Re: How to sort list of sublists as per key/keys of sublist?

On 12 Jun 2009, at 9:23 am, Levan Cheishvili wrote:

I have the following structure:

object('obj1', [id='12', color='blue']).
object('obj4', [id='13', color='red', weight=120]).
object('obj21', [id='15 a', color='yellow', weight=1000, price=23]).
object('obj21', [id='16 a', color='blue', weight=200, price=230]).

While selecting a proper structure for data representation I read a newsletter: http://www.ainewsletter.com/newsletters/aix_0309.htm#ontology
and did not know that in prolog (=)/2 is not supported as (-)/2. (I am learning prolog while working on my very first project in this language).

I think it might be better to read a good book.
And while Prolog is _better_ than SQL, a lot of the stuff in
data base _theory_ (which SQL little respects) is useful to know.

How come you have two facts for obj21?

My data is like:

object(type1, id, title, [
 prop1-val1,
 prop2-val2,...
 propN-valN
]).

Data come from external source in form of XML, and I have no idea how many properties will be, but I know that there will be type, id and title which will form a unique key.

The point of an "id" is to be a unique identifier.
In fact, surely 'obj1', 'obj4', and so on *are* "ids".

With all the earnestness I possess, I tell you that life
will be *far* simpler if the things you use as unique keys
are SINGLE atoms or SINGLE numbers.  (What William Kent in
his wonderful book "Data and Reality" calls "surrogates".)

In general you will need relations mapping between surrogates
(internal keys) and external names.  So if external names are
(type,id,title) keys, use

surrogate(Atom, Type, Id, Title).
 % fd  Atom -> Type, Id, Title
 % fd Type, Id, Title -> Atom


I'm still very troubled by something being *called* an Id that is none.

If your data are provided in XML, then it simply isn't try that
your data "is (sic.) like:
       object(type1, id1, title, [ ... ])."
Your data are like *whatever you want them to be like*.

The big difference between
       prop1(key1, val1).
and
       object(key1, [...,prop1-val1,...])
is that

  If you just want to know "what is the value of prop1 for key1"
  then the first version (one relation per property) JUST RETURNS
  THE INFORMATION YOU WANT, while the second version COPIES
  EACH AND EVERY PROPERTY for that object EVERY TIME.

That is to say, using the list-of-properties version is
seriously inefficient, and that's throwing roses at it.

During my work on this project I have tried several structures. And first one was binary relation. But I found it difficult for me to write predicates for querying it,

Give specific examples.  The data structure you displayed in your
message only really handles binary relations, so there could be no
simple queries easily expressible that way that are not easier
expressed using binary relations.

as I have complex properties, like e.g. prop(obj, val1, val2, val3...) and each value may be connectedc to another property etc.

This is not clear.  Give an example.  There is no reason why the
*value* of a property cannot be a list or arbitrary term.

Why don't you tell us what your _abstract_ problem is, your
"logical data model", as it were, and _then_ we can discuss how
it might be mapped well onto Prolog.

However I think that RDF would be better in this case....I will do it in RDF when I learn it :-)

The _basic_ data model in RDF is binary relations, where admittedly
the relation names can be very long.  RDF can be written in Notation3,
after all.

Please do tell us more about what you are trying to model/do.
Let's see how well it can be done.