如何在Pyspark中启用Apache Arrow

Question

我正在尝试使Apache能够转换为Pandas。我正在使用：

熊猫0.25.1矮小1.17.2

这是示例代码

spark.conf.set("spark.sql.execution.arrow.enabled", "true")
x = pd.Series([1, 2, 3])
df = spark.createDataFrame(pd.DataFrame(x, columns=["x"]))

我收到一条警告信息

c:\users\administrator\appdata\local\programs\python\python37\lib\site-packages\pyspark\sql\session.py:714: UserWarning: createDataFrame attempted Arrow optimization because 'spark.sql.execution.arrow.enabled' is set to true; however, failed by the reason below:
  An error occurred while calling z:org.apache.spark.sql.api.python.PythonSQLUtils.readArrowStreamFromFile.
: java.lang.IllegalArgumentException
    at java.nio.ByteBuffer.allocate(ByteBuffer.java:334)
    at org.apache.arrow.vector.ipc.message.MessageSerializer.readMessage(MessageSerializer.java:543)
    at org.apache.spark.sql.execution.arrow.ArrowConverters$$anon$3.readNextBatch(ArrowConverters.scala:243)
    at org.apache.spark.sql.execution.arrow.ArrowConverters$$anon$3.<init>(ArrowConverters.scala:229)
    at org.apache.spark.sql.execution.arrow.ArrowConverters$.getBatchesFromStream(ArrowConverters.scala:228)
    at org.apache.spark.sql.execution.arrow.ArrowConverters$$anonfun$readArrowStreamFromFile$2.apply(ArrowConverters.scala:216)
    at org.apache.spark.sql.execution.arrow.ArrowConverters$$anonfun$readArrowStreamFromFile$2.apply(ArrowConverters.scala:214)
    at org.apache.spark.util.Utils$.tryWithResource(Utils.scala:2543)
    at org.apache.spark.sql.execution.arrow.ArrowConverters$.readArrowStreamFromFile(ArrowConverters.scala:214)
    at org.apache.spark.sql.api.python.PythonSQLUtils$.readArrowStreamFromFile(PythonSQLUtils.scala:46)
    at org.apache.spark.sql.api.python.PythonSQLUtils.readArrowStreamFromFile(PythonSQLUtils.scala)
    at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method)
    at sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:62)
    at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43)
    at java.lang.reflect.Method.invoke(Method.java:498)
    at py4j.reflection.MethodInvoker.invoke(MethodInvoker.java:244)
    at py4j.reflection.ReflectionEngine.invoke(ReflectionEngine.java:357)
    at py4j.Gateway.invoke(Gateway.java:282)
    at py4j.commands.AbstractCommand.invokeMethod(AbstractCommand.java:132)
    at py4j.commands.CallCommand.execute(CallCommand.java:79)
    at py4j.GatewayConnection.run(GatewayConnection.java:238)
    at java.lang.Thread.run(Thread.java:748)

Attempting non-optimization as 'spark.sql.execution.arrow.fallback.enabled' is set to true.
  warnings.warn(msg)

Answer 1

我们在0.15.0中做了一个更改，这使得py箭头的默认行为与Java中的Arrow的旧版本不兼容--您的Spark环境似乎使用的是旧版本。

你的选择是

从使用Python

• Downgrade的位置开始，
• 将环境变量ARROW_PRE_0_15_IPC_FORMAT=1设置为py箭头< 0.15.0。

Answer 2

用pyarrow==0.15在我的Spark2.4.4集群中给我的熊猫命名为UDF。正如上面提到的，我很难设置ARROW_PRE_0_15_IPC_FORMAT=1标志。

我在以下位置设置了标志：(1)在头节点上通过export设置命令行；(2)在集群中的所有节点上通过spark-env.sh和yarn-env.sh设置标志；(3)在我在head节点上的脚本中设置吡火花代码本身。由于未知的原因，所有这些都无法真正在udf中设置此标志。

我找到的最简单的解决方案是将这个在中称为udf：

    @pandas_udf("integer", PandasUDFType.SCALAR)
    def foo(*args):
        import os
        os.environ["ARROW_PRE_0_15_IPC_FORMAT"] = "1"
        #...

希望这能省下其他人几个小时。