incubator-couchdb-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Elf <elf2...@gmail.com>
Subject Re: sort by key
Date Wed, 23 Sep 2009 10:04:18 GMT
2009/9/23 Jan Lehnardt <jan@apache.org>:
>
> On 22 Sep 2009, at 18:39, Elf wrote:
>
>> 2009/9/22 Ning Tan <ningtan@gmail.com>:
>>>
>>> On Tue, Sep 22, 2009 at 11:43 AM, Elf <elf2001@gmail.com> wrote:
>>>>
>>>> Do not use emit(something, doc), prefer to use include_docs=true (so,
>>>> emit(something, something_less_than_large_document) when requesting a
>>>> view.
>>>
>>> Can you elaborate on this? What if I have a mostly read database (i.e.
>>> I don't care too much about view update time and space limitation,
>>> etc) and want to get to the data very quickly? Having to use
>>> include_docs means internally I need to query a second b-tree to get
>>> the data, right?
>>
>> So, I will recommend you to store documents in a view :-)
>> But topicstarter is novice in couchdb, so I can't recommend that tactics
>> for it.
>> Also, I think, second b-tree lookup is not a bottleneck. Bottleneck
>> may be disk or network bandwidth, so... It's need to test your
>> environment, not to guessing.
>
> It'd be a secondary b-tree lookup per view result row. So this is a
> significant hit
> on view read time). Emitting the full doc is also a significant hit (on view
> index
> time). Emitting only the values you need from a view is best for both cases
> :)

I agree with you. I forgot that documents lookup must be done several
at one view request.


>
> Cheers
> Jan
> --
>
>



-- 
----------------
Best regards
Elf
mailto:elf2001@gmail.com

Mime
View raw message