From: Nasir Khan <rubylearner@...>
Date: 2007-06-11T10:08:19+09:00
Subject: Re: bug in facets 1.8.54

------=_Part_102890_18892848.1181524099865
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: quoted-printable
Content-Disposition: inline

Thanks Olivier.
IMHO  def initialize(compareProc=3Dlambda{|x,y| x>y},  elements=3Dnil)
could have preserved backwards compatibility for PQueue users and also
preserved semantics of passing the comparator as an argument rather than a
block.
Shall work around for now.

Thanks
Nasir


On 6/10/07, Olivier Renaud <o.renaud@laposte.net> wrote:
>
> Le samedi 09 juin 2007 22:14, Nasir Khan a =E9crit:
> > Ah! pqueue.rb was changed to be backward incompatible from facets
> > 1.8.51 -> 1.8.54
> >
> >
> > # 1.8.51
> >  def initialize(compareProc=3Dlambda{|x,y| x>y})
> >     # By default, retrieves maximal elements first.
> >     @qarray=3D[nil]; @size=3D0; @gt=3DcompareProc; make_legal
> >  end
> >
> >
> > # 1.8.54
> >  def initialize(elements=3Dnil, &block) # :yields: a, b
> >     @qarray =3D [nil]
> >     @size =3D 0
> >     @gt =3D block || lambda {|a,b| a > b}
> >     replace(elements) if elements
> >   end
> >
> >
> > Is there a way to load 1.8.51 version of one file (in this case
> > pqueue.rb) while use rest of the facets gem 1.8.54?
> > The other option is to monkey patch :-(
> > Any other option?
> >
> > - nasir
> >
>
> Hi,
>
> Yes you are right, pqueue was rewritten and the api changed a lot. In
> Facets,
> there was two different classes for the same purpose, PQueue and Heap,
> which
> both provided a heap based priority queue. The latter was broken, althoug=
h
> it
> had functionalities that pqueue lacked.
>
> So, I merged both classes into one full-featured class, with a different
> API,
> closer to the Array one and more ruby-like to me. I submitted it, and as =
I
> based my work on pqueue, it replaced the previous version of the class,
> and
> Heap was removed.
>
> Here is the notes I transmitted to Trans, who maintains Facets :
>
> * Initialization can take both initial values and comparison block, both
> optionnaly. Whereas the original PQueue take only a block, and Heap take
> only
> the initial values and needs to be subclassed in order to implement the
> comparison
> * Every methods which take an array as a parameter (initialize, push_all,
> replace) can now take any object that responds to #to_a, and are optimize=
d
> in
> the case where the given object is a PQueue itself.
> * Creating a queue from scratch (with #initialize or with #replace) is
> faster
> than for the original PQueue, by adapting the #heapify method from Heap.
> * Added consistent methods #inspect, #=3D=3D, #initialize_copy.
> * Methods are documented and unit-tested.
>
> I dropped some methods from both original classes :
> * PQueue#make_legal, useless since the PQueue is now always legal
> * PQueue#replace_top_low, which I can't figure out the purpose :)
> * PQueue#each_with_index, already provided by Enumerator
> * Heap#sort_internal, was like an inplace-sort, but it destroyed the
> PQueue.
>
> Actually, I didn't know it was already commited :)
> If you used only push, pop, clear and empty? in your client code, then th=
e
> only api change is for the instanciation method, which doesn't take a Pro=
c
> object as an argument, but a block.
>
> I see no other solution than monkey patching, if you really don't want to
> modify your client code.
>
> Regards
>
> --
> Olivier Renaud
>
>

------=_Part_102890_18892848.1181524099865--