Github user HyukjinKwon commented on the issue:
https://github.com/apache/spark/pull/14279
Sigh.. I will proceed but please let me note that this will also bring a
breaking change for reading as well not only writing..
It seems there is a backwards compatibility problem with using default date
format for `dateFormat` and `timestampFormat` instead of `null` when reading.
When it was `null`, it uses `DateTimeUtils.stringToTime` for
`TimestampType` and `DateTimeUtils.millisToDays` for `DateType`. It seems okay
with `DateTimeUtils.millisToDays` for `DateType` but the problem is with
`DateTimeUtils.stringToTime()`.
It seems `DateTimeUtils.stringToTime()` for `TimestampType` supports to
read several other date formats such as
- `2000-01-01T00:00GMT+01:00`
- `2000-01-01T00:00+01:00`
- `2000-01-01 00:00:00.0`
As previous Spark version writes `2000-01-01 00:00:00.0`, setting the
default value for `timestampFormat` to ISO8601 format will also break the
backwards compatibility.
we should explicitly set this for reading as well not only writing for the
same timestamps handling between Spark 1.x and Spark 2.x.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastruct...@apache.org or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
