Github user omalley commented on the issue:
https://github.com/apache/orc/pull/233
So the semantics of the current timestamp type are that it does not change
with the reader's timezone. It should always come back the same. If you put
'2018-03-23 12:34:56.789' in, it should always come back exactly the same.
The easiest way to get that effect is to use UTC in the ColumnVectorBatch,
so that is what we do. This is different from the Java side because there we
needed to use Java's Timestamp class, which always uses local.
We are going to add a new type that is "timestamp with local timezone" in
ORC-189. That is for the instance in time where timezone is taken in to account.
The encoding of timestamps in ORC 1.x is overly complicated because I
messed it up and we've tried to maintain backwards and forwards compatibility.
In ORC 2 we'll just use UTC for encoding both types.
---
