A compiler could be a good option for doing this, but AFAICT many Pig examples would seem to operate as a sequence of view operations. So, if I had to implement this as a front-end/client DSL, it seems that I would lose the potential to interleave Pig operations to take advantage of pipelining optimizations. Oh well...
Perhaps a higher-order JavaScript library, to DRY up views, would be more pragmatic...? On Mon, Mar 2, 2009 at 12:58 PM, Chris Anderson <[email protected]> wrote: > On Mon, Mar 2, 2009 at 10:32 AM, Zachary Zolton > <[email protected]> wrote: >> Hey all, >> >> Has anyone looked into how a high-level data-processing language could >> integrate with CouchDB? >> >> http://wiki.apache.org/pig/PigOverview >> http://www.cs.cmu.edu/~olston/publications/sigmod08.pdf >> >> I've been reading a bit about (fellow Apache project) Pig, which >> implements features analogous to CouchDB views and combinative >> operations over views. I'm guessing it's not a perfect fit —otherwise >> I wager I'd have heard some mention of it on this mailing list. > > I think people are generally pretty happy writing views as JS > functions. Pig would probably be a decent fit, except that Hadoop M/R > does not really reduce, at least not in the sense that CouchDB does. > > That said, it may be possible to write a "compiler" for Pig (or a > dialect of Pig Latin) that generates JS map/reduce views for CouchDB. > > If such things float your boat, I say go for it. > > -- > Chris Anderson > http://jchris.mfdz.com >
