From 3c1c9e9070db11a06ca3154a545faff5e3c15d77 Mon Sep 17 00:00:00 2001
From: Nico Poggi <nico.poggi@databricks.com>
Date: Mon, 17 Sep 2018 15:18:16 +0200
Subject: [PATCH] Rebase for PR 87: Add -m for custom master, use SBT_HOME if
 set (#169)

* Add -m for custom master
* Add ability to use own sbt jar, update readme to include -m option
* Add stddev percentage showing
---
 README.md                                     |  2 ++
 build/sbt-launch-lib.bash                     | 11 +++++++-
 .../spark/sql/perf/RunBenchmark.scala         | 28 +++++++++++++------
 3 files changed, 31 insertions(+), 10 deletions(-)

diff --git a/README.md b/README.md
index 794537b..833a80a 100644
--- a/README.md
+++ b/README.md
@@ -18,6 +18,8 @@ Usage: spark-sql-perf [options]
 
   -b <value> | --benchmark <value>
         the name of the benchmark to run
+  -m <value> | --master <value
+        the master url to use
   -f <value> | --filter <value>
         a filter on the name of the queries to run
   -i <value> | --iterations <value>
diff --git a/build/sbt-launch-lib.bash b/build/sbt-launch-lib.bash
index 615f848..2a39936 100755
--- a/build/sbt-launch-lib.bash
+++ b/build/sbt-launch-lib.bash
@@ -26,6 +26,13 @@ else
     declare java_cmd=java
 fi
 
+if test -x "$SBT_HOME"; then
+  echo -e "Using $SBT_HOME as default SBT_HOME - should be the jar name!"
+  # Could be at /usr/share/sbt-launcher-packaging/bin/sbt-launch.jar
+  # so this would be export SBT_HOME=/usr/share/sbt-launcher-packaging/bin/sbt-launch.jar
+  sbt_jar=${SBT_HOME}
+fi
+
 echoerr () {
   echo 1>&2 "$@"
 }
@@ -165,7 +172,9 @@ process_args () {
 }
 
 run() {
-  # no jar? download it.
+  # first check SBT_HOME is present so we use what's already available
+  sbt_jar=$SBT_HOME
+  # if there's no jar let's download it.
   [[ -f "$sbt_jar" ]] || acquire_sbt_jar "$sbt_version" || {
     # still no jar? uh-oh.
     echo "Download failed. Obtain the sbt-launch.jar manually and place it at $sbt_jar"
diff --git a/src/main/scala/com/databricks/spark/sql/perf/RunBenchmark.scala b/src/main/scala/com/databricks/spark/sql/perf/RunBenchmark.scala
index ea557b0..f5c5a93 100644
--- a/src/main/scala/com/databricks/spark/sql/perf/RunBenchmark.scala
+++ b/src/main/scala/com/databricks/spark/sql/perf/RunBenchmark.scala
@@ -17,14 +17,14 @@
 package com.databricks.spark.sql.perf
 
 import java.net.InetAddress
-
+import java.io.File
 import org.apache.spark.sql.SQLContext
 import org.apache.spark.sql.functions._
 import org.apache.spark.{SparkContext, SparkConf}
-
 import scala.util.Try
 
 case class RunConfig(
+    master: String = "local[*]",
     benchmarkName: String = null,
     filter: Option[String] = None,
     iterations: Int = 3,
@@ -37,6 +37,9 @@ object RunBenchmark {
   def main(args: Array[String]): Unit = {
     val parser = new scopt.OptionParser[RunConfig]("spark-sql-perf") {
       head("spark-sql-perf", "0.2.0")
+      opt[String]('m', "master")
+        .action { (x, c) => c.copy(master = x) }
+        .text("the Spark master to use, default to local[*]")
       opt[String]('b', "benchmark")
         .action { (x, c) => c.copy(benchmarkName = x) }
         .text("the name of the benchmark to run")
@@ -64,14 +67,16 @@ object RunBenchmark {
 
   def run(config: RunConfig): Unit = {
     val conf = new SparkConf()
-        .setMaster("local[*]")
-        .setAppName(getClass.getName)
+      .setMaster(config.master)
+      .setAppName(getClass.getName)
 
     val sc = SparkContext.getOrCreate(conf)
     val sqlContext = SQLContext.getOrCreate(sc)
     import sqlContext.implicits._
 
-    sqlContext.setConf("spark.sql.perf.results", new java.io.File("performance").toURI.toString)
+    sqlContext.setConf("spark.sql.perf.results",
+      new File("performance").toURI.toString)
+
     val benchmark = Try {
       Class.forName(config.benchmarkName)
           .newInstance()
@@ -102,7 +107,8 @@ object RunBenchmark {
     experiment.waitForFinish(1000 * 60 * 30)
 
     sqlContext.setConf("spark.sql.shuffle.partitions", "1")
-    experiment.getCurrentRuns()
+      
+    val toShow = experiment.getCurrentRuns()
         .withColumn("result", explode($"results"))
         .select("result.*")
         .groupBy("name")
@@ -110,9 +116,13 @@ object RunBenchmark {
           min($"executionTime") as 'minTimeMs,
           max($"executionTime") as 'maxTimeMs,
           avg($"executionTime") as 'avgTimeMs,
-          stddev($"executionTime") as 'stdDev)
+          stddev($"executionTime") as 'stdDev,
+          (stddev($"executionTime") / avg($"executionTime") * 100) as 'stdDevPercent)
         .orderBy("name")
-        .show(truncate = false)
+        
+    println("Showing at most 100 query results now")
+    toShow.show(100)
+      
     println(s"""Results: sqlContext.read.json("${experiment.resultPath}")""")
 
     config.baseline.foreach { baseTimestamp =>
@@ -136,4 +146,4 @@ object RunBenchmark {
       data.show(truncate = false)
     }
   }
-}
\ No newline at end of file
+}