misc(Python): Replace python 2 with python 3

Python 2 has reached EOL last year and should not be used anymore. This commit replaces all references to the "python" binary with the more explicit "python3" binary. If desired, the build can still be performed for Python 2 by settings the "PYTHON_EXECUTABLE" environment variable to an appropriate version. Additionally, python wheels are the preferred way to distribute python code (see https://packaging.python.org/discussions/wheel-vs-egg/). This commit additionally builds the job-server-python wheel.

Spark-2.4 does not support python >= 3.8 (see apache/spark#26194) leading to failed test cases (TypeError: an integer is required (got type bytes)). If you encounter these issues try to state a python executable < 3.8 explicitly.

build.sbt

@@ -115,8 +115,8 @@ lazy val jobServerExtrasSettings = revolverSettings ++ Assembly.settings ++ publ
 lazy val jobServerApiSettings = Seq(libraryDependencies ++= sparkDeps ++ sparkExtraDeps)
 
 lazy val testPython = taskKey[Unit]("Launch a sub process to run the Python tests")
-lazy val buildPython = taskKey[Unit]("Build the python side of python support into an egg")
-lazy val buildPyExamples = taskKey[Unit]("Build the examples of python jobs into an egg")
+lazy val buildPython = taskKey[Unit]("Build the python side of python support into a wheel")
+lazy val buildPyExamples = taskKey[Unit]("Build the examples of python jobs into a wheel")
 
 lazy val jobServerPythonSettings = revolverSettings ++ Assembly.settings ++ publishSettings ++ Seq(
   libraryDependencies ++= sparkPythonDeps,

ci/install-python-dependencies.sh

@@ -1,6 +1,5 @@
 #!/usr/bin/env bash
 set -e
-pip install --upgrade pip
-pip install --user pyhocon
+pip3 install --upgrade pip
 pip3 install --user pyhocon
-pip install --user pycodestyle
+pip3 install --user pycodestyle

doc/python.md

@@ -45,10 +45,10 @@ A basic config supporting Python might look like:
         python {
           paths = [
             ${SPARK_HOME}/python,
-            "/home/user/spark-jobserver/job-server-extras/job-server-python/target/python/spark_jobserver_python-0.8.0-py2.7.egg"
+            "/home/user/spark-jobserver/job-server-extras/job-server-python/target/python/spark_jobserver_python-0.10.1-py3-none-any.whl"
           ]
 
-          # The default value in application.conf is "python"
+          # The default value in application.conf is "python3"
           executable = "python3"
         }
       }

job-server-extras/src/test/scala/spark/jobserver/python/PythonSessionContextFactorySpec.scala

@@ -165,8 +165,8 @@ with BeforeAndAfterAll {
     it("should successfully run jobs using python3", WindowsIgnore) {
       val factory = new TestPythonSessionContextFactory()
       val p3Config = ConfigFactory.parseString(
-        """
-          |python.executable = "python3"
+        s"""
+          |python.executable = "${sys.env.getOrElse("PYTHON_EXECUTABLE", "python3")}"
         """.stripMargin).withFallback(config)
       context = factory.makeContext(sparkConf, p3Config, "test-create")
       runSessionTest(factory, context, p3Config)

job-server-extras/src/test/scala/spark/jobserver/python/PythonSparkContextFactorySpec.scala

@@ -104,7 +104,7 @@ object PythonSparkContextFactorySpec {
        |  "${originalPythonPath.getOrElse("")}"
        |]
        |
-      |python.executable = "python"
+      |python.executable = "${sys.env.getOrElse("PYTHON_EXECUTABLE", "python3")}"
       |${JobserverConfig.IS_SPARK_SESSION_HIVE_ENABLED} = true
     """.replace("\\", "\\\\") // Windows-compatibility
       .stripMargin)

job-server-python/src/python/build.cmd

@@ -3,8 +3,13 @@ set SJS_VERSION=%1
 setlocal
 cd %~dp0
 
-python %~dp0%2 build --build-base ../../target/python^
+set PYTHON_EXECUTABLE=%2
+
+PYTHON_EXECUTABLE %~dp0%3 build --build-base ../../target/python^
  egg_info --egg-base ../../target/python^
  bdist_egg --bdist-dir /tmp/bdist --dist-dir ../../target/python --skip-build
 
+PYTHON_EXECUTABLE %~dp0%3 build --build-base ../../target/python^
+ bdist_wheel --bdist-dir /tmp/bdist --dist-dir ../../target/python --skip-build
+
 endlocal

job-server-python/src/python/build.sh

@@ -1,5 +1,8 @@
 #!/usr/bin/env bash
 export SJS_VERSION=$1
-python $2 build --build-base ../../target/python \
+$2 $3 build --build-base ../../target/python \
 egg_info --egg-base ../../target/python \
 bdist_egg --bdist-dir /tmp/bdist --dist-dir ../../target/python --skip-build
+
+$2 $3 build --build-base ../../target/python \
+bdist_wheel --bdist-dir /tmp/bdist --dist-dir ../../target/python --skip-build

job-server-python/src/python/run-tests.cmd

@@ -15,7 +15,8 @@ setlocal
 cd %~dp0
 
 set PYTHONPATH=%~dp0;%SPARK_HOME%\python\lib\pyspark.zip;%SPARK_HOME%\python\lib\py4j-0.9-src.zip;%PYTHONPATH%
-python test\apitests.py
+set PYSPARK_PYTHON=%1
+PYSPARK_PYTHON test\apitests.py
 set exitCode=%ERRORLEVEL%
 REM This sleep is here so that all of Spark's shutdown stdout if written before we exit,
 REM so that we return cleanly to the command prompt.

job-server-python/src/python/run-tests.sh

@@ -1,5 +1,7 @@
 #!/usr/bin/env bash
-PYTHONPATH=.:$SPARK_HOME/python/lib/pyspark.zip:$SPARK_HOME/python/lib/py4j-0.10.7-src.zip:$PYTHONPATH python test/apitests.py
+PYTHONPATH=.:$SPARK_HOME/python/lib/pyspark.zip:$SPARK_HOME/python/lib/py4j-0.10.7-src.zip:$PYTHONPATH \
+PYSPARK_PYTHON=$1 \
+PYSPARK_PYTHON test/apitests.py
 exitCode=$?
 #This sleep is here so that all of Spark's shutdown stdout if written before we exit,
 #so that we return cleanly to the command prompt.

job-server-python/src/test/scala/spark/jobserver/python/SubprocessSpec.scala

@@ -152,10 +152,17 @@ class SubprocessSpec extends FunSpec with Matchers with BeforeAndAfter with Befo
   )
 
   private def setupPythonProcess(port: String, token: String): scala.sys.process.ProcessBuilder = {
+    // Spark-2.4 does not support python >= 3.8 (see https://github.com/apache/spark/pull/26194) leading to
+    // failed test cases (TypeError: an integer is required (got type bytes)). If you encounter these issues
+    // try to state a python executable < 3.8 explicitly.
+    // TODO: remove comment after migration to Spark-3.0
+    val pythonExecutable = sys.env.getOrElse("PYTHON_EXECUTABLE", "python3")
     Process(
-      Seq("python", "-m", "sparkjobserver.subprocess", port, token),
+      Seq(pythonExecutable, "-m", "sparkjobserver.subprocess", port, token),
       None,
-      "PYTHONPATH" -> pythonPath)
+      "PYTHONPATH" -> pythonPath,
+      "PYSPARK_PYTHON" -> pythonExecutable
+    )
   }
 
   describe("The python subprocess") {

job-server/src/main/resources/application.conf

@@ -207,7 +207,7 @@ spark {
       #Any locations of python libraries to be included on the PYTHONPATH when running Python jobs
       paths = []
       #The shell command to run when launching the subprocess for python jobs
-      executable = "python"
+      executable = "python3"
     }
 
     # All the above sections have higher precedence because those properties are added directly to

notes/0.11.0.markdown

@@ -1,7 +1,11 @@
-#Scala #akka @ApacheSpark
+# Scala #akka @ApacheSpark
 
 ## Breaking Changes:
-The authentication management has been refactored. As a consequence, existing `shiro` configuration sections have to be
-adapted. Previously, `shiro` configuration had to be provided at root level without a prefix. Now, `shiro` configuration
-has to be embedded in an additional `access-control` block. An example configuration is available in the
-[sources](../job-server/src/main/resources/application.conf#L322).
+
+- The authentication management has been refactored. As a consequence, existing `shiro` configuration sections have to
+  be adapted. Previously, `shiro` configuration had to be provided at root level without a prefix. Now, `shiro`
+  configuration has to be embedded in an additional `access-control` block. An example configuration is available in the
+  [sources](../job-server/src/main/resources/application.conf#L322).
+- The default Python version is switched from Python 2 to Python 3 (namely, from `python` to `python3`). If you would
+  still like to use Python 2, adapt the Python executable in
+  the [application.conf](../job-server/src/main/resources/application.conf#L210).

project/PythonTasks.scala

@@ -3,15 +3,16 @@ import java.io.File
 import scala.sys.process.Process
 
 object PythonTasks {
-  val ext : String = if(System.getProperty("os.name").indexOf("Win") >= 0) "cmd" else "sh"
+  val ext : String = if (System.getProperty("os.name").indexOf("Win") >= 0) "cmd" else "sh"
+  val pythonExecutable: String = sys.env.getOrElse("PYTHON_EXECUTABLE", "python3")
 
   def workingDirectory(baseDirectory: File): File =
     new File(baseDirectory.getAbsolutePath + Seq("src", "python")
       .mkString("/", "/", ""))
 
   def testPythonTask(baseDirectory: File): Unit = {
     val cwd = workingDirectory(baseDirectory)
-    val exitCode = Process(cwd.getAbsolutePath + "/run-tests." + ext, cwd).!
+    val exitCode = Process(Seq(cwd.getAbsolutePath + "/run-tests." + ext, pythonExecutable), cwd).!
     if(exitCode != 0) {
       sys.error(s"Running python tests received non-zero exit code $exitCode")
     }
@@ -20,15 +21,15 @@ object PythonTasks {
   def buildPythonTask(baseDirectory: File, version: String): Unit = {
     val cwd = workingDirectory(baseDirectory)
     val exitCode = Process(Seq(cwd.getAbsolutePath + "/build." + ext,
-      version, "setup.py"), cwd).!
+      version, pythonExecutable, "setup.py"), cwd).!
     if(exitCode != 0) {
       sys.error(s"Building python API received non-zero exit code $exitCode")
     }
   }
 
   def buildExamplesTask(baseDirectory: File, version: String): Unit = {
     val cwd = workingDirectory(baseDirectory)
-    val exitCode = Process(Seq(cwd.getAbsolutePath + "/build." + ext, version,
+    val exitCode = Process(Seq(cwd.getAbsolutePath + "/build." + ext, version, pythonExecutable,
       "setup-examples.py"), cwd).!
     if(exitCode != 0) {
       sys.error(s"Building python examples received non-zero exit code $exitCode")