question on custom indexes

[Tommaso Teofili <te...@adobe.com>]

Hi all,

maybe that's a silly question but I wonder what is currently considered the best practice for where to create index nodes.
Do they all have to (or _should_) be under a same (e.g. query:index) node or rather (as I remember it was discussed) on specific paths? Or it may depend on the specific implementation?

Thanks in advance for your help,
Tommaso

Re: question on custom indexes

[Tommaso Teofili <te...@adobe.com>]

On 24/gen/2013, at 10:59, Thomas Mueller wrote:

> Hi,
> 
> The current code assumes the index configuration nodes are under
> /oak:index.

Ok thanks for the clarification Thomas, that sounds good for now.

> 
> We discussed that index nodes could be local, where the content is, if the
> index only applies to a subtree (is not repository-wide). But this would
> make things more complicated in the index code: when updating the content,
> all relevant indexes need to be updated, so the indexes must be known at
> that time or discovered during the update. The same when running a query.
> It's possible to do, but so far we didn't start implementing this feature.

maybe keeping a prefix tree of all the indexes could be a fast way to retrieve all the indexes which lay under a certain path (avoiding node-is-an-index checks), where the prefix tree gets updated by QueryIndexProviders.

My 2 cents,
Tommaso

> 
> Regards,
> Thomas
> 
> 
> 
> On 1/24/13 10:46 AM, "Tommaso Teofili" <te...@adobe.com> wrote:
> 
>> Hi all,
>> 
>> maybe that's a silly question but I wonder what is currently considered
>> the best practice for where to create index nodes.
>> Do they all have to (or _should_) be under a same (e.g. query:index) node
>> or rather (as I remember it was discussed) on specific paths? Or it may
>> depend on the specific implementation?
>> 
>> Thanks in advance for your help,
>> Tommaso
> 

Re: question on custom indexes

[Thomas Mueller <mu...@adobe.com>]

Hi,

The current code assumes the index configuration nodes are under
/oak:index.

We discussed that index nodes could be local, where the content is, if the
index only applies to a subtree (is not repository-wide). But this would
make things more complicated in the index code: when updating the content,
all relevant indexes need to be updated, so the indexes must be known at
that time or discovered during the update. The same when running a query.
It's possible to do, but so far we didn't start implementing this feature.

Regards,
Thomas



On 1/24/13 10:46 AM, "Tommaso Teofili" <te...@adobe.com> wrote:

>Hi all,
>
>maybe that's a silly question but I wonder what is currently considered
>the best practice for where to create index nodes.
>Do they all have to (or _should_) be under a same (e.g. query:index) node
>or rather (as I remember it was discussed) on specific paths? Or it may
>depend on the specific implementation?
>
>Thanks in advance for your help,
>Tommaso