diff options
author | Yanbo Liang <ybliang8@gmail.com> | 2016-09-29 00:54:26 -0700 |
---|---|---|
committer | Yanbo Liang <ybliang8@gmail.com> | 2016-09-29 00:54:26 -0700 |
commit | a19a1bb59411177caaf99581e89098826b7d0c7b (patch) | |
tree | 649a504d904cce2f0783def6e0114ab68a9e1024 /mllib/src | |
parent | 37eb9184f1e9f1c07142c66936671f4711ef407d (diff) | |
download | spark-a19a1bb59411177caaf99581e89098826b7d0c7b.tar.gz spark-a19a1bb59411177caaf99581e89098826b7d0c7b.tar.bz2 spark-a19a1bb59411177caaf99581e89098826b7d0c7b.zip |
[SPARK-16356][FOLLOW-UP][ML] Enforce ML test of exception for local/distributed Dataset.
## What changes were proposed in this pull request?
#14035 added ```testImplicits``` to ML unit tests and promoted ```toDF()```, but left one minor issue at ```VectorIndexerSuite```. If we create the DataFrame by ```Seq(...).toDF()```, it will throw different error/exception compared with ```sc.parallelize(Seq(...)).toDF()``` for one of the test cases.
After in-depth study, I found it was caused by different behavior of local and distributed Dataset if the UDF failed at ```assert```. If the data is local Dataset, it throws ```AssertionError``` directly; If the data is distributed Dataset, it throws ```SparkException``` which is the wrapper of ```AssertionError```. I think we should enforce this test to cover both case.
## How was this patch tested?
Unit test.
Author: Yanbo Liang <ybliang8@gmail.com>
Closes #15261 from yanboliang/spark-16356.
Diffstat (limited to 'mllib/src')
-rw-r--r-- | mllib/src/test/scala/org/apache/spark/ml/feature/VectorIndexerSuite.scala | 13 |
1 files changed, 9 insertions, 4 deletions
diff --git a/mllib/src/test/scala/org/apache/spark/ml/feature/VectorIndexerSuite.scala b/mllib/src/test/scala/org/apache/spark/ml/feature/VectorIndexerSuite.scala index 4da1b133e8..b28ce2ab45 100644 --- a/mllib/src/test/scala/org/apache/spark/ml/feature/VectorIndexerSuite.scala +++ b/mllib/src/test/scala/org/apache/spark/ml/feature/VectorIndexerSuite.scala @@ -88,9 +88,7 @@ class VectorIndexerSuite extends SparkFunSuite with MLlibTestSparkContext densePoints1 = densePoints1Seq.map(FeatureData).toDF() sparsePoints1 = sparsePoints1Seq.map(FeatureData).toDF() - // TODO: If we directly use `toDF` without parallelize, the test in - // "Throws error when given RDDs with different size vectors" is failed for an unknown reason. - densePoints2 = sc.parallelize(densePoints2Seq, 2).map(FeatureData).toDF() + densePoints2 = densePoints2Seq.map(FeatureData).toDF() sparsePoints2 = sparsePoints2Seq.map(FeatureData).toDF() badPoints = badPointsSeq.map(FeatureData).toDF() } @@ -121,10 +119,17 @@ class VectorIndexerSuite extends SparkFunSuite with MLlibTestSparkContext model.transform(densePoints1) // should work model.transform(sparsePoints1) // should work - intercept[SparkException] { + // If the data is local Dataset, it throws AssertionError directly. + intercept[AssertionError] { model.transform(densePoints2).collect() logInfo("Did not throw error when fit, transform were called on vectors of different lengths") } + // If the data is distributed Dataset, it throws SparkException + // which is the wrapper of AssertionError. + intercept[SparkException] { + model.transform(densePoints2.repartition(2)).collect() + logInfo("Did not throw error when fit, transform were called on vectors of different lengths") + } intercept[SparkException] { vectorIndexer.fit(badPoints) logInfo("Did not throw error when fitting vectors of different lengths in same RDD.") |