Upgrade Spark and PySpark version to 3.3.0 (#146)

Co-authored-by: Navin Soni <[email protected]>

Author: navinsoni

buildspec-deploy.yml

@@ -18,6 +18,8 @@ phases:
 
   build:
     commands:
+      - export SBT_OPTS="-Xms1024M -Xmx4G -Xss2M -XX:MaxMetaspaceSize=2G"
+
       # ignore reuse error to allow retry of this build stage
       # when sonatype step has transient error
       - publish-pypi-package --ignore-reuse-error $CODEBUILD_SRC_DIR_ARTIFACT_1/sagemaker-pyspark-sdk/dist/sagemaker_pyspark-*.tar.gz

buildspec-release.yml

@@ -18,23 +18,21 @@ phases:
 
   build:
     commands:
+      - export SBT_OPTS="-Xms1024M -Xmx4G -Xss2M -XX:MaxMetaspaceSize=2G"
+
       # prepare the release (update versions, changelog etc.)
       - git-release --prepare
 
       # spark unit tests and package (no coverage)
       - cd $CODEBUILD_SRC_DIR/sagemaker-spark-sdk
-      - AWS_ACCESS_KEY_ID= AWS_SECRET_ACCESS_KEY= AWS_SESSION_TOKEN=
-        AWS_CONTAINER_CREDENTIALS_RELATIVE_URI=
-        sbt -Dsbt.log.noformat=true clean test package
+      - sbt -Dsbt.log.noformat=true clean test package
 
       # pyspark linters, package and doc build tests
       - cd $CODEBUILD_SRC_DIR/sagemaker-pyspark-sdk
       - tox -e flake8,twine,sphinx
 
       # pyspark unit tests (no coverage)
-      - AWS_ACCESS_KEY_ID= AWS_SECRET_ACCESS_KEY= AWS_SESSION_TOKEN=
-        AWS_CONTAINER_CREDENTIALS_RELATIVE_URI= IGNORE_COVERAGE=-
-        tox -e py27,py36 -- tests/
+      - tox -e py37 -- tests/
 
       # todo consider adding subset of integration tests
 

buildspec.yml

@@ -10,7 +10,7 @@ phases:
       - export JAVA_HOME=/usr/lib/jvm/java-8-openjdk-amd64/bin
 
       # install sbt launcher
-      - curl -LO https://github.com/sbt/sbt/releases/download/v1.1.6/sbt-1.1.6.tgz
+      - curl -LO https://github.com/sbt/sbt/releases/download/v1.7.1/sbt-1.7.1.tgz
       - tar -xf sbt-*.tgz
       - export PATH=$CODEBUILD_SRC_DIR/sbt/bin/:$PATH
       - cd $CODEBUILD_SRC_DIR/sagemaker-spark-sdk
@@ -26,13 +26,13 @@ phases:
 
   build:
     commands:
+      - export SBT_OPTS="-Xms1024M -Xmx4G -Xss2M -XX:MaxMetaspaceSize=2G"
+
       # build spark sdk first, since pyspark package depends on it (even linters)
 
       # spark unit tests
       - cd $CODEBUILD_SRC_DIR/sagemaker-spark-sdk
-      - AWS_ACCESS_KEY_ID= AWS_SECRET_ACCESS_KEY= AWS_SESSION_TOKEN=
-        AWS_CONTAINER_CREDENTIALS_RELATIVE_URI=
-        sbt -Dsbt.log.noformat=true clean coverage test coverageReport
+      - sbt -Dsbt.log.noformat=true clean coverage test coverageReport
 
       # rebuild without coverage instrumentation
       - cd $CODEBUILD_SRC_DIR/sagemaker-spark-sdk
@@ -41,16 +41,16 @@ phases:
       # pyspark linters and unit tests
       - cd $CODEBUILD_SRC_DIR/sagemaker-pyspark-sdk
       - tox -e flake8,twine,sphinx
-      - AWS_ACCESS_KEY_ID= AWS_SECRET_ACCESS_KEY= AWS_SESSION_TOKEN=
-        AWS_CONTAINER_CREDENTIALS_RELATIVE_URI=
-        tox -e py36,stats -- tests/
+      - tox -e py37,stats -- tests/
 
       # spark integration tests
       - cd $CODEBUILD_SRC_DIR/integration-tests/sagemaker-spark-sdk
-      - test_cmd="sbt -Dsbt.log.noformat=true it:test"
-      - execute-command-if-has-matching-changes "$test_cmd" "src/" "test/" "build.sbt" "buildspec.yml"
+      - sbt -Dsbt.log.noformat=true it:test
+        # - test_cmd="sbt -Dsbt.log.noformat=true it:test"
+        # - execute-command-if-has-matching-changes "$test_cmd" "src/" "test/" "build.sbt" "buildspec.yml"
 
       # pyspark integration tests
       - cd $CODEBUILD_SRC_DIR/sagemaker-pyspark-sdk
-      - test_cmd="IGNORE_COVERAGE=- tox -e py36 -- $CODEBUILD_SRC_DIR/integration-tests/sagemaker-pyspark-sdk/tests/ -n 10 --boxed --reruns 2"
-      - execute-command-if-has-matching-changes "$test_cmd" "src/" "tests/" "setup.*" "requirements.txt" "tox.ini" "buildspec.yml"
+      - IGNORE_COVERAGE=- tox -e py37 -- $CODEBUILD_SRC_DIR/integration-tests/sagemaker-pyspark-sdk/tests/ -n 10 --boxed --reruns 2
+        # - test_cmd="IGNORE_COVERAGE=- tox -e py37 -- $CODEBUILD_SRC_DIR/integration-tests/sagemaker-pyspark-sdk/tests/ -n 10 --boxed --reruns 2"
+        # - execute-command-if-has-matching-changes "$test_cmd" "src/" "tests/" "setup.*" "requirements.txt" "tox.ini" "buildspec.yml"

sagemaker-pyspark-sdk/setup.py

@@ -36,17 +36,19 @@ def read_version():
             print("Could not create dir {0}".format(TEMP_PATH), file=sys.stderr)
             exit(1)
 
-        p = subprocess.Popen("sbt printClasspath".split(),
-                             stdout=subprocess.PIPE,
-                             stderr=subprocess.PIPE,
-                             cwd="../sagemaker-spark-sdk/")
+        p = subprocess.Popen(
+            "sbt printClasspath".split(),
+            stdout=subprocess.PIPE,
+            stderr=subprocess.PIPE,
+            cwd="../sagemaker-spark-sdk/",
+        )
 
         output, errors = p.communicate()
 
         classpath = []
         # Java Libraries to include.
-        java_libraries = ['aws', 'sagemaker', 'hadoop', 'htrace']
-        for line in output.decode('utf-8').splitlines():
+        java_libraries = ["aws", "sagemaker", "hadoop", "htrace"]
+        for line in output.decode("utf-8").splitlines():
             path = str(line.strip())
             if path.endswith(".jar") and os.path.exists(path):
                 jar = os.path.basename(path).lower()
@@ -65,8 +67,10 @@ def read_version():
 
     else:
         if not os.path.exists(JARS_TARGET):
-            print("You need to be in the sagemaker-pyspark-sdk root folder to package",
-                  file=sys.stderr)
+            print(
+                "You need to be in the sagemaker-pyspark-sdk root folder to package",
+                file=sys.stderr,
+            )
             exit(-1)
 
