Details
-
Improvement
-
Status: Open
-
Major
-
Resolution: Unresolved
-
None
-
None
-
None
Description
Related to the thread on the arrow dev mailing list: https://lists.apache.org/thread.html/rf1a377c66990ae5ac0693119d416c93a7e19228d3eaaea8bd90acb17%40%3Cdev.arrow.apache.org%3E
Currently, when writing parquet files with Arrow (parquet-cpp), we default to parquet format "1.0". In practice, this means that we don't use certain LogicalTypes (eg we don't write integers other than int32/int64, and we don't write the nanosecond timestamps).
I think it would be nice to enable nanosecond timestamps by default, but I also have no idea how widely this is already supported by other readers.
To be clear, this is not about enabling data page version 2 by default, in Arrow that is governed by a separate option.
While checking this, I made an overview of which types were introduced in
which parquet format version, in case someone wants to see the details ->
https://nbviewer.jupyter.org/gist/jorisvandenbossche/3cc9942eaffb53564df65395e5656702