Incidentally, we were discussing this yesterday. Here are some thoughts on null handling in SQL/DataFrames. Would be great to get some feedback.
1. Treat floating point NaN and null as the same "null" value. This would be consistent with most SQL databases, and Pandas. This would also require some inbound conversion. 2. Internally, when we see a NaN value, we should mark the null bit as true, and keep the NaN value. When we see a null value for a floating point field, we should mark the null bit as true, and update the field to store NaN. 3. Externally, for floating point values, return NaN when the value is null. 4. For all other types, return null for null values. 5. For UDFs, if the argument is primitive type only (i.e. does not handle null) and not a floating point field, simply evaluate the expression to null. This is consistent with most SQL UDFs and most programming languages' treatment of NaN. Any thoughts on this semantics? On Thu, Apr 2, 2015 at 5:51 AM, Dean Wampler <deanwamp...@gmail.com> wrote: > I'm afraid you're a little stuck. In Scala, the types Int, Long, Float, > Double, Byte, and Boolean look like reference types in source code, but > they are compiled to the corresponding JVM primitive types, which can't be > null. That's why you get the warning about ==. > > It might be your best choice is to use NaN as the placeholder for null, > then create one DF using a filter that removes those values. Use that DF to > compute the mean. Then apply a map step to the original DF to translate the > NaN's to the mean. > > dean > > Dean Wampler, Ph.D. > Author: Programming Scala, 2nd Edition > <http://shop.oreilly.com/product/0636920033073.do> (O'Reilly) > Typesafe <http://typesafe.com> > @deanwampler <http://twitter.com/deanwampler> > http://polyglotprogramming.com > > On Thu, Apr 2, 2015 at 7:54 AM, Peter Rudenko <petro.rude...@gmail.com> > wrote: > > > Hi i need to implement MeanImputor - impute missing values with mean. If > i > > set missing values to null - then dataframe aggregation works properly, > but > > in UDF it treats null values to 0.0. Here’s example: > > > > |val df = sc.parallelize(Array(1.0,2.0, null, 3.0, 5.0, null)).toDF > > df.agg(avg("_1")).first //res45: org.apache.spark.sql.Row = [2.75] > > df.withColumn("d2", callUDF({(value: Double) => value}, DoubleType, > > df("d"))),show() d d2 1.0 1.0 2.0 2.0 null 0.0 3.0 3.0 5.0 5.0 null 0.0 > val > > df = sc.parallelize(Array(1.0,2.0, Double.NaN, 3.0, 5.0, > Double.NaN)).toDF > > df.agg(avg("_1")).first //res46: org.apache.spark.sql.Row = [Double.NaN] > | > > > > In UDF i cannot compare scala’s Double to null: > > > > |comparing values of types Double and Null using `==' will always yield > > false [warn] if (value==null) meanValue else value | > > > > With Double.NaN instead of null i can compare in UDF, but aggregation > > doesn’t work properly. Maybe it’s related to : > https://issues.apache.org/ > > jira/browse/SPARK-6573 > > > > Thanks, > > Peter Rudenko > > > > > > >
