Thanks you, I didn't know about it.
Guillaume Polaert | Cyrès Conseil
De : Gabi D [mailto:gabi...@gmail.com]
Envoyé : lundi 11 juin 2012 12:14
À : user@hive.apache.org
Cc : Matouk Iftissen
Objet : Re: Trouble with sum function
float is known to have precision issues, because of the way
>From the code here
http://svn.apache.org/viewvc/hive/branches/branch-0.7/ql/src/java/org/apache/hadoop/hive/ql/udf/generic/GenericUDAFSum.java?view=markup
For float , doble and string the implementation points to common function
GenericUDAFSumDouble()
if (parameters[0].getCategory() != ObjectIn
float is known to have precision issues, because of the way it is
implemented. If you are working with money data you should definitely move
to double.
google 'float precision' and you'll find a bunch of explanations.
On Mon, Jun 11, 2012 at 12:49 PM, Guillaume Polaert wrote:
> Hi,
>
> We're expe
Hi,
We're expecting some issue with the sum function in Hive 0.7.1.
The precision of float number isn't correct (0.320484484676 instead of 0.32)
We aren't expecting this error with double format.
For instance, "select id, sum(col1), sum(col2) from test_table group by id"
returns incorrect values
