Per Bothner <per@bothner.com> wrote:

> On 05/25/2013 02:06 PM, Vassil Nikolov wrote:
> >
> >
> > Per Bothner <per@bothner.com> wrote:
> >
> >> On 05/25/2013 01:50 PM, Vassil Nikolov wrote:
> >>>
> >>> Per Bothner <per@bothner.com> wrote:
> >>>
> >>>> ...
> >>>> a bag of T is just
> >>>> a minor optimization of a map (hash-table) from T to integers.
> >>>
> >>>     Except when equality only depends on
> >>>     parts of the elements.
> >>
> >> I don't understand this comment.
> >
> >    If a bag's notion of element equality only
> >    takes into account a part of each element
> >    and ignores the rest of the element,
> >    keeping a count instead of the elements
> >    themselves would lose information.
>
> I don't the think the bag API supports this.  I.e. using
> a bag will also lose information, unless the bag API is
> clearly specified to support this use case.  And I don't
> believe it does.

  In any case, I believe it will be better if it
  does, or it would be better if it did.  My 2e-2.

> ...
> To support this use-case you could use a map that maps
> T to list-of-T.  So there is still no strong need for a
> separate bag API.

  I'd rather abstain on the latter point, at
  least for now.

  ---Vassil.

  2013-05-26.

--
Would you like your metaphors shaken or stirred?

Vassil Nikolov | Васил Николов | <vnikolov@pobox.com>

_______________________________________________
Scheme-reports mailing list
Scheme-reports@scheme-reports.org
http://lists.scheme-reports.org/cgi-bin/mailman/listinfo/scheme-reports