How to pass an object to RUDF in hive

[rajasekhariitbbs