     setup(
@@ -76,32 +80,30 @@ def read_version():
         author="Amazon Web Services",
         url="https://github.com/aws/sagemaker-spark",
         license="Apache License 2.0",
+        python_requires=">= 3.7",
         zip_safe=False,
-
-        packages=["sagemaker_pyspark",
-                  "sagemaker_pyspark.algorithms",
-                  "sagemaker_pyspark.transformation",
-                  "sagemaker_pyspark.transformation.deserializers",
-                  "sagemaker_pyspark.transformation.serializers",
-                  "sagemaker_pyspark.jars",
-                  "sagemaker_pyspark.licenses"],
-
+        packages=[
+            "sagemaker_pyspark",
+            "sagemaker_pyspark.algorithms",
+            "sagemaker_pyspark.transformation",
+            "sagemaker_pyspark.transformation.deserializers",
+            "sagemaker_pyspark.transformation.serializers",
+            "sagemaker_pyspark.jars",
+            "sagemaker_pyspark.licenses",
+        ],
         package_dir={
             "sagemaker_pyspark": "src/sagemaker_pyspark",
             "sagemaker_pyspark.jars": "deps/jars",
-            "sagemaker_pyspark.licenses": "licenses"
+            "sagemaker_pyspark.licenses": "licenses",
         },
         include_package_data=True,
-
         package_data={
             "sagemaker_pyspark.jars": ["*.jar"],
-            "sagemaker_pyspark.licenses": ["*.txt"]
+            "sagemaker_pyspark.licenses": ["*.txt"],
         },
-
         scripts=["bin/sagemakerpyspark-jars", "bin/sagemakerpyspark-emr-jars"],
-
         install_requires=[
-            "pyspark==2.4.0",
+            "pyspark==3.3.0",
             "numpy",
         ],
     )

sagemaker-pyspark-sdk/src/sagemaker_pyspark/algorithms/XGBoostSageMakerEstimator.py

@@ -380,7 +380,7 @@ def __init__(self,
         if uid is None:
             uid = Identifiable._randomUID()
 
-        kwargs = locals()
+        kwargs = locals().copy()
         del kwargs['self']
         super(XGBoostSageMakerEstimator, self).__init__(**kwargs)
 

sagemaker-pyspark-sdk/tests/namepolicy_test.py

@@ -28,29 +28,29 @@ def with_spark_context():
 def test_CustomNamePolicyFactory():
     policy_factory = CustomNamePolicyFactory("jobName", "modelname", "epconfig", "ep")
     java_obj = policy_factory._to_java()
-    assert(isinstance(java_obj, JavaObject))
-    assert(java_obj.getClass().getSimpleName() == "CustomNamePolicyFactory")
+    assert (isinstance(java_obj, JavaObject))
+    assert (java_obj.getClass().getSimpleName() == "CustomNamePolicyFactory")
     policy_name = java_obj.createNamePolicy().getClass().getSimpleName()
-    assert(policy_name == "CustomNamePolicy")
+    assert (policy_name == "CustomNamePolicy")
 
 
 def test_CustomNamePolicyWithTimeStampSuffixFactory():
     policy_factory = CustomNamePolicyWithTimeStampSuffixFactory("jobName", "modelname",
                                                                 "epconfig", "ep")
     java_obj = policy_factory._to_java()
-    assert(isinstance(java_obj, JavaObject))
+    assert (isinstance(java_obj, JavaObject))
     assert (java_obj.getClass().getSimpleName() == "CustomNamePolicyWithTimeStampSuffixFactory")
     policy_name = java_obj.createNamePolicy().getClass().getSimpleName()
-    assert(policy_name == "CustomNamePolicyWithTimeStampSuffix")
+    assert (policy_name == "CustomNamePolicyWithTimeStampSuffix")
 
 
 def test_CustomNamePolicyWithTimeStampSuffix():
     name_policy = CustomNamePolicyWithTimeStampSuffix("jobName", "modelname", "epconfig", "ep")
-    assert(isinstance(name_policy._to_java(), JavaObject))
-    assert(name_policy._call_java("trainingJobName") != "jobName")
-    assert(name_policy._call_java("modelName") != "modelname")
-    assert(name_policy._call_java("endpointConfigName") != "epconfig")
-    assert(name_policy._call_java("endpointName") != "ep")
+    assert (isinstance(name_policy._to_java(), JavaObject))
+    assert (name_policy._call_java("trainingJobName") != "jobName")
+    assert (name_policy._call_java("modelName") != "modelname")
+    assert (name_policy._call_java("endpointConfigName") != "epconfig")
+    assert (name_policy._call_java("endpointName") != "ep")
 
     assert (name_policy._call_java("trainingJobName").startswith("jobName"))
     assert (name_policy._call_java("modelName").startswith("modelname"))

sagemaker-pyspark-sdk/tox.ini

@@ -1,5 +1,5 @@
 [tox]
-envlist = flake8,twine,sphinx,py36,stats
+envlist = flake8,twine,sphinx,py37,stats
 skip_missing_interpreters = False
 
