[SPARK-48025][SQL][TESTS] Fix org.apache.spark.sql.execution.benchmark.DateTimeBenchmark #46261

yaooqinn · 2024-04-28T07:17:33Z

What changes were proposed in this pull request?

This PR fixes several issues in org.apache.spark.sql.execution.benchmark.DateTimeBenchmark

Misuse trunc function, a.k.a, the parameter order is wrong in reverse order
Some benchmarks not compatible with ANSI by default

Why are the changes needed?

restore benchmark cases

Does this PR introduce any user-facing change?

no

How was this patch tested?

benchmark

Was this patch authored or co-authored using generative AI tooling?

no

…k.DateTimeBenchmark

yaooqinn · 2024-04-28T07:32:49Z

sql/core/src/test/scala/org/apache/spark/sql/execution/benchmark/DateTimeBenchmark.scala

@@ -75,7 +81,7 @@ object DateTimeBenchmark extends SqlBasedBenchmark {
            doBenchmark(N, s"$dt + interval 1 month 2 day")
          }
          benchmark.addCase("date + interval(m, d, ms)") { _ =>
-            doBenchmark(N, s"$dt + interval 1 month 2 day 5 hour")
+            doBenchmarkAnsiOff(N, s"$dt + interval 1 month 2 day 5 hour")


illegal hour portion for ansi date add

yaooqinn · 2024-04-28T07:33:16Z

sql/core/src/test/scala/org/apache/spark/sql/execution/benchmark/DateTimeBenchmark.scala

@@ -161,7 +167,7 @@ object DateTimeBenchmark extends SqlBasedBenchmark {
          }
          val dateExpr = "cast(timestamp_seconds(id) as date)"
          Seq("year", "yyyy", "yy", "mon", "month", "mm").foreach { level =>
-            run(N, s"trunc $level", s"trunc('$level', $dateExpr)")
+            run(N, s"trunc $level", s"trunc($dateExpr, '$level')")


the parameter order is wrong

yaooqinn · 2024-04-28T07:33:58Z

sql/core/src/test/scala/org/apache/spark/sql/execution/benchmark/DateTimeBenchmark.scala

@@ -171,7 +177,7 @@ object DateTimeBenchmark extends SqlBasedBenchmark {
          run(n, "to timestamp str", timestampStrExpr)
          run(n, "to_timestamp", s"to_timestamp($timestampStrExpr, $pattern)")
          run(n, "to_unix_timestamp", s"to_unix_timestamp($timestampStrExpr, $pattern)")
-          val dateStrExpr = "concat('2019-01-', lpad(mod(id, 25), 2, '0'))"
+          val dateStrExpr = "concat('2019-01-', lpad(mod(id, 25) + 1, 2, '0'))"


avoid invalid date value 2019-01-00

yaooqinn · 2024-04-28T07:49:43Z

sql/core/benchmarks/DateTimeBenchmark-jdk21-results.txt

+trunc year wholestage off                           866            867           2         11.5          86.6       1.0X
+trunc year wholestage on                            886            888           2         11.3          88.6       1.0X

-OpenJDK 64-Bit Server VM 21.0.2+13-LTS on Linux 6.5.0-1016-azure
+OpenJDK 64-Bit Server VM 21.0.3+9-LTS on Linux 6.5.0-1018-azure
 AMD EPYC 7763 64-Core Processor
 trunc yyyy:                               Best Time(ms)   Avg Time(ms)   Stdev(ms)    Rate(M/s)   Per Row(ns)   Relative
 ------------------------------------------------------------------------------------------------------------------------
-trunc yyyy wholestage off                           301            305           5         33.2          30.1       1.0X
-trunc yyyy wholestage on                            271            275           3         36.9          27.1       1.1X
+trunc yyyy wholestage off                           882            883           1         11.3          88.2       1.0X
+trunc yyyy wholestage on                            881            887           5         11.4          88.1       1.0X

-OpenJDK 64-Bit Server VM 21.0.2+13-LTS on Linux 6.5.0-1016-azure
+OpenJDK 64-Bit Server VM 21.0.3+9-LTS on Linux 6.5.0-1018-azure
 AMD EPYC 7763 64-Core Processor
 trunc yy:                                 Best Time(ms)   Avg Time(ms)   Stdev(ms)    Rate(M/s)   Per Row(ns)   Relative
 ------------------------------------------------------------------------------------------------------------------------
-trunc yy wholestage off                             295            295           0         33.9          29.5       1.0X
-trunc yy wholestage on                              270            277           4         37.0          27.0       1.1X
+trunc yy wholestage off                             875            877           3         11.4          87.5       1.0X
+trunc yy wholestage on                              881            890           9         11.3          88.1       1.0X

-OpenJDK 64-Bit Server VM 21.0.2+13-LTS on Linux 6.5.0-1016-azure
+OpenJDK 64-Bit Server VM 21.0.3+9-LTS on Linux 6.5.0-1018-azure
 AMD EPYC 7763 64-Core Processor
 trunc mon:                                Best Time(ms)   Avg Time(ms)   Stdev(ms)    Rate(M/s)   Per Row(ns)   Relative
 ------------------------------------------------------------------------------------------------------------------------
-trunc mon wholestage off                            296            297           1         33.8          29.6       1.0X
-trunc mon wholestage on                             275            278           2         36.4          27.5       1.1X
+trunc mon wholestage off                            838            839           2         11.9          83.8       1.0X
+trunc mon wholestage on                             833            839           5         12.0          83.3       1.0X

