niketanpansare commented on issue #857: [SYSTEMML-2523] Update SystemML to 
Support Spark 2.3.0
URL: https://github.com/apache/systemml/pull/857#issuecomment-475413637
 
 
   Interestingly, running a similar code with `1.2.0` jars in 
`spark-2.3.0../spark-shell` succeeds, i.e. behaves similar to setup 5 rather 
than setup 6. Here is the Scala code used for testing:
   
   ```
   val ml = new org.apache.sysml.api.mlcontext.MLContext(spark)
   System.out.println(ml.version())
   val df = spark.read.parquet("shake.parquet")
   df.show()
   df.createOrReplaceTempView("df")
   ```
   
   Based on the above experiments, here are my thoughts:
   1. We can continue to support older Spark 2.1 version and can get away with 
warning on Spark 2.3 in the following setups:
   - Invoked without any Spark SQL code
   - Part of Scala/Java pipeline (for example: if invoked via spark-shell)
   - With PySpark if and only if we recommend our users to not provide any jars 
in the `driver-class-path or jars` (see setup 5 and 6)
   2. If we are uncomfortable with the above restriction, we should consider 
merging this PR.
   - Though I have validated that above Python code works with Spark 2.2.3 with 
a warning, I did not run exhaustive testing to guarantee backward compatibility 
support for older Spark 2.1 and 2.2 (with the exception of warning).
      

----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
 
For queries about this service, please contact Infrastructure at:
[email protected]


With regards,
Apache Git Services

Reply via email to