cloud-fan commented on a change in pull request #28576:
URL: https://github.com/apache/spark/pull/28576#discussion_r427847756
##########
File path:
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/util/DateTimeFormatterHelper.scala
##########
@@ -31,17 +31,50 @@ import org.apache.spark.sql.internal.SQLConf
import org.apache.spark.sql.internal.SQLConf.LegacyBehaviorPolicy._
trait DateTimeFormatterHelper {
+ private def getOrDefault(accessor: TemporalAccessor, field: ChronoField,
default: Int): Int = {
+ if (accessor.isSupported(field)) {
+ accessor.get(field)
+ } else {
+ default
+ }
+ }
+
+ protected def toLocalDate(accessor: TemporalAccessor, allowMissingYear:
Boolean): LocalDate = {
+ val year = if (accessor.isSupported(ChronoField.YEAR)) {
+ accessor.get(ChronoField.YEAR)
+ } else if (allowMissingYear) {
+ // To keep backward compatibility with Spark 2.x, we pick 1970 as the
default value of year.
+ 1970
+ } else {
+ throw new SparkUpgradeException("3.0",
+ "Year must be given in the date/timestamp string to be parsed. You can
set " +
+ SQLConf.LEGACY_ALLOW_MISSING_YEAR_DURING_PARSING.key + " to true, to
pick 1970 as " +
+ "the default value of year.", null)
+ }
+ val month = getOrDefault(accessor, ChronoField.MONTH_OF_YEAR, 1)
Review comment:
ah this looks better!
I'm wondering if we still need the legacy config though. It's very hard to
hit the missing year exception, and if it happens it's very likely to be a data
error. Users should fix the data or change the pattern and provide the default
year.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
