I didn’t realize that Spark was using that method. Thanks for the report 
Dongjoon.

We shouldn’t break compatibility, so I’ll add a new method to use a default of 
the previous value of false. My understanding is that Spark 3 actually uses the 
proleptic calendar and thus should use true.

.. Owen


> On Jan 26, 2020, at 1:10 AM, Dongjoon Hyun <[email protected]> wrote:
> 
> Hi, All.
> 
> ORC 1.5.9 seems to have some disruptive changes which breaks downstream.
> Unlike 1.5.8, the following Apache ORC changes cause a failure.
> 
> "ORC-578: Addendum fixing up the stripe statistics" seems to remove `public` 
> API.
> 
>    +++ b/java/core/src/java/org/apache/orc/impl/OrcTail.java
>    -  public List<StripeStatistics> getStripeStatistics() throws IOException {
>    +  public List<StripeStatistics> getStripeStatistics(
>    +      boolean writerUsedProlepticGregorian, boolean 
> convertToProlepticGregorian)
>    +      throws IOException {
> 
> In addition, ORC-27 changes `hive-storage-api` from 2.6.0 (Apr, 2018) to 
> 2.7.1 (Dec, 2019) in order to bring `DateColumnVector ` from the following 
> Hive patch.
> 
>    HIVE-22405 Add ColumnVector support for ProlepticCalendar
> 
> From user side, 2.7.1 seems to work with Hive 3.1.3+.
> Is there any expected incompatibility?
> I'm wondering if this works with Hive 1.2.1 and Hive 2.3.6?
> 
> Bests,
> Dongjoon.
> 
> On 2020/01/24 23:00:31, "Owen O'Malley" <[email protected]> wrote: 
>> All,
>> 
>> Should we release the following artifacts as ORC 1.5.9?
>> 
>> tar: http://home.apache.org/~omalley/orc-1.5.9/
>> tag: https://github.com/apache/orc/releases/tag/release-1.5.9rc0
>> jiras: https://issues.apache.org/jira/browse/ORC/fixforversion/12346546
>> 
>> Thanks!
>> 

Reply via email to