you're not building the hash up yourself, then yes, I'd use what you
> wroten in an each loop and override the existing value.
>
> Regards
> Stefan
> --
> Posted via http://www.ruby-forum.com/.
>