gistfile1.txt

06:41 <jjthrash> Not an urgent thing, but I'd like to understand better the 
cost of anon js functions in map reduce queries. 

06:42 <jjthrash> If I'm map-reducing over 2 or 3 nodes with thousands of 
records, how much impact is an anonymous JS function going to have? 

06:42 <jjthrash> vs. stored in bucket, vs. loaded in a JS library? 

06:42 <jjthrash> thousands, not tens of thousands of records. 

06:43 <seancribbs> stored in a JS lib is definitely the fastest, because 
to doesn't ever have to reevaluate it 

06:43 <seancribbs> bucket requires an extra fetch, but then should be cached 
on the node for a time 

06:43 <seancribbs> but I don't have hard numbers, Kevin would be the better 
person to ask 

06:44 <jjthrash> OK. 

06:45 <jjthrash> Thanks. For now, everything is anonymous JS, and I'm 
wondering where to spend my time. We're doing some report queries that are 
hammering both of our nodes for a couple minutes. 

06:47 <seancribbs> Riak tip will improve that 

06:47 <seancribbs> both in terms of JS memory usage, marshalling, key listing 

06:47 <seancribbs> s/both// 

06:47 <jjthrash> Is it "stable"? :) 

06:48 <seancribbs> might be good to try in dev 

06:48 <seancribbs> a new release is coming soon, so i hear 

06:49 <jjthrash> You're thinking it's a bigger bang/buck to upgrade than to 
tinker with where the JS functions live? 

06:49 <seancribbs> no, certainly you'll want them in a library in the long run 

06:50 <seancribbs> but I'm not saying it'll dramatically improve performance 
of your job 

06:50 <jjthrash> OK. 

06:52 <jjthrash> Thanks for the help.