I have small doubt regarding Teradata Columnar feature consider a column partitioned table with three column partitions C1,C2 and C3. The table has say a billion rows. is a datablock allowed to contain only one kind of Container C1,C2 or C3 contiguously. or is it mandatory that a datablock should have combination of all Container C1,C2 and C3 for the given sets of row.
Basically is a logical row allowed to exist across multiple datablocks?
if so then the Conventional Master Index descriptors and Cylinder Index Descriptors to identify the location of the row(s) wouldnt work right?