I am trying to implement a simple UDF but that doesn't seem to work for some
reason. It looks like Hive is not able to cast the arguments to the right
type.
select price, mavg(0, price, 2) from prices limit 1;
FAILED: Error in semantic analysis: line 1:14 Wrong Arguments 2: No matching
method for class com.priceyeti.hive.udf.MovingAvg with (int, float, int).
Possible choices: _FUNC_(org.apache.hadoop.io.Writable,
org.apache.hadoop.io.DoubleWritable, int)
Interestingly, the last argument is recognized as an int. The first one is
allowed to be of any type but it looks like I have to define a method for
each supported type. Is that true?
Lastly, the middle argument is not recognized as a DoubleWritable.
Is that because I should be using
org.apache.hadoop.hive.serde2.io.DoubleWritable instead of
org.apache.hadoop.io.DoubleWritable That's very confusing but it seems to be
the case:
http://svn.apache.org/viewvc/hive/trunk/ql/src/java/org/apache/hadoop/hive/ql/udf/UDFCeil.java?revision=1027721&view=markup
The arg types have to be from hive.serde2 but he return type has to be from
hadoop. Is that right?
It's even more confusing that there isn't a serde2 type for Int or Text:
http://svn.apache.org/viewvc/hive/trunk/serde/src/java/org/apache/hadoop/hive/serde2/io/
What's going on here?
Thanks!
igor
