Yes, and also to avoid returning an incomplete row for the same primary key because of different timestamp for the column's cell.
On Sat, Jun 24, 2017 at 4:20 AM, Randy Hu <ruw...@gmail.com> wrote: > First HBase does not have a concept of "row timestamp". Timestamp is part > of each cell. The closest to row timestamp is probably the latest timestamp > from all cells with same row key. > > The reason that timestamp column need to be part of primary key is that > there could be multiple values with different timestamps but same row key, > family. and qualifier. The SQL result data model does not support such two > dimension structure well. The best Phoenix can do in the SQL model is to > return list of values in array, but then the cell timestamp need to be > returned in another array with same length under different column name. You > can imagine the mess and awkwardness it could be if going to that > direction. It's much straightforward to make time stamp part of primary > key, so the two dimension structure could be represented by multiple rows > naturally in SQL result. > > Randy > > > > > -- > View this message in context: http://apache-phoenix-user- > list.1124778.n5.nabble.com/phoenix-query-modtime-tp3702p3716.html > Sent from the Apache Phoenix User List mailing list archive at Nabble.com. >