 [testenv]
@@ -38,8 +38,8 @@ basepython = python3
 deps =
     twine>=1.12.0
 commands =
-    python setup.py sdist
-    twine check dist/*.tar.gz
+    - python setup.py sdist
+    - twine check dist/*.tar.gz
 
 [testenv:flake8]
 basepython=python3

sagemaker-spark-sdk/build.sbt

@@ -14,11 +14,11 @@ scmInfo := Some(
 )
 licenses := Seq("Apache License, Version 2.0" -> url("https://aws.amazon.com/apache2.0"))
 
-scalaVersion := "2.11.7"
+scalaVersion := "2.12.16"
 
 // to change the version of spark add -DSPARK_VERSION=2.x.x when running sbt
 // for example: "sbt -DSPARK_VERSION=2.1.1 clean compile test doc package"
-val sparkVersion = System.getProperty("SPARK_VERSION", "2.4.0")
+val sparkVersion = System.getProperty("SPARK_VERSION", "3.3.0")
 
 lazy val SageMakerSpark = (project in file("."))
 
@@ -29,16 +29,18 @@ version := {
 }
 
 libraryDependencies ++= Seq(
-  "org.apache.hadoop" % "hadoop-aws" % "2.8.1",
-  "com.amazonaws" % "aws-java-sdk-s3" % "1.11.835",
-  "com.amazonaws" % "aws-java-sdk-sts" % "1.11.835",
-  "com.amazonaws" % "aws-java-sdk-sagemaker" % "1.11.835",
-  "com.amazonaws" % "aws-java-sdk-sagemakerruntime" % "1.11.835",
+  "org.apache.hadoop" % "hadoop-aws" % "3.3.1",
+  "com.amazonaws" % "aws-java-sdk-s3" % "1.12.262",
+  "com.amazonaws" % "aws-java-sdk-sts" % "1.12.262",
+  "com.amazonaws" % "aws-java-sdk-sagemaker" % "1.12.262",
+  "com.amazonaws" % "aws-java-sdk-sagemakerruntime" % "1.12.262",
   "org.apache.spark" %% "spark-core" % sparkVersion % "provided",
   "org.apache.spark" %% "spark-mllib" % sparkVersion % "provided",
   "org.apache.spark" %% "spark-sql" % sparkVersion % "provided",
-  "org.scalatest" %% "scalatest" % "3.0.4" % "test",
-  "org.mockito" % "mockito-all" % "1.10.19" % "test"
+  "org.scoverage" %% "scalac-scoverage-plugin" % "1.4.2" % "provided",
+  "org.scalatest" %% "scalatest" % "3.0.9" % "test",
+  "org.scala-sbt" %% "compiler-bridge" % "1.7.1" % "test",
+  "org.mockito" % "mockito-all" % "2.0.2-beta" % "test"
 )
 
 // add a task to print the classpath. Also use the packaged JAR instead
@@ -48,8 +50,11 @@ lazy val printClasspath = taskKey[Unit]("Dump classpath")
 printClasspath := (fullClasspath in Runtime value) foreach { e => println(e.data) }
 
 // set coverage threshold
-coverageMinimum := 90
 coverageFailOnMinimum := true
+coverageMinimumStmtTotal := 90
+coverageMinimumBranchTotal := 90
+coverageMinimumStmtPerPackage := 83
+coverageMinimumBranchPerPackage := 75
 
 // make scalastyle gate the build
 (compile in Compile) := {

sagemaker-spark-sdk/project/build.properties

@@ -1 +1 @@
-sbt.version=1.2.1
+sbt.version=1.7.1

sagemaker-spark-sdk/project/plugins.sbt

@@ -1,4 +1,4 @@
 addSbtPlugin("org.scalastyle" %% "scalastyle-sbt-plugin" % "1.0.0")
 addSbtPlugin("com.jsuereth" % "sbt-pgp" % "1.1.0")
 addSbtPlugin("org.xerial.sbt" % "sbt-sonatype" % "3.8.1")
-addSbtPlugin("org.scoverage" % "sbt-scoverage" % "1.6.0")
+addSbtPlugin("org.scoverage" % "sbt-scoverage" % "2.0.0")