Ferret talk - Oct 2006 - Need help for coding an extension to ferret

Hi,

i''m working on a project using Ferret for indexing it''s datas.
I''m very
happy with it but i need to code an extension to implement a .to_json 
method to TopDocs class, because ruby''s json implementation is really 
really slow...

It''s my second (the first was the tutorial :/ ) ruby C extension, so
i''m
not really at ease with ruby C bindings, even with the C experience...

Here is my problem :

I would like to load each document from ids in my TopDoc object, to make 
the json string myself from that, but i don''t know how to load my 
documents from this class... it''s really weirdo to me actually :\ hope 
somebody can help !

Here is my code, situated in r_search.c :

static VALUE
frt_td_to_json(VALUE self)
{
	int i;
	VALUE rhits = rb_funcall(self, id_hits, 0);
	VALUE rhit;
	const int len = RARRAY(rhits)->len;
	long pos;

	for (i = 0; i < len; i++)
	{
		rhit = RARRAY(rhits)->ptr[i];
		pos = FIX2INT(rb_funcall(rhit, id_doc, 0));
//
//   HERE I WOUlD LIKE TO LOAD THE DOCUMENTS, ID IS THE GOOD DOC_ID..
//   I WOULD LIKE TO GET THIS IN FACT (ruby):
//   doc_id = INDEX.search(''query'').hits.first
//   INDEX[doc_id].load <== THAT''S WHAT I WOULD LIKE TO GET !
//
	}
	rstr = rb_str_new2(str);
	free(str);
	return (argv[0]);
}

of course, i''ve bound the method to the good object etc...

Hope somebody''ll help !
Thank you by advance,
Jeremie ''ahFeel'' BORDIER

-- 
Posted via http://www.ruby-forum.com/.

Damn, i''ve forgotten some debug / test code in the past :P
of course, you don''t have to care about these lines :
> 	rstr = rb_str_new2(str);
> 	free(str);
> 	return (argv[0]);
> }
> Thank you by advance,
> Jeremie ''ahFeel'' BORDIER
-- 
Posted via http://www.ruby-forum.com/.

On 10/11/06, ahFeel <ahfeel at rift.fr> wrote:
> Hi,
>
> i''m working on a project using Ferret for indexing it''s
datas. I''m very
> happy with it but i need to code an extension to implement a .to_json
> method to TopDocs class, because ruby''s json implementation is
really
> really slow...
>
> It''s my second (the first was the tutorial :/ ) ruby C extension,
so i''m
> not really at ease with ruby C bindings, even with the C experience...
>
> Here is my problem :
>
> I would like to load each document from ids in my TopDoc object, to make
> the json string myself from that, but i don''t know how to load my
> documents from this class... it''s really weirdo to me actually :\
hope
> somebody can help !
>
> Here is my code, situated in r_search.c :
>
> static VALUE
> frt_td_to_json(VALUE self)
> {
>         int i;
>         VALUE rhits = rb_funcall(self, id_hits, 0);
>         VALUE rhit;
>         const int len = RARRAY(rhits)->len;
>         long pos;
>
>         for (i = 0; i < len; i++)
>         {
>                 rhit = RARRAY(rhits)->ptr[i];
>                 pos = FIX2INT(rb_funcall(rhit, id_doc, 0));
> //
> //   HERE I WOUlD LIKE TO LOAD THE DOCUMENTS, ID IS THE GOOD DOC_ID..
> //   I WOULD LIKE TO GET THIS IN FACT (ruby):
> //   doc_id = INDEX.search(''query'').hits.first
> //   INDEX[doc_id].load <== THAT''S WHAT I WOULD LIKE TO GET !
> //
>         }
>         rstr = rb_str_new2(str);
>         free(str);
>         return (argv[0]);
> }
>
> of course, i''ve bound the method to the good object etc...
>
> Hope somebody''ll help !
> Thank you by advance,
> Jeremie ''ahFeel'' BORDIER
>
The frt_td_to_s method is doing almost exactly this. Make sure you
have have version 0.10.9 or later. Each TopDocs object has a reference
to the searcher that created it. You can get a LazyDoc object for the
searcher with the following call:

    LazyDoc *lazy_doc = sea->get_lazy_doc(searcher, id);

By the way, id should be an int, not a long. To turn this into the
lazy loading Hash object that you see in Ruby you use this method:

    VALUE frt_get_lazy_doc(LazyDoc *lazy_doc)

The code for cLazyDoc is in r_index.c.

Can I ask why you need to do this in the C code? It would seem to me
to make a lot more sense to code this in Ruby but you probably have a
good reason.

Cheers,
Dave

David Balmain wrote:
> On 10/11/06, ahFeel <ahfeel at rift.fr> wrote:
>> Here is my problem :
>> {
>> //
>>
>> of course, i''ve bound the method to the good object etc...
>>
>> Hope somebody''ll help !
>> Thank you by advance,
>> Jeremie ''ahFeel'' BORDIER
>>
> 
> The frt_td_to_s method is doing almost exactly this. Make sure you
> have have version 0.10.9 or later. Each TopDocs object has a reference
> to the searcher that created it. You can get a LazyDoc object for the
> searcher with the following call:
> 
>     LazyDoc *lazy_doc = sea->get_lazy_doc(searcher, id);
> 
> By the way, id should be an int, not a long. To turn this into the
> lazy loading Hash object that you see in Ruby you use this method:
> 
>     VALUE frt_get_lazy_doc(LazyDoc *lazy_doc)
> 
> The code for cLazyDoc is in r_index.c.
> 
> Can I ask why you need to do this in the C code? It would seem to me
> to make a lot more sense to code this in Ruby but you probably have a
> good reason.
> 
> Cheers,
> Dave
Hi dave !

First, thank you a lot for this anwser !

I''ve been looking and trying a lot of things such as

sea = (Searcher *)DATA_PTR(self)

to use the sea->get_lazy_doc function, but it returns a null pointer... 
maybe i did something wrong, i''ll retry, and be sure to have the last 
ferret release :)

Concerning the reasons i''m doing this directly in C, it''s just
because
we need a really fast implementation.. I''m working with Florent Solt
who
you already knows, works on a big project using Ruby / Ferret. These 
datas are supposed to transit over a RPC protocol, and there will be a 
lot of queries, so we really need a fast to_json method :)

Thank you again,
Jeremie ''ahFeel'' BORDIER

-- 
Posted via http://www.ruby-forum.com/.

Ok, i''ve updated the gem, and the last to_s method is helping a LOT !

thank you again Dave,
ahf :)

-- 
Posted via http://www.ruby-forum.com/.