The current timestamp type is based on milliseconds. This patch changes timestamp type to be based on unixtime.
committed it to master.
FAILURE: Integrated in Tajo-trunk-postcommit #638 (See https://builds.apache.org/job/Tajo-trunk-postcommit/638/)
TAJO-448: Timestamp should be based on unixtime. (hyunsik) (hyunsik: https://git-wip-us.apache.org/repos/asf?p=incubator-tajo.git&a=commit&h=fce88829d58885197f67c477699b57c632e7246d)
Thank you for the kind reply.
Now, I understand and totally agree with you.
Actually, it works well for CSV file. But, we need more time to think of the fixed binary data format for storing timestamp into some binary format. After the binary format is fixed once, it is hard to change the format.
I understand your point.
Is there any reason that we do not store a timestamp type data?
Even though It is based on unix time, it can optionally has milliseconds ranging from 0 to 999999999. 32 bits and 64bits all are not enough to represent Timestamp's full representation. 64bit log may be a temporal solution. It may be OK because currently we can only use timestamp values during query processing, and we cannot store a data including timestamp type to any file.
It looks good to me, too.
But, are 32 bits enough to represent unix time?
32 bits unix time only supports up to 2038-01-19 and PostgreSQL looks to support unix time bigger than 32 bits.
It looks good.
'mvn clean install' finished successfully.