it might be the case that there are few empty spaces at the end of each row which are being handled when you are reading and writing from disc
but when you set autoconvert then looks like one of these tables is really small and it is converted into mapside join which means the entire table is loaded into map memory and there is no need of reduce On Wed, Aug 15, 2012 at 9:13 PM, Himanish Kushary <himan...@gmail.com> wrote: > Hi, > > I have uploaded few csv files from windows into hive and configured few > external tables using them. When I am trying to run a join on two tables one > of the int columns > get changed to 0. The structure of the tables are as follows: > > > Table-1 Table-2 > ------------ ----------- > > Id(int) id(int) datetime > eid(int) > -- ---- ------------ > ----- > 1 1 2011-02-01 3 > 2 1 2011-03-01 4 > 3 2 2011-04-01 5 > 4 2011-05-01 6 > 6 2011-06-01 7 > > > The join query is - select a.* from Table-2 a join Table-1 b on (a.id=b.id); > > The output is: > > 1 2011-02-01 0 > 1 2011-03-01 0 > 2 2011-04-01 0 > > > I checked the logs and noticed the following warning : WARN > org.apache.hadoop.hive.serde2.lazybinary.LazyBinaryStruct: Extra bytes > detected at the end of the row! Ignoring similar problems.Could this be > causing it ? > > When I turn on hive.auto.convert.join=true , the error goes away as there is > no reduce phase.The output is: > > 1 2011-02-01 3 > 1 2011-03-01 4 > 2 2011-04-01 5 > > Could somebody please help me figure out why we get the wrong results when > running through the reducer. > -- > Thanks -- Nitin Pawar
