Skip to content

[SPARK-31426][SQL] Fix perf regressions of toJavaTimestamp/fromJavaTimestamp #28189

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
MaxGekk wants to merge 8 commits into from
Closed

[SPARK-31426][SQL] Fix perf regressions of toJavaTimestamp/fromJavaTimestamp #28189

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
8 commits
Select commit Hold shift + click to select a range
a56c752
Rewrite toJavaTimestamp/fromJavaTimestamp
MaxGekk Apr 11, 2020
794cc5e
Re-gen DateTimeRebaseBenchmark results on JDK 8
MaxGekk Apr 11, 2020
fd06ecc
Re-gen DateTimeBenchmark results on JDK 8
MaxGekk Apr 11, 2020
94dc31e
Re-gen benchmark results on JDK 11
MaxGekk Apr 11, 2020
055ac97
Merge remote-tracking branch 'origin/master' into optimize-to-from-ja…
MaxGekk Apr 13, 2020
b331ad8
Merge remote-tracking branch 'origin/master' into optimize-to-from-ja…
MaxGekk Apr 13, 2020
5dc9a3a
Re-gen DateTimeRebaseBenchmark results on JDK 11
MaxGekk Apr 13, 2020
a28f2a1
Re-gen DateTimeRebaseBenchmark results on JDK 8
MaxGekk Apr 13, 2020
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
21 changes: 8 additions & 13 deletions sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/util/DateTimeUtils.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -148,8 +148,12 @@ object DateTimeUtils {
* @return A `java.sql.Timestamp` from number of micros since epoch.
*/
def toJavaTimestamp(us: SQLTimestamp): Timestamp = {
val ldt = microsToInstant(us).atZone(ZoneId.systemDefault()).toLocalDateTime
Timestamp.valueOf(ldt)
val rebasedMicros = rebaseGregorianToJulianMicros(us)
val seconds = Math.floorDiv(rebasedMicros, MICROS_PER_SECOND)
val ts = new Timestamp(seconds * MILLIS_PER_SECOND)
val nanos = (rebasedMicros - seconds * MICROS_PER_SECOND) * NANOS_PER_MICROS
ts.setNanos(nanos.toInt)
ts
}

/**
Expand All @@ -172,17 +176,8 @@ object DateTimeUtils {
* @return The number of micros since epoch from `java.sql.Timestamp`.
*/
def fromJavaTimestamp(t: Timestamp): SQLTimestamp = {
val era = if (t.before(julianCommonEraStart)) 0 else 1
val localDateTime = LocalDateTime.of(
t.getYear + 1900, t.getMonth + 1, 1,
t.getHours, t.getMinutes, t.getSeconds, t.getNanos)
.`with`(ChronoField.ERA, era)
// Add days separately to convert dates existed in Julian calendar but not
// in Proleptic Gregorian calendar. For example, 1000-02-29 is valid date
// in Julian calendar because 1000 is a leap year but 1000 is not a leap
// year in Proleptic Gregorian calendar. And 1000-02-29 doesn't exist in it.
.plusDays(t.getDate - 1) // Returns the next valid date after `date.getDate - 1` days
instantToMicros(localDateTime.atZone(ZoneId.systemDefault).toInstant)
val micros = millisToMicros(t.getTime) + (t.getNanos / NANOS_PER_MICROS) % MICROS_PER_MILLIS
Copy link
Member

@dongjoon-hyun dongjoon-hyun Apr 14, 2020 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This seems to be missing in branch-3.0.

$ git grep millisToMicros
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/util/DateTimeUtils.scala: 
val micros = millisToMicros(t.getTime) + (t.getNanos / NANOS_PER_MICROS) % MICROS_PER_MILLIS

rebaseJulianToGregorianMicros(micros)
}

/**
Expand Down
4 changes: 2 additions & 2 deletions sql/catalyst/src/test/scala/org/apache/spark/sql/catalyst/util/DateTimeUtilsSuite.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -496,8 +496,8 @@ class DateTimeUtilsSuite extends SparkFunSuite with Matchers with SQLHelper {
test("2016-03-13 02:00:00", tz, "2016-03-13 10:00:00.0")
test("2016-03-13 03:00:00", tz, "2016-03-13 10:00:00.0")
test("2016-11-06 00:59:59", tz, "2016-11-06 07:59:59.0")
test("2016-11-06 01:00:00", tz, "2016-11-06 08:00:00.0")
test("2016-11-06 01:59:59", tz, "2016-11-06 08:59:59.0")
test("2016-11-06 01:00:00", tz, "2016-11-06 09:00:00.0")
Copy link
Member Author

@MaxGekk MaxGekk Apr 11, 2020

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Screenshot 2020-04-11 at 19 01 34

Copy link
Contributor

@cloud-fan cloud-fan Apr 13, 2020

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

what's the behavior of Spark 2.4?

Copy link
Member Author

@MaxGekk MaxGekk Apr 13, 2020

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

See the same test in 2.4 https://github.com/apache/spark/blob/branch-2.4/sql/catalyst/src/test/scala/org/apache/spark/sql/catalyst/util/DateTimeUtilsSuite.scala#L580-L581 but it sets PST time zone.

Copy link
Contributor

@cloud-fan cloud-fan Apr 13, 2020

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

as we have looked into the JDK code before, PST is treated as LA. So the behavior is the same as 2.4, and it's great!

can we highlight it in the Does this PR introduce any user-facing change? section? Say it's changed back to the 2.4 behavior.

Copy link
Member Author

@MaxGekk MaxGekk Apr 13, 2020

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

done

test("2016-11-06 01:59:59", tz, "2016-11-06 09:59:59.0")
test("2016-11-06 02:00:00", tz, "2016-11-06 10:00:00.0")
}
}
Expand Down
Loading