From: Intransition <transfire@...>
Date: 2010-10-13T06:33:14+09:00
Subject: Re: map-1.3.0



On Oct 12, 12:53 pm, "ara.t.howard" <ara.t.how...@gmail.com> wrote:
> > Hi Ara,
>
> > There is also Stash library (gem install stash -or- gem install
> > hashery).
>
> > One difference between Stash and your Map, is that it uses Strings
> > rather then Symbols for keys so they can be garbage collected.
>
> > Looks like we created these libraries for the same reason. I agree
> > with your statement "the ruby container you've always wanted". IMO
> > Ruby's built-in Hash should work like this because most of the time
> > that's really what we want. A Hash that allows for any type key could
> > be separate class instantiated explicitly.
>
> map does use strings:http://github.com/ahoward/map/blob/master/lib/map.rb#L123

Ah, your example's heavy use of symbols made me think otherwise. Very
good.

> big differences between the two libs actually: map is *always*
> ordered.  stash does not appear to be...

Stash is a subclass of Hash so it will be as ordered as actual Hash,
i.e. 1.8.x no, 1.9 yes.

> map also works recursively:
>
> ruby-1.8.7-p302 > require 'map'
>  => true
> ruby-1.8.7-p302 > m = Map.new
>  => {}
> ruby-1.8.7-p302 > m.update :k => {:a => {:x => 42}}
>  => {"k"=>{"a"=>{"x"=>42}}}
> ruby-1.8.7-p302 > m['k']['a']['x']
>  => 42
>
> stash does not:
>
> ruby-1.8.7-p302 > require 'stash'
>  => true
> ruby-1.8.7-p302 > s = Stash.new
>  => {}
> ruby-1.8.7-p302 > s.update :k => {:a => {:x => 42}}
>  => {"k"=>{:a=>{:x=>42}}}
> ruby-1.8.7-p302 > s['k']['a']['x']
> NoMethodError: undefined method `[]' for nil:NilClass
>         from (irb):14

The error is technically correct. You've assigned a Hash to a Stash as
a value, so you can't treat the Hash as if it were a Stash. Otherwise
you would never be able to assign an actual Hash as a value. To work
correctly one needs to do:

   s.update :k => Stash[:a => Stash[:x => 42]]

I realize automatic conversion might be convenient, but I would be
concerned that it could cause issues.

I might add a #normalize method though that could descend the Stash
and convert all Hash entries and sub-Hash entries to Stashes. I'll
give it some thought. Thanks.

> so there are definitely big differences between the libs at this
> point.  just FYI.

or big differences in our definition of "big"?