map/reduce fails with documents in jiak

Bryan Fink bryan at basho.com
Mon Feb 8 09:42:38 EST 2010


On Sun, Feb 7, 2010 at 4:10 PM, francisco treacy
<francisco.treacy at gmail.com> wrote:
> I noticed that the example at
> http://bitbucket.org/basho/riak/src/tip/doc/js-mapreduce.org stores
> data in /raw.  So I did the same with my (json) data (instead of
> storing at /jiak) and now map/reduce works.
>
> Is this behaviour intended?  If so, I cannot find it documented anywhere.


Hi, Francisco.  This behavior is not entirely unintended.  We have
been circulating the idea of deprecating the Jiak interface among Riak
users.  So far, we have yet to find much opposition to the idea.
Given that the raw interface can handle more than just JSON, more
efficiently than the Jiak interface, we plan to go ahead with this
deprecation.


> (And btw, is there really a difference between /jiak and /raw, if all
> I'm doing is just storing data with a Content-Type?)


The biggest difference is that /raw won't be enforcing any schema
restrictions, like Jiak did.  There will be client libraries in major
languages for /raw , just like there are for /jiak now.  The "raw"
interface will handle JSON just as well.  Differences at the HTTP
layer involve where links and vclocks are stored (in headers instead
of in JSON), but this will largely be abstracted by the new client
libraries.

Which is all a long way of saying that, yes, /raw is considered the
preferred interface for data that will be used in Javascript
map/reduce jobs.

-Bryan



More information about the riak-users mailing list