Обсуждение: Getting relation/attribute names from inside UDT input/output function

Hello,

I wrote the UDT input/output functions for my complex type
and everything is working well.
But now, in the input function, I need to get the relation/attribute
names where a new object is being inserted.

PG_FUNCTION_INFO_V1(complex_in);
Datum complex_in(PG_FUNCTION_ARGS) {
   char *arg1 = PG_GETARG_CSTRING(0);
   char *relname = ???
   char *attribname = ???
   ...
   PG_RETURN_POINTER(result);
}

Does anybody have an idea of how I get that?

Thanks in advance

Humberto Razente
hlr@icmc.usp.br

On Wed, May 10, 2006 at 09:58:04AM -0300, Humberto Luiz Razente wrote:
> I wrote the UDT input/output functions for my complex type
> and everything is working well.
> But now, in the input function, I need to get the relation/attribute
> names where a new object is being inserted.

Why would the type's input function need to know that?  What are
you trying to do?

--
Michael Fuhr

"Humberto Luiz Razente" <hlr@icmc.usp.br> writes:
> I wrote the UDT input/output functions for my complex type
> and everything is working well.
> But now, in the input function, I need to get the relation/attribute
> names where a new object is being inserted.

Since, in general, there *is* no such place where the value would be
inserted, this is impossible.  Consider

    SELECT 'foo'::yourtype

Better rethink what you're trying to accomplish.

            regards, tom lane

[Please copy the mailing list on replies so others can contribute
to and learn from the discussion.]

On Wed, May 10, 2006 at 02:50:26PM -0300, Humberto Luiz Razente wrote:
> I'm trying to index data characteristics extracted from multimedia
> files (like a color distribution histogram from an image), in
> order to make k-nearest neighbor queries (similarity-based queries).
> The problem is that I can have a variable number of characteristics (its
> called adimensional features), so I cannot index with R-tree
> like GIST. I already have the tree implemented in C, so I would
> like to build a tree for each complex type in each
> attribute/relation. Am I going in the wrong direction?

I'm not familiar with the problem domain but maybe somebody else
on the mailing list is.  Could you explain a bit more, perhaps with
an example showing how you're currently using the type and what
you'd like to be able to do with it (e.g., a hypothetical table,
data set, query, and query result)?

--
Michael Fuhr

On Wed, May 10, 2006 at 09:12:33PM -0600, Michael Fuhr wrote:
> On Wed, May 10, 2006 at 02:50:26PM -0300, Humberto Luiz Razente wrote:
> > I'm trying to index data characteristics extracted from multimedia
> > files (like a color distribution histogram from an image), in
> > order to make k-nearest neighbor queries (similarity-based queries).
> > The problem is that I can have a variable number of characteristics (its
> > called adimensional features), so I cannot index with R-tree
> > like GIST. I already have the tree implemented in C, so I would
> > like to build a tree for each complex type in each
> > attribute/relation. Am I going in the wrong direction?
>
> I'm not familiar with the problem domain but maybe somebody else
> on the mailing list is.  Could you explain a bit more, perhaps with
> an example showing how you're currently using the type and what
> you'd like to be able to do with it (e.g., a hypothetical table,
> data set, query, and query result)?

I'm not entirely sure, but this seems a similar issue to what tsearch
had (indexing documents). CVS HEAD has a new index type called GIN
(Generalised Inverted Index) which may be closer to your needs.

Have a nice day,
--
Martijn van Oosterhout   <kleptog@svana.org>   http://svana.org/kleptog/
> From each according to his ability. To each according to his ability to litigate.