Interesting, I can't comment on the applicability of msgpack within Cassandra, but I posted something a while back about a serialization technique I was using for implementing inverted indexes at the application level on top of Cassandra (http://github.com/edanuff/CassandraCompositeType). I used a very primitive serialization scheme, I'll have to take a closer look at msgpack.
On Sun, Jul 11, 2010 at 9:12 AM, Thomas Heller <i...@zilence.net> wrote: > Hey, > > I was wondering what your plans are concerning secondary indexes and > supported operators? > > I looked over the code and so far there is only EQ which of course is > the easiest since it just compares byte values and that will work for > (almost) everything. Most index operations will however require > de-serializing the data. > > I chose msgpack binaries to store my data because it was the most > compact and it keeps basic type information intact (string, int, > floats, arrays, etc), but a msgpack int is not 4 bytes and wouldnt be > castable to a standard 4 byte int so for example a "simple less than" > wouldnt work. > > Are there any plans on this topic or is it still open to discussion? > > Cheers, > /thomas >
