On Tue, Oct 6, 2009 at 2:20 PM, Edmond Lau <[email protected]> wrote: > The get_slice() API takes a ColumnParent and a predicate, which means > that I can either 1) pick one super column and slice its constituent > subcolumns or 2) slice the super columns and pick up all constituent > subcolumns. What I'd like to do instead is to take a slice of > supercolumns and then a slice of columns within those supercolumns. > > With the current API design, this would involve issuing multiple > get_slice calls, one for each super column that I want. Supporting > this with one call doesn't seem like a limitation of the data model, > but rather of the thrift api. It seems like get_slice could take in a > pair of optional predicates instead, one for the supercolumn and one > for the column. Granted, the semantics of a parameter like "count" > would be harder to understand with two slice predicates. > > Is this functionality not supported because no one else had a use for it > (yet)?
Since supercolumn contents are always fully deserialized into memory, it wouldn't be too hard to have one predicate to pull out supercolumns, and use the other on its contents. But no, clearly nobody with a use enough needed it badly enough to submit a patch so far. :) -Jonathan
