[SPARK-6953] [PySpark] speed up python tests

This PR try to speed up some python tests:

```
tests.py                       144s -> 103s      -41s
mllib/classification.py         24s -> 17s        -7s
mllib/regression.py             27s -> 15s       -12s
mllib/tree.py                   27s -> 13s       -14s
mllib/tests.py                  64s -> 31s       -33s
streaming/tests.py             185s -> 84s      -101s
```
Considering python3, the total saving will be 558s (almost 10 minutes) (core, and streaming run three times, mllib runs twice).

During testing, it will show used time for each test file:
```
Run core tests ...
Running test: pyspark/rdd.py ... ok (22s)
Running test: pyspark/context.py ... ok (16s)
Running test: pyspark/conf.py ... ok (4s)
Running test: pyspark/broadcast.py ... ok (4s)
Running test: pyspark/accumulators.py ... ok (4s)
Running test: pyspark/serializers.py ... ok (6s)
Running test: pyspark/profiler.py ... ok (5s)
Running test: pyspark/shuffle.py ... ok (1s)
Running test: pyspark/tests.py ... ok (103s)   144s
```

Author: Reynold Xin <rxin@databricks.com>
Author: Xiangrui Meng <meng@databricks.com>

Closes #5605 from rxin/python-tests-speed and squashes the following commits:

d08542d [Reynold Xin] Merge pull request #14 from mengxr/SPARK-6953
89321ee [Xiangrui Meng] fix seed in tests
3ad2387 [Reynold Xin] Merge pull request #5427 from davies/python_tests

Author: rxin

python/pyspark/mllib/classification.py

@@ -86,7 +86,7 @@ class LogisticRegressionModel(LinearClassificationModel):
     ...     LabeledPoint(0.0, [0.0, 1.0]),
     ...     LabeledPoint(1.0, [1.0, 0.0]),
     ... ]
-    >>> lrm = LogisticRegressionWithSGD.train(sc.parallelize(data))
+    >>> lrm = LogisticRegressionWithSGD.train(sc.parallelize(data), iterations=10)
     >>> lrm.predict([1.0, 0.0])
     1
     >>> lrm.predict([0.0, 1.0])
@@ -95,15 +95,15 @@ class LogisticRegressionModel(LinearClassificationModel):
     [1, 0]
     >>> lrm.clearThreshold()
     >>> lrm.predict([0.0, 1.0])
-    0.123...
+    0.279...
 
     >>> sparse_data = [
     ...     LabeledPoint(0.0, SparseVector(2, {0: 0.0})),
     ...     LabeledPoint(1.0, SparseVector(2, {1: 1.0})),
     ...     LabeledPoint(0.0, SparseVector(2, {0: 1.0})),
     ...     LabeledPoint(1.0, SparseVector(2, {1: 2.0}))
     ... ]
-    >>> lrm = LogisticRegressionWithSGD.train(sc.parallelize(sparse_data))
+    >>> lrm = LogisticRegressionWithSGD.train(sc.parallelize(sparse_data), iterations=10)
     >>> lrm.predict(array([0.0, 1.0]))
     1
     >>> lrm.predict(array([1.0, 0.0]))
@@ -129,7 +129,8 @@ class LogisticRegressionModel(LinearClassificationModel):
     ...     LabeledPoint(1.0, [1.0, 0.0, 0.0]),
     ...     LabeledPoint(2.0, [0.0, 0.0, 1.0])
     ... ]
-    >>> mcm = LogisticRegressionWithLBFGS.train(data=sc.parallelize(multi_class_data), numClasses=3)
+    >>> data = sc.parallelize(multi_class_data)
+    >>> mcm = LogisticRegressionWithLBFGS.train(data, iterations=10, numClasses=3)
     >>> mcm.predict([0.0, 0.5, 0.0])
     0
     >>> mcm.predict([0.8, 0.0, 0.0])
@@ -298,7 +299,7 @@ def train(cls, data, iterations=100, initialWeights=None, regParam=0.01, regType
         ...     LabeledPoint(0.0, [0.0, 1.0]),
         ...     LabeledPoint(1.0, [1.0, 0.0]),
         ... ]
-        >>> lrm = LogisticRegressionWithLBFGS.train(sc.parallelize(data))
+        >>> lrm = LogisticRegressionWithLBFGS.train(sc.parallelize(data), iterations=10)
         >>> lrm.predict([1.0, 0.0])
         1
         >>> lrm.predict([0.0, 1.0])
@@ -330,22 +331,22 @@ class SVMModel(LinearClassificationModel):
     ...     LabeledPoint(1.0, [2.0]),
     ...     LabeledPoint(1.0, [3.0])
     ... ]
-    >>> svm = SVMWithSGD.train(sc.parallelize(data))
+    >>> svm = SVMWithSGD.train(sc.parallelize(data), iterations=10)
     >>> svm.predict([1.0])
     1
     >>> svm.predict(sc.parallelize([[1.0]])).collect()
     [1]
     >>> svm.clearThreshold()
     >>> svm.predict(array([1.0]))
-    1.25...
+    1.44...
 
     >>> sparse_data = [
     ...     LabeledPoint(0.0, SparseVector(2, {0: -1.0})),
     ...     LabeledPoint(1.0, SparseVector(2, {1: 1.0})),
     ...     LabeledPoint(0.0, SparseVector(2, {0: 0.0})),
     ...     LabeledPoint(1.0, SparseVector(2, {1: 2.0}))
     ... ]
-    >>> svm = SVMWithSGD.train(sc.parallelize(sparse_data))
+    >>> svm = SVMWithSGD.train(sc.parallelize(sparse_data), iterations=10)
     >>> svm.predict(SparseVector(2, {1: 1.0}))
     1
     >>> svm.predict(SparseVector(2, {0: -1.0}))

python/pyspark/mllib/regression.py

@@ -108,7 +108,8 @@ class LinearRegressionModel(LinearRegressionModelBase):
     ...     LabeledPoint(3.0, [2.0]),
     ...     LabeledPoint(2.0, [3.0])
     ... ]
-    >>> lrm = LinearRegressionWithSGD.train(sc.parallelize(data), initialWeights=np.array([1.0]))
+    >>> lrm = LinearRegressionWithSGD.train(sc.parallelize(data), iterations=10,
+    ...     initialWeights=np.array([1.0]))
     >>> abs(lrm.predict(np.array([0.0])) - 0) < 0.5
     True
     >>> abs(lrm.predict(np.array([1.0])) - 1) < 0.5
@@ -135,12 +136,13 @@ class LinearRegressionModel(LinearRegressionModelBase):
     ...     LabeledPoint(3.0, SparseVector(1, {0: 2.0})),
     ...     LabeledPoint(2.0, SparseVector(1, {0: 3.0}))
     ... ]
-    >>> lrm = LinearRegressionWithSGD.train(sc.parallelize(data), initialWeights=array([1.0]))
+    >>> lrm = LinearRegressionWithSGD.train(sc.parallelize(data), iterations=10,
+    ...     initialWeights=array([1.0]))
     >>> abs(lrm.predict(array([0.0])) - 0) < 0.5
     True
     >>> abs(lrm.predict(SparseVector(1, {0: 1.0})) - 1) < 0.5
     True
-    >>> lrm = LinearRegressionWithSGD.train(sc.parallelize(data), iterations=100, step=1.0,
+    >>> lrm = LinearRegressionWithSGD.train(sc.parallelize(data), iterations=10, step=1.0,
     ...    miniBatchFraction=1.0, initialWeights=array([1.0]), regParam=0.1, regType="l2",
     ...    intercept=True, validateData=True)
     >>> abs(lrm.predict(array([0.0])) - 0) < 0.5
@@ -238,7 +240,7 @@ class LassoModel(LinearRegressionModelBase):
     ...     LabeledPoint(3.0, [2.0]),
     ...     LabeledPoint(2.0, [3.0])
     ... ]
-    >>> lrm = LassoWithSGD.train(sc.parallelize(data), initialWeights=array([1.0]))
+    >>> lrm = LassoWithSGD.train(sc.parallelize(data), iterations=10, initialWeights=array([1.0]))
     >>> abs(lrm.predict(np.array([0.0])) - 0) < 0.5
     True
     >>> abs(lrm.predict(np.array([1.0])) - 1) < 0.5
@@ -265,12 +267,13 @@ class LassoModel(LinearRegressionModelBase):
     ...     LabeledPoint(3.0, SparseVector(1, {0: 2.0})),
     ...     LabeledPoint(2.0, SparseVector(1, {0: 3.0}))
     ... ]
-    >>> lrm = LinearRegressionWithSGD.train(sc.parallelize(data), initialWeights=array([1.0]))
+    >>> lrm = LinearRegressionWithSGD.train(sc.parallelize(data), iterations=10,
+    ...     initialWeights=array([1.0]))
     >>> abs(lrm.predict(np.array([0.0])) - 0) < 0.5
     True
     >>> abs(lrm.predict(SparseVector(1, {0: 1.0})) - 1) < 0.5
     True
-    >>> lrm = LassoWithSGD.train(sc.parallelize(data), iterations=100, step=1.0,
+    >>> lrm = LassoWithSGD.train(sc.parallelize(data), iterations=10, step=1.0,
     ...     regParam=0.01, miniBatchFraction=1.0, initialWeights=array([1.0]), intercept=True,
     ...     validateData=True)
     >>> abs(lrm.predict(np.array([0.0])) - 0) < 0.5
@@ -321,7 +324,8 @@ class RidgeRegressionModel(LinearRegressionModelBase):
     ...     LabeledPoint(3.0, [2.0]),
     ...     LabeledPoint(2.0, [3.0])
     ... ]
-    >>> lrm = RidgeRegressionWithSGD.train(sc.parallelize(data), initialWeights=array([1.0]))
+    >>> lrm = RidgeRegressionWithSGD.train(sc.parallelize(data), iterations=10,
+    ...     initialWeights=array([1.0]))
     >>> abs(lrm.predict(np.array([0.0])) - 0) < 0.5
     True
     >>> abs(lrm.predict(np.array([1.0])) - 1) < 0.5
@@ -348,12 +352,13 @@ class RidgeRegressionModel(LinearRegressionModelBase):
     ...     LabeledPoint(3.0, SparseVector(1, {0: 2.0})),
     ...     LabeledPoint(2.0, SparseVector(1, {0: 3.0}))
     ... ]
-    >>> lrm = LinearRegressionWithSGD.train(sc.parallelize(data), initialWeights=array([1.0]))
+    >>> lrm = LinearRegressionWithSGD.train(sc.parallelize(data), iterations=10,
+    ...     initialWeights=array([1.0]))
     >>> abs(lrm.predict(np.array([0.0])) - 0) < 0.5
     True
     >>> abs(lrm.predict(SparseVector(1, {0: 1.0})) - 1) < 0.5
     True
-    >>> lrm = RidgeRegressionWithSGD.train(sc.parallelize(data), iterations=100, step=1.0,
+    >>> lrm = RidgeRegressionWithSGD.train(sc.parallelize(data), iterations=10, step=1.0,
     ...     regParam=0.01, miniBatchFraction=1.0, initialWeights=array([1.0]), intercept=True,
     ...     validateData=True)
     >>> abs(lrm.predict(np.array([0.0])) - 0) < 0.5
@@ -396,7 +401,7 @@ def _test():
     from pyspark import SparkContext
     import pyspark.mllib.regression
     globs = pyspark.mllib.regression.__dict__.copy()
-    globs['sc'] = SparkContext('local[4]', 'PythonTest', batchSize=2)
+    globs['sc'] = SparkContext('local[2]', 'PythonTest', batchSize=2)
     (failure_count, test_count) = doctest.testmod(globs=globs, optionflags=doctest.ELLIPSIS)
     globs['sc'].stop()
     if failure_count: