FW: Re: Nil in an index

[Chris Muller]

Interesting idea.  Should acceptNulls mean, "ignoreNulls" or are you
suggesting the index should support "null" (can we say nil?) as an
indexable position?  Presumably the lowest possible hash value, 0
(zero)?

Probably the former because usually the presence of a non-nil value is
what is interesting about the object.  The ones with nil are not
interesting w.r.t. that index.

Right?


On Jan 14, 2008 4:15 AM,  <brent at zamail.co.za> wrote:
> Hi,
>
> Actually the MaDateIndex was used in some real work.
>
> It mught be prudent to accept a tested patch to the MaDateIndex that,
> conditionally, allows null date values.
>
> Since this null feature will probably crop up in a lot of index
> classes, it might be prudent to agree on a naming convention.
>
> E.g.
>      index := MaDateIndex new
>                  acceptNulls;
>                  yourself.
>
> Chris should have the final word on the naming IMHO.
>
> Cheers
>
> Brent
>
>
> From: asqueaker at gmail.com
> To: nmanzanos at gmail.com,
> Subject: FW: Re: Nil in an index
> Date: Sat, 12 Jan 2008 22:48:11 -0500
>
> The indexes provided with Magma are really just simple examples.
> Robust applications will almost certainly want to define their own
> custom indexes.
>
> That is your path to a clean solution.  Make your own DateIndex that
> accepts a nil.  Just override #canIndex: in addition to
> #indexHashForIndexObject:, et al...
>
>
> On Jan 11, 2008 9:43 AM, Norberto Manzanos <nmanzanos at gmail.com>
> wrote:
> > Hi!
> >
> > Is it a good idea avoid nil in magma indexes?
> > This question rised when I tried MaDateIndex. Until this moment, I
> used
> > String or Integer indexes having care to answer allways a String
> (an empty
> > string if the object is nil) and an  Integer (a zero if nil). But
> with dates
> > there is a conceptual problem: there is not a null date. In some
> > applications, business stuff, for example, you can think in an
> ovbiously
> > unused date (01-01-1500, for example) that acts as a null date.
> It's a
> > little dirty, but it would work. But in other applications every
> possible
> > date has a meaning. And if the object could be inexistent in some
> cases, you
> > have no choice: it's nil.
> > That's my case, and I'm forced to use a StringIndex, converting the
> dates to
> > Strings, answering an empty string if the object is nil. That's
> really
> > dirty.
> > Would'nt be better if magma indexes just ignore nil objects? Is it a
> > requirement for magma that every object from a MagmaCollection must
> be
> > present in all indexes?
> > I hope there is something I misunderstood and there is a more
> elegant
> > solution.
> >
> > Regards
> > Norberto
> >
> >
> > _______________________________________________
> > Magma mailing list
> > Magma at lists.squeakfoundation.org
> > http://lists.squeakfoundation.org/mailman/listinfo/magma
> >
> >
> _______________________________________________
> Magma mailing list
> Magma at lists.squeakfoundation.org
> http://lists.squeakfoundation.org/mailman/listinfo/magma
>
>
> _______________________________________________
> Magma mailing list
> Magma at lists.squeakfoundation.org
> http://lists.squeakfoundation.org/mailman/listinfo/magma
>