Hyukjin Kwon created SPARK-14787:
------------------------------------
Summary: Upgrade Joda-Time library from 2.9 to 2.9.3
Key: SPARK-14787
URL: https://issues.apache.org/jira/browse/SPARK-14787
Project: Spark
Issue Type: Improvement
Components: SQL
Affects Versions: 2.0.0
Reporter: Hyukjin Kwon
Priority: Trivial
Currently, Spark uses {{joda-time}} library version 2.9.
This might have to be upgraded to 2.9.3 because it fixes some minor bugs as
below:
{quote}
Changes in 2.9.3
----------------
- DateTimeZone data updated to version 2016c
- Make DateTimeUtils.SYSTEM_MILLIS_PROVIDER public [#357]
- Fix bug when adding months at the maximum limits of integer [#361]
- Add Turkish period translations [#364]
Changes in 2.9.2
----------------
- DateTimeZone data updated to version 2016a (version 2.9 had time-zone data
2015g)
- Fix bug in time-zone binary search [#332]
- Minor fixes to code internals [#339, #326, #344, #350, #343]
- Better document behaviour [#325]
Changes in 2.9.1
----------------
- Fix bug introduced by Long.MIN_VALUE and Long.MAX_VALUE changes [#328]
{quote}
For Spark, there can be some issues by bugs below:
1. Fix bug introduced by Long.MIN_VALUE and Long.MAX_VALUE changes
{code}
DateTime minDT = new DateTime(Long.MIN_VALUE);
{code}
2. Fix bug in time-zone binary search
It looks there is a bug for parsing time-zone. It looks a binary search is
introduced for parsing time-zone but there is a bug. So, If Spark tries to
parse some timezones, this could be possibly an issue.
3.Fix bug when adding months at the maximum limits of integer
It looks it throws an exception, {{ArrayIndexOutOfBoundsException}}, the codes
below:
{code}
DateTime dateTime = new DateTime(1455993437373l);
DateTime date = dateTime.plusMonths(2147483647);
{code}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
