gistfile1.txt

grncdr # can you make http requests from JS in a map reduce query?

seancribbs # grncdr: not sure - why do you need to?

grncdr # and (somewhat related) can you write back to riak in a map reduce query

seancribbs # grncdr: you _can_ write back to riak if you write the phase in Erlang

grncdr # hmm, I could probably learn enough erlang to do that, and it 
seems like I'm going to need erlang for the best performance anyways what 
I've done is stuck about 130million keys in a single bucket, then found out 
that was not a good idea ;)

seancribbs # ah, right

grncdr # so I'm wanting to run a mapreduce job that just moves keys into smaller buckets

seancribbs # well, you could also stream the keys back, fetch and update in a 
separate process it's still going to take a long time to move 130 million objects, though

grncdr # yep, this is not urgent, I'm just playing around really

seancribbs # if you have the data separately, why not just put it in the
new place and then delete it from the old on the client side?

grncdr # less fun/perverse?

seancribbs # or… more brutally, delete the data directories and then reload the data
heh

grncdr # I actually might need http requests for a legitimate use as well though

seancribbs # well you could definitely do them in Erlang. I'm not sure what's
built-in to Spidermonkey though and it might not be a good idea because of 
hard phase timeouts hmm, there's a CommonJS support project out there, 
but we don't build SM with it linked in

grncdr # the use case scenario could definitely be done client side, 
I'm just interested in what can be done in mapred

seancribbs # yeah - its purpose is really for queries only although you 
can twist its arm a little ;)

grncdr # well I'd rather it was optimized for the 99% of uses rather 
than build it double-jointed (to twist the metaphor a bit)

seancribbs # ha