use SQLContext.getOrCreate in mllib

Author: Davies Liu

python/pyspark/mllib/common.py

@@ -102,7 +102,7 @@ def _java2py(sc, r, encoding="bytes"):
             return RDD(jrdd, sc)
 
         if clsName == 'DataFrame':
-            return DataFrame(r, SQLContext(sc))
+            return DataFrame(r, SQLContext.getOrCreate(sc))
 
         if clsName in _picklable_classes:
             r = sc._jvm.SerDe.dumps(r)
@@ -125,7 +125,7 @@ def callJavaFunc(sc, func, *args):
 
 def callMLlibFunc(name, *args):
     """ Call API in PythonMLLibAPI """
-    sc = SparkContext._active_spark_context
+    sc = SparkContext.getOrCreate()
     api = getattr(sc._jvm.PythonMLLibAPI(), name)
     return callJavaFunc(sc, api, *args)
 
@@ -135,7 +135,7 @@ class JavaModelWrapper(object):
     Wrapper for the model in JVM
     """
     def __init__(self, java_model):
-        self._sc = SparkContext._active_spark_context
+        self._sc = SparkContext.getOrCreate()
         self._java_model = java_model
 
     def __del__(self):

python/pyspark/mllib/evaluation.py

@@ -44,7 +44,7 @@ class BinaryClassificationMetrics(JavaModelWrapper):
 
     def __init__(self, scoreAndLabels):
         sc = scoreAndLabels.ctx
-        sql_ctx = SQLContext(sc)
+        sql_ctx = SQLContext.getOrCreate(sc)
         df = sql_ctx.createDataFrame(scoreAndLabels, schema=StructType([
             StructField("score", DoubleType(), nullable=False),
             StructField("label", DoubleType(), nullable=False)]))
@@ -103,7 +103,7 @@ class RegressionMetrics(JavaModelWrapper):
 
     def __init__(self, predictionAndObservations):
         sc = predictionAndObservations.ctx
-        sql_ctx = SQLContext(sc)
+        sql_ctx = SQLContext.getOrCreate(sc)
         df = sql_ctx.createDataFrame(predictionAndObservations, schema=StructType([
             StructField("prediction", DoubleType(), nullable=False),
             StructField("observation", DoubleType(), nullable=False)]))
@@ -197,7 +197,7 @@ class MulticlassMetrics(JavaModelWrapper):
 
     def __init__(self, predictionAndLabels):
         sc = predictionAndLabels.ctx
-        sql_ctx = SQLContext(sc)
+        sql_ctx = SQLContext.getOrCreate(sc)
         df = sql_ctx.createDataFrame(predictionAndLabels, schema=StructType([
             StructField("prediction", DoubleType(), nullable=False),
             StructField("label", DoubleType(), nullable=False)]))
@@ -338,7 +338,7 @@ class RankingMetrics(JavaModelWrapper):
 
     def __init__(self, predictionAndLabels):
         sc = predictionAndLabels.ctx
-        sql_ctx = SQLContext(sc)
+        sql_ctx = SQLContext.getOrCreate(sc)
         df = sql_ctx.createDataFrame(predictionAndLabels,
                                      schema=sql_ctx._inferSchema(predictionAndLabels))
         java_model = callMLlibFunc("newRankingMetrics", df._jdf)
@@ -424,7 +424,7 @@ class MultilabelMetrics(JavaModelWrapper):
 
     def __init__(self, predictionAndLabels):
         sc = predictionAndLabels.ctx
-        sql_ctx = SQLContext(sc)
+        sql_ctx = SQLContext.getOrCreate(sc)
         df = sql_ctx.createDataFrame(predictionAndLabels,
                                      schema=sql_ctx._inferSchema(predictionAndLabels))
         java_class = sc._jvm.org.apache.spark.mllib.evaluation.MultilabelMetrics

python/pyspark/mllib/feature.py

@@ -100,8 +100,7 @@ def transform(self, vector):
         :return: normalized vector. If the norm of the input is zero, it
                  will return the input vector.
         """
-        sc = SparkContext._active_spark_context
-        assert sc is not None, "SparkContext should be initialized first"
+        sc = SparkContext.getOrCreate()
         if isinstance(vector, RDD):
             vector = vector.map(_convert_to_vector)
         else: