Mailing-List: contact user-help@couchdb.apache.org; run by ezmlm
Precedence: bulk
Reply-To: user@couchdb.apache.org
Received-SPF: neutral (athena.apache.org: local policy)
MIME-Version: 1.0
In-Reply-To: <53410712-5299-43D7-9AAD-578779BCC881@me.com>
References: <1300861645.89657.ezmlm@couchdb.apache.org>
	<53410712-5299-43D7-9AAD-578779BCC881@me.com>
Date: Thu, 24 Mar 2011 10:44:31 +1300
Message-ID: <AANLkTi=8aVcT=8LsT=efHjT6Hz703efQxzfrbDcicO4A@mail.gmail.com>
Subject: Re: Re: Cross/Joined Database Access?
From: Dave Cottlehuber <dave@muse.net.nz>
To: user@couchdb.apache.org
Content-Type: text/plain; charset=ISO-8859-1

On 23 March 2011 20:47, Marc Schwering <marcschwering@me.com> wrote:
> So there is no way to span queries along multiple connections right? I thought there might be a solution using the replication / sharding features of couch?
>
> Am 22.03.2011 um 12:26 schrieb dominic.tarr@gmail.com:
>>
>> you would have to do multiple http requests, i think.
>>
>> would it be possible to combine all of the dbs into one database and just
>> have a field = user1,user2 etc?

This is very common - see
http://guide.couchdb.org/draft/views.html#comments and the nice
picture below it for an example.

>> somethings like this seem weird, coming from relational databases, but thats
>> couchdb.
>>
>> On Tue, Mar 22, 2011 at 9:02 PM, Marc Schwering <mschwering@weluse.de>wrote:
>>
>>> Hi there.
>>>
>>> Is it possible // How do i Search on multiple Databases?
>>>
>>> Given one Layout/Schema of Entities:
>>> Postings => id; title; content; timestamp
>>>
>>> This Schema exists due multiple instances of one Application (for different
>>> Users) on n Databases. Like: db1,db2,db3
>>>
>>> I want to do two fetches:
>>>
>>> a) all Posts since time X until Y sorted by timestamp. from db1 & db3.
>>> b) all Posts since time X until Y sorted by timestamp. from db2 & db3.
>>>
>>> Is that possible? If so, how?
>>>
>>> Thanks in advance
>>>
>>> Marc
>>>
>>> --
>>> M.Schwering

Long story short if this is a critical requirement for your app, it
may make more sense to keep all data in a single DB, and use
validation functions, replication filters, and either duck typing or
tagging to ensure you know who "owns" or can access each doc.

If you *need* to keep data private, another workaround is to use
multiple couches:

one db per user, private - they would replicate / talk to only this personal DB.
have one-way replication from each per-user db to a larger DB which
consolidates all user data, excluding any private stuff where
necessary.
run your spanning queries from this DB.
optionally use client-side code to pull in any private data from the
person's private DB.

for time-bound queries, try something like emitting an array of JSON {
"year": "2001", "month" and ensure you use
http://guide.couchdb.org/draft/recipes.html#fast
more info on your app might help give a better answer.

A+
Dave