Question

这是我在 scala 中使用 toDebugString 时得到的结果：

scala> val a  = sc.parallelize(Array(1,2,3)).distinct
a: org.apache.spark.rdd.RDD[Int] = MappedRDD[3] at distinct at <console>:12

scala> a.toDebugString
res0: String = 
(4) MappedRDD[3] at distinct at <console>:12
 |  ShuffledRDD[2] at distinct at <console>:12
 +-(4) MappedRDD[1] at distinct at <console>:12
    |  ParallelCollectionRDD[0] at parallelize at <console>:12

这相当于 python ：

>>> a = sc.parallelize([1,2,3]).distinct()
>>> a.toDebugString()
'(4) PythonRDD[6] at RDD at PythonRDD.scala:43\n |  MappedRDD[5] at values at NativeMethodAccessorImpl.java:-2\n |  ShuffledRDD[4] at partitionBy at NativeMethodAccessorImpl.java:-2\n +-(4) PairwiseRDD[3] at RDD at PythonRDD.scala:261\n    |  PythonRDD[2] at RDD at PythonRDD.scala:43\n    |  ParallelCollectionRDD[0] at parallelize at PythonRDD.scala:315'

正如您所看到的，python中的输出不像scala那样好。有没有任何技巧可以更好地输出这个功能？

我正在使用Spark 1.1.0。

Answer 1

尝试添加print语句，以便实际打印调试字符串，而不是显示其__repr__：

>>> a = sc.parallelize([1,2,3]).distinct()
>>> print a.toDebugString()
(8) PythonRDD[27] at RDD at PythonRDD.scala:44 [Serialized 1x Replicated]
 |  MappedRDD[26] at values at NativeMethodAccessorImpl.java:-2 [Serialized 1x Replicated]
 |  ShuffledRDD[25] at partitionBy at NativeMethodAccessorImpl.java:-2 [Serialized 1x Replicated]
 +-(8) PairwiseRDD[24] at distinct at <stdin>:1 [Serialized 1x Replicated]
    |  PythonRDD[23] at distinct at <stdin>:1 [Serialized 1x Replicated]
    |  ParallelCollectionRDD[21] at parallelize at PythonRDD.scala:358 [Serialized 1x Replicated]

Answer 2

它没有执行，只是缓存你应该使用：

a = sc.parallelize([1,2,3]).distinct()
a.collect()
[1, 2, 3]

Spark中的toDebugString在python中并不好用

2 个答案: