[SPARK-2334] fix AttributeError when call PipelineRDD.id()

The underline JavaRDD for PipelineRDD is created lazily, it's delayed until call _jrdd.

The id of JavaRDD is cached as `_id`, it saves a RPC call in py4j for later calls.

closes #1276

Author: Davies Liu <davies.liu@gmail.com>

Closes #2296 from davies/id and squashes the following commits:

e197958 [Davies Liu] fix style
9721716 [Davies Liu] fix id of PipelineRDD

1 files changed, 6 insertions, 0 deletions

diff --git a/python/pyspark/rdd.py b/python/pyspark/rdd.py
index aa90297855..266090e3ae 100644
--- a/python/pyspark/rdd.py
+++ b/python/pyspark/rdd.py
@@ -2075,6 +2075,7 @@ class PipelinedRDD(RDD):
         self.ctx = prev.ctx
         self.prev = prev
         self._jrdd_val = None
+        self._id = None
         self._jrdd_deserializer = self.ctx.serializer
         self._bypass_serializer = False
         self._partitionFunc = prev._partitionFunc if self.preservesPartitioning else None
@@ -2105,6 +2106,11 @@ class PipelinedRDD(RDD):
         self._jrdd_val = python_rdd.asJavaRDD()
         return self._jrdd_val
 
+    def id(self):
+        if self._id is None:
+            self._id = self._jrdd.id()
+        return self._id
+
     def _is_pipelinable(self):
         return not (self.is_cached or self.is_checkpointed)