You are viewing a plain text version of this content. The canonical link for it is here.
Posted to solr-user@lucene.apache.org by Juan Pablo Mora <ju...@informa.es> on 2011/12/03 12:43:41 UTC
Grouping or Facet ?
I need to do some counts on a StrField field to suggest options from two different categories, and I don´t know what option is the best:
My schema looks:
- id
- name
- category: XX or YY
with Grouping I do:
http://localhost:8983/?q=name:prefix*&group=true&group.field=category
But I can change my schema to to:
- id
- nameXX
- nameYY
- category: XX or YY (only 1 value in nameXX or nameYY)
With facet:
http://localhost:8983/?q=*:*&facet=true&facet.field=nameXX&facet.field=nameYY&facet.prefix=prefix
What option have the best performance ?
Best,
Juampa.
Re: Grouping or Facet ?
Posted by Juan Pablo Mora <ju...@informa.es>.
Sorry if I don´t explain my problem clearly...
I need to do a suggester of names based on a prefix. My data are from two categories of people, admins and developers for example. So when the client write "SAN" my results should be:
Prefix: San
Developers: Sanchez Garcia, Juan (5)
Sanchez Roman, Ivan (2)
San...
Admins: Sanchez, Pedro (7)
Sanchez Garcia, Javier (2)
And the most common a name is, the upper position will have. So I think is not posible to do that with grouping. So finally my schema will be:
id
nameDeveloper or nameAdmin : both String fields, but only one will have values in a doc.
And my query with facet will be:
/q=*:*&facet=true&facet.field=nameDeveloper&facet.field=nameAdmin&facet.prefix=SAN&facet.minCounts=1
If I try to do that with grouping I need something like group.pivot=category,name , and is not posible in Solr yet.
Best,
Juampa.
El 08/12/2011, a las 02:23, Darren Govoni escribió:
> Yes. That's what I would expect. I guess I didn't understand when you said
>
> "The facet counts are the counts of the *values* in that field"
>
> Because it seems its the count of the number of matching documents irrespective
> if one document has 20 values for that field and another 10, the facet count will be 2,
> one for each document in the results.
>
> On 12/07/2011 09:04 AM, Erick Erickson wrote:
>> In your example you'll have 10 facets returned each with a value of 1.
>>
>> Best
>> Erick
>>
>> On Tue, Dec 6, 2011 at 9:54 AM,<da...@ontrenet.com> wrote:
>>> Sorry to jump into this thread, but are you saying that the facet count is
>>> not # of result hits?
>>>
>>> So if I have 1 document with field CAT that has 10 values and I do a query
>>> that returns this 1 document with faceting, that the CAT facet count will
>>> be 10 not 1? I don't seem to be seeing that behavior in my app (Solr 3.5).
>>>
>>> Thanks.
>>>
>>>> OK, I'm not understanding here. You get the counts and the results if you
>>>> facet
>>>> on a single category field. The facet counts are the counts of the
>>>> *values* in that
>>>> field. So it would help me if you showed the output of faceting on a
>>>> single
>>>> category field and why that didn't work for you....
>>>>
>>>> But either way, faceting will probably outperform grouping.
>>>>
>>>> Best
>>>> Erick
>>>>
>>>> On Mon, Dec 5, 2011 at 9:05 AM, Juan Pablo Mora<ju...@informa.es> wrote:
>>>>> Because I need the count and the result to return back to the client
>>>>> side. Both the grouping and the facet offers me a solution to do that,
>>>>> but my doubt is about performance ...
>>>>>
>>>>> With Grouping my results are:
>>>>>
>>>>> "grouped":{
>>>>> "category":{
>>>>> "matches": ...,
>>>>> "groups":[{
>>>>> "groupValue":"categoryXX",
>>>>> "doclist":{"numFound":Important_number,"start":0,"docs":[
>>>>> {
>>>>> doc:id
>>>>> category:XX
>>>>> }
>>>>> "groupValue":"categoryYY",
>>>>> "doclist":{"numFound":Important_number,"start":0,"docs":[
>>>>> {
>>>>> doc: id
>>>>> category:YY
>>>>> }
>>>>>
>>>>> And with faceting my results are :
>>>>> "facet.prefix=whatever"
>>>>> "facet_counts":{
>>>>> "facet_queries":{},
>>>>> "facet_fields":{
>>>>> "namesXX":[
>>>>> "whatever_name_in_category",76,
>>>>> ...
>>>>> "namesYY":[
>>>>> "whatever_name_in_category",76,
>>>>> ...
>>>>>
>>>>> Both results are OK to me.
>>>>>
>>>>>
>>>>> ________________________________________
>>>>> De: Erick Erickson [erickerickson@gmail.com]
>>>>> Enviado el: lunes, 05 de diciembre de 2011 14:48
>>>>> Para: solr-user@lucene.apache.org
>>>>> Asunto: Re: Grouping or Facet ?
>>>>>
>>>>> Why not just use the first form of the document
>>>>> and just facet.field=category? You'll get
>>>>> two different facet counts for XX and YY
>>>>> that way.
>>>>>
>>>>> I don't think grouping is the way to go here.
>>>>>
>>>>> Best
>>>>> Erick
>>>>>
>>>>> On Sat, Dec 3, 2011 at 6:43 AM, Juan Pablo Mora<ju...@informa.es>
>>>>> wrote:
>>>>>> I need to do some counts on a StrField field to suggest options from
>>>>>> two different categories, and I don´t know what option is the best:
>>>>>>
>>>>>> My schema looks:
>>>>>>
>>>>>> - id
>>>>>> - name
>>>>>> - category: XX or YY
>>>>>>
>>>>>> with Grouping I do:
>>>>>>
>>>>>> http://localhost:8983/?q=name:prefix*&group=true&group.field=category
>>>>>>
>>>>>> But I can change my schema to to:
>>>>>>
>>>>>> - id
>>>>>> - nameXX
>>>>>> - nameYY
>>>>>> - category: XX or YY (only 1 value in nameXX or nameYY)
>>>>>>
>>>>>> With facet:
>>>>>> http://localhost:8983/?q=*:*&facet=true&facet.field=nameXX&facet.field=nameYY&facet.prefix=prefix
>>>>>>
>>>>>>
>>>>>> What option have the best performance ?
>>>>>>
>>>>>> Best,
>>>>>> Juampa.
>
Re: Grouping or Facet ?
Posted by Darren Govoni <da...@ontrenet.com>.
Yes. That's what I would expect. I guess I didn't understand when you said
"The facet counts are the counts of the *values* in that field"
Because it seems its the count of the number of matching documents
irrespective
if one document has 20 values for that field and another 10, the facet
count will be 2,
one for each document in the results.
On 12/07/2011 09:04 AM, Erick Erickson wrote:
> In your example you'll have 10 facets returned each with a value of 1.
>
> Best
> Erick
>
> On Tue, Dec 6, 2011 at 9:54 AM,<da...@ontrenet.com> wrote:
>> Sorry to jump into this thread, but are you saying that the facet count is
>> not # of result hits?
>>
>> So if I have 1 document with field CAT that has 10 values and I do a query
>> that returns this 1 document with faceting, that the CAT facet count will
>> be 10 not 1? I don't seem to be seeing that behavior in my app (Solr 3.5).
>>
>> Thanks.
>>
>>> OK, I'm not understanding here. You get the counts and the results if you
>>> facet
>>> on a single category field. The facet counts are the counts of the
>>> *values* in that
>>> field. So it would help me if you showed the output of faceting on a
>>> single
>>> category field and why that didn't work for you....
>>>
>>> But either way, faceting will probably outperform grouping.
>>>
>>> Best
>>> Erick
>>>
>>> On Mon, Dec 5, 2011 at 9:05 AM, Juan Pablo Mora<ju...@informa.es> wrote:
>>>> Because I need the count and the result to return back to the client
>>>> side. Both the grouping and the facet offers me a solution to do that,
>>>> but my doubt is about performance ...
>>>>
>>>> With Grouping my results are:
>>>>
>>>> "grouped":{
>>>> "category":{
>>>> "matches": ...,
>>>> "groups":[{
>>>> "groupValue":"categoryXX",
>>>> "doclist":{"numFound":Important_number,"start":0,"docs":[
>>>> {
>>>> doc:id
>>>> category:XX
>>>> }
>>>> "groupValue":"categoryYY",
>>>> "doclist":{"numFound":Important_number,"start":0,"docs":[
>>>> {
>>>> doc: id
>>>> category:YY
>>>> }
>>>>
>>>> And with faceting my results are :
>>>> "facet.prefix=whatever"
>>>> "facet_counts":{
>>>> "facet_queries":{},
>>>> "facet_fields":{
>>>> "namesXX":[
>>>> "whatever_name_in_category",76,
>>>> ...
>>>> "namesYY":[
>>>> "whatever_name_in_category",76,
>>>> ...
>>>>
>>>> Both results are OK to me.
>>>>
>>>>
>>>> ________________________________________
>>>> De: Erick Erickson [erickerickson@gmail.com]
>>>> Enviado el: lunes, 05 de diciembre de 2011 14:48
>>>> Para: solr-user@lucene.apache.org
>>>> Asunto: Re: Grouping or Facet ?
>>>>
>>>> Why not just use the first form of the document
>>>> and just facet.field=category? You'll get
>>>> two different facet counts for XX and YY
>>>> that way.
>>>>
>>>> I don't think grouping is the way to go here.
>>>>
>>>> Best
>>>> Erick
>>>>
>>>> On Sat, Dec 3, 2011 at 6:43 AM, Juan Pablo Mora<ju...@informa.es>
>>>> wrote:
>>>>> I need to do some counts on a StrField field to suggest options from
>>>>> two different categories, and I don´t know what option is the best:
>>>>>
>>>>> My schema looks:
>>>>>
>>>>> - id
>>>>> - name
>>>>> - category: XX or YY
>>>>>
>>>>> with Grouping I do:
>>>>>
>>>>> http://localhost:8983/?q=name:prefix*&group=true&group.field=category
>>>>>
>>>>> But I can change my schema to to:
>>>>>
>>>>> - id
>>>>> - nameXX
>>>>> - nameYY
>>>>> - category: XX or YY (only 1 value in nameXX or nameYY)
>>>>>
>>>>> With facet:
>>>>> http://localhost:8983/?q=*:*&facet=true&facet.field=nameXX&facet.field=nameYY&facet.prefix=prefix
>>>>>
>>>>>
>>>>> What option have the best performance ?
>>>>>
>>>>> Best,
>>>>> Juampa.
Re: Grouping or Facet ?
Posted by Erick Erickson <er...@gmail.com>.
In your example you'll have 10 facets returned each with a value of 1.
Best
Erick
On Tue, Dec 6, 2011 at 9:54 AM, <da...@ontrenet.com> wrote:
> Sorry to jump into this thread, but are you saying that the facet count is
> not # of result hits?
>
> So if I have 1 document with field CAT that has 10 values and I do a query
> that returns this 1 document with faceting, that the CAT facet count will
> be 10 not 1? I don't seem to be seeing that behavior in my app (Solr 3.5).
>
> Thanks.
>
>> OK, I'm not understanding here. You get the counts and the results if you
>> facet
>> on a single category field. The facet counts are the counts of the
>> *values* in that
>> field. So it would help me if you showed the output of faceting on a
>> single
>> category field and why that didn't work for you....
>>
>> But either way, faceting will probably outperform grouping.
>>
>> Best
>> Erick
>>
>> On Mon, Dec 5, 2011 at 9:05 AM, Juan Pablo Mora <ju...@informa.es> wrote:
>>> Because I need the count and the result to return back to the client
>>> side. Both the grouping and the facet offers me a solution to do that,
>>> but my doubt is about performance ...
>>>
>>> With Grouping my results are:
>>>
>>> "grouped":{
>>> "category":{
>>> "matches": ...,
>>> "groups":[{
>>> "groupValue":"categoryXX",
>>> "doclist":{"numFound":Important_number,"start":0,"docs":[
>>> {
>>> doc:id
>>> category:XX
>>> }
>>> "groupValue":"categoryYY",
>>> "doclist":{"numFound":Important_number,"start":0,"docs":[
>>> {
>>> doc: id
>>> category:YY
>>> }
>>>
>>> And with faceting my results are :
>>> "facet.prefix=whatever"
>>> "facet_counts":{
>>> "facet_queries":{},
>>> "facet_fields":{
>>> "namesXX":[
>>> "whatever_name_in_category",76,
>>> ...
>>> "namesYY":[
>>> "whatever_name_in_category",76,
>>> ...
>>>
>>> Both results are OK to me.
>>>
>>>
>>> ________________________________________
>>> De: Erick Erickson [erickerickson@gmail.com]
>>> Enviado el: lunes, 05 de diciembre de 2011 14:48
>>> Para: solr-user@lucene.apache.org
>>> Asunto: Re: Grouping or Facet ?
>>>
>>> Why not just use the first form of the document
>>> and just facet.field=category? You'll get
>>> two different facet counts for XX and YY
>>> that way.
>>>
>>> I don't think grouping is the way to go here.
>>>
>>> Best
>>> Erick
>>>
>>> On Sat, Dec 3, 2011 at 6:43 AM, Juan Pablo Mora <ju...@informa.es>
>>> wrote:
>>>> I need to do some counts on a StrField field to suggest options from
>>>> two different categories, and I don´t know what option is the best:
>>>>
>>>> My schema looks:
>>>>
>>>> - id
>>>> - name
>>>> - category: XX or YY
>>>>
>>>> with Grouping I do:
>>>>
>>>> http://localhost:8983/?q=name:prefix*&group=true&group.field=category
>>>>
>>>> But I can change my schema to to:
>>>>
>>>> - id
>>>> - nameXX
>>>> - nameYY
>>>> - category: XX or YY (only 1 value in nameXX or nameYY)
>>>>
>>>> With facet:
>>>> http://localhost:8983/?q=*:*&facet=true&facet.field=nameXX&facet.field=nameYY&facet.prefix=prefix
>>>>
>>>>
>>>> What option have the best performance ?
>>>>
>>>> Best,
>>>> Juampa.
>>
>
Re: Grouping or Facet ?
Posted by da...@ontrenet.com.
Sorry to jump into this thread, but are you saying that the facet count is
not # of result hits?
So if I have 1 document with field CAT that has 10 values and I do a query
that returns this 1 document with faceting, that the CAT facet count will
be 10 not 1? I don't seem to be seeing that behavior in my app (Solr 3.5).
Thanks.
> OK, I'm not understanding here. You get the counts and the results if you
> facet
> on a single category field. The facet counts are the counts of the
> *values* in that
> field. So it would help me if you showed the output of faceting on a
> single
> category field and why that didn't work for you....
>
> But either way, faceting will probably outperform grouping.
>
> Best
> Erick
>
> On Mon, Dec 5, 2011 at 9:05 AM, Juan Pablo Mora <ju...@informa.es> wrote:
>> Because I need the count and the result to return back to the client
>> side. Both the grouping and the facet offers me a solution to do that,
>> but my doubt is about performance ...
