Hi chenchun, About the second part of your question, I agree that we should change it. I checked MySQL and it returns the same type as was sent to the round function. I created https://issues.apache.org/jira/browse/HIVE-4174 to track this. It's presently assigned to me but if you like to contribute a patch (along with updated tests), please feel free to steal it from me.
Mark On Thu, Mar 14, 2013 at 12:44 AM, chenchun <chenchun.f...@gmail.com> wrote: > Hi, > I have a question about HIVE-3675 NaN does not work correctly for round(n). > The issue did solve the NaN problem, but the hql below result in a different > display with hive0.10. > I checked the code UDFRound.java, found the function "LongWritable > evaluate(DoubleWritable n)" changed to "DoubleWritable > evaluate(DoubleWritable n) ". > I think it maybe better to add a "LongWritable evaluate(LongWritable n)" > function and modify the toString function of LongWritable that could solve > the problem. > > > select round(1234560), round(12345670) from test limit 1; > > //hive 0.10 > 1234560.0 1.234567E7 > > //hive 0.9 > 1234560 12345670 > > select round(cast(1234560 as BIGINT)), round(cast(12345670 as BIGINT)) from > test limit 1; > > //hive 0.10 > 1234560.0 1.234567E7 > > > regards, > chenchun >