-OpenJDK 64-Bit Server VM 21.0.2+13-LTS on Linux 6.5.0-1016-azure
+OpenJDK 64-Bit Server VM 21.0.3+9-LTS on Linux 6.5.0-1018-azure
 AMD EPYC 7763 64-Core Processor
 trunc month:                              Best Time(ms)   Avg Time(ms)   Stdev(ms)    Rate(M/s)   Per Row(ns)   Relative
 ------------------------------------------------------------------------------------------------------------------------
-trunc month wholestage off                          304            308           6         32.9          30.4       1.0X
-trunc month wholestage on                           276            280           4         36.2          27.6       1.1X
+trunc month wholestage off                          839            841           4         11.9          83.9       1.0X
+trunc month wholestage on                           837            839           2         12.0          83.7       1.0X

-OpenJDK 64-Bit Server VM 21.0.2+13-LTS on Linux 6.5.0-1016-azure
+OpenJDK 64-Bit Server VM 21.0.3+9-LTS on Linux 6.5.0-1018-azure
 AMD EPYC 7763 64-Core Processor
 trunc mm:                                 Best Time(ms)   Avg Time(ms)   Stdev(ms)    Rate(M/s)   Per Row(ns)   Relative
 ------------------------------------------------------------------------------------------------------------------------
-trunc mm wholestage off                             297            298           1         33.6          29.7       1.0X
-trunc mm wholestage on                              274            276           2         36.5          27.4       1.1X
+trunc mm wholestage off                             846            847           1         11.8          84.6       1.0X
+trunc mm wholestage on                              834            845          11         12.0          83.4       1.0X


The results are corrected

HyukjinKwon

Seems fine but do we have this issue in other banchmarks too?

yaooqinn · 2024-04-28T09:07:37Z

Thank you @HyukjinKwon. I have dispatched a CI task to audit, https://github.com/yaooqinn/spark/actions/runs/8866497258

yaooqinn · 2024-04-28T09:53:41Z

Merged to master

…k.DateTimeBenchmark ### What changes were proposed in this pull request? This PR fixes several issues in org.apache.spark.sql.execution.benchmark.DateTimeBenchmark - Misuse `trunc` function, a.k.a, the parameter order is wrong in reverse order - Some benchmarks not compatible with ANSI by default ### Why are the changes needed? restore benchmark cases ### Does this PR introduce _any_ user-facing change? no ### How was this patch tested? benchmark ### Was this patch authored or co-authored using generative AI tooling? no Closes apache#46261 from yaooqinn/SPARK-48025. Authored-by: Kent Yao <yao@apache.org> Signed-off-by: Kent Yao <yao@apache.org>

[SPARK-48025][SQL][TESTS] Fix org.apache.spark.sql.execution.benchmar…

e237f57

…k.DateTimeBenchmark

github-actions bot added the SQL label Apr 28, 2024

yaooqinn commented Apr 28, 2024

View reviewed changes

results

b515ba0

yaooqinn commented Apr 28, 2024

View reviewed changes

HyukjinKwon approved these changes Apr 28, 2024

View reviewed changes

yaooqinn closed this in 0bf3945 Apr 28, 2024

yaooqinn deleted the SPARK-48025 branch April 28, 2024 12:02

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[SPARK-48025][SQL][TESTS] Fix org.apache.spark.sql.execution.benchmark.DateTimeBenchmark #46261

[SPARK-48025][SQL][TESTS] Fix org.apache.spark.sql.execution.benchmark.DateTimeBenchmark #46261

yaooqinn commented Apr 28, 2024

yaooqinn Apr 28, 2024

yaooqinn Apr 28, 2024

yaooqinn Apr 28, 2024

yaooqinn Apr 28, 2024

HyukjinKwon left a comment

yaooqinn commented Apr 28, 2024

yaooqinn commented Apr 28, 2024

[SPARK-48025][SQL][TESTS] Fix org.apache.spark.sql.execution.benchmark.DateTimeBenchmark #46261

[SPARK-48025][SQL][TESTS] Fix org.apache.spark.sql.execution.benchmark.DateTimeBenchmark #46261

Conversation

yaooqinn commented Apr 28, 2024

What changes were proposed in this pull request?

Why are the changes needed?

Does this PR introduce any user-facing change?

How was this patch tested?

Was this patch authored or co-authored using generative AI tooling?

yaooqinn Apr 28, 2024

Choose a reason for hiding this comment

yaooqinn Apr 28, 2024

Choose a reason for hiding this comment

yaooqinn Apr 28, 2024

Choose a reason for hiding this comment

yaooqinn Apr 28, 2024

Choose a reason for hiding this comment

HyukjinKwon left a comment

Choose a reason for hiding this comment

yaooqinn commented Apr 28, 2024

yaooqinn commented Apr 28, 2024