Yes, it is possible to do that, and it is intended :) - your table still has col1 and col3 as part of its schema, and when people read that table, they will get nulls for col1 and col3.
-Sushanth On Thu, Nov 3, 2011 at 2:57 PM, Francis Christopher Liu <[email protected]> wrote: > Hi, > I'd like to clarify wether this selective column updates are supported as > part of HCatOutputFormat.setSchema(). > If I have an unpartitioned table: > foo(int col1, int col2, int col3) > Would it be possible for me to just pass column "col2" when calling > setSchema? For columnar stores (ie Hbase) it would be more efficient just to > write out the values that we want updated instead of the entire row. > -Francis