>>
>> With Grouping my results are:
>>
>> "grouped":{
>> "category":{
>> "matches": ...,
>> "groups":[{
>> "groupValue":"categoryXX",
>> "doclist":{"numFound":Important_number,"start":0,"docs":[
>> {
>> doc:id
>> category:XX
>> }
>> "groupValue":"categoryYY",
>> "doclist":{"numFound":Important_number,"start":0,"docs":[
>> {
>> doc: id
>> category:YY
>> }
>>
>> And with faceting my results are :
>> "facet.prefix=whatever"
>> "facet_counts":{
>> "facet_queries":{},
>> "facet_fields":{
>> "namesXX":[
>> "whatever_name_in_category",76,
>> ...
>> "namesYY":[
>> "whatever_name_in_category",76,
>> ...
>>
>> Both results are OK to me.
>>
>>
>> ________________________________________
>> De: Erick Erickson [erickerickson@gmail.com]
>> Enviado el: lunes, 05 de diciembre de 2011 14:48
>> Para: solr-user@lucene.apache.org
>> Asunto: Re: Grouping or Facet ?
>>
>> Why not just use the first form of the document
>> and just facet.field=category? You'll get
>> two different facet counts for XX and YY
>> that way.
>>
>> I don't think grouping is the way to go here.
>>
>> Best
>> Erick
>>
>> On Sat, Dec 3, 2011 at 6:43 AM, Juan Pablo Mora <ju...@informa.es>
>> wrote:
>>> I need to do some counts on a StrField field to suggest options from
>>> two different categories, and I don´t know what option is the best:
>>>
>>> My schema looks:
>>>
>>> - id
>>> - name
>>> - category: XX or YY
>>>
>>> with Grouping I do:
>>>
>>> http://localhost:8983/?q=name:prefix*&group=true&group.field=category
>>>
>>> But I can change my schema to to:
>>>
>>> - id
>>> - nameXX
>>> - nameYY
>>> - category: XX or YY (only 1 value in nameXX or nameYY)
>>>
>>> With facet:
>>> http://localhost:8983/?q=*:*&facet=true&facet.field=nameXX&facet.field=nameYY&facet.prefix=prefix
>>>
>>>
>>> What option have the best performance ?
>>>
>>> Best,
>>> Juampa.
>
Re: Grouping or Facet ?
Posted by Erick Erickson <er...@gmail.com>.
OK, I'm not understanding here. You get the counts and the results if you facet
on a single category field. The facet counts are the counts of the
*values* in that
field. So it would help me if you showed the output of faceting on a single
category field and why that didn't work for you....
But either way, faceting will probably outperform grouping.
Best
Erick
On Mon, Dec 5, 2011 at 9:05 AM, Juan Pablo Mora <ju...@informa.es> wrote:
> Because I need the count and the result to return back to the client side. Both the grouping and the facet offers me a solution to do that, but my doubt is about performance ...
>
> With Grouping my results are:
>
> "grouped":{
> "category":{
> "matches": ...,
> "groups":[{
> "groupValue":"categoryXX",
> "doclist":{"numFound":Important_number,"start":0,"docs":[
> {
> doc:id
> category:XX
> }
> "groupValue":"categoryYY",
> "doclist":{"numFound":Important_number,"start":0,"docs":[
> {
> doc: id
> category:YY
> }
>
> And with faceting my results are :
> "facet.prefix=whatever"
> "facet_counts":{
> "facet_queries":{},
> "facet_fields":{
> "namesXX":[
> "whatever_name_in_category",76,
> ...
> "namesYY":[
> "whatever_name_in_category",76,
> ...
>
> Both results are OK to me.
>
>
> ________________________________________
> De: Erick Erickson [erickerickson@gmail.com]
> Enviado el: lunes, 05 de diciembre de 2011 14:48
> Para: solr-user@lucene.apache.org
> Asunto: Re: Grouping or Facet ?
>
> Why not just use the first form of the document
> and just facet.field=category? You'll get
> two different facet counts for XX and YY
> that way.
>
> I don't think grouping is the way to go here.
>
> Best
> Erick
>
> On Sat, Dec 3, 2011 at 6:43 AM, Juan Pablo Mora <ju...@informa.es> wrote:
>> I need to do some counts on a StrField field to suggest options from two different categories, and I don´t know what option is the best:
>>
>> My schema looks:
>>
>> - id
>> - name
>> - category: XX or YY
>>
>> with Grouping I do:
>>
>> http://localhost:8983/?q=name:prefix*&group=true&group.field=category
>>
>> But I can change my schema to to:
>>
>> - id
>> - nameXX
>> - nameYY
>> - category: XX or YY (only 1 value in nameXX or nameYY)
>>
>> With facet:
>> http://localhost:8983/?q=*:*&facet=true&facet.field=nameXX&facet.field=nameYY&facet.prefix=prefix
>>
>>
>> What option have the best performance ?
>>
>> Best,
>> Juampa.
RE: Grouping or Facet ?
Posted by Juan Pablo Mora <ju...@informa.es>.
Because I need the count and the result to return back to the client side. Both the grouping and the facet offers me a solution to do that, but my doubt is about performance ...
With Grouping my results are:
"grouped":{
"category":{
"matches": ...,
"groups":[{
"groupValue":"categoryXX",
"doclist":{"numFound":Important_number,"start":0,"docs":[
{
doc:id
category:XX
}
"groupValue":"categoryYY",
"doclist":{"numFound":Important_number,"start":0,"docs":[
{
doc: id
category:YY
}
And with faceting my results are :
"facet.prefix=whatever"
"facet_counts":{
"facet_queries":{},
"facet_fields":{
"namesXX":[
"whatever_name_in_category",76,
...
"namesYY":[
"whatever_name_in_category",76,
...
Both results are OK to me.
________________________________________
De: Erick Erickson [erickerickson@gmail.com]
Enviado el: lunes, 05 de diciembre de 2011 14:48
Para: solr-user@lucene.apache.org
Asunto: Re: Grouping or Facet ?
Why not just use the first form of the document
and just facet.field=category? You'll get
two different facet counts for XX and YY
that way.
I don't think grouping is the way to go here.
Best
Erick
On Sat, Dec 3, 2011 at 6:43 AM, Juan Pablo Mora <ju...@informa.es> wrote:
> I need to do some counts on a StrField field to suggest options from two different categories, and I don´t know what option is the best:
>
> My schema looks:
>
> - id
> - name
> - category: XX or YY
>
> with Grouping I do:
>
> http://localhost:8983/?q=name:prefix*&group=true&group.field=category
>
> But I can change my schema to to:
>
> - id
> - nameXX
> - nameYY
> - category: XX or YY (only 1 value in nameXX or nameYY)
>
> With facet:
> http://localhost:8983/?q=*:*&facet=true&facet.field=nameXX&facet.field=nameYY&facet.prefix=prefix
>
>
> What option have the best performance ?
>
> Best,
> Juampa.
Re: Grouping or Facet ?
Posted by Erick Erickson <er...@gmail.com>.
Why not just use the first form of the document
and just facet.field=category? You'll get
two different facet counts for XX and YY
that way.
I don't think grouping is the way to go here.
Best
Erick
On Sat, Dec 3, 2011 at 6:43 AM, Juan Pablo Mora <ju...@informa.es> wrote:
> I need to do some counts on a StrField field to suggest options from two different categories, and I don´t know what option is the best:
>
> My schema looks:
>
> - id
> - name
> - category: XX or YY
>
> with Grouping I do:
>
> http://localhost:8983/?q=name:prefix*&group=true&group.field=category
>
> But I can change my schema to to:
>
> - id
> - nameXX
> - nameYY
> - category: XX or YY (only 1 value in nameXX or nameYY)
>
> With facet:
> http://localhost:8983/?q=*:*&facet=true&facet.field=nameXX&facet.field=nameYY&facet.prefix=prefix
>
>
> What option have the best performance ?
>
> Best,
> Juampa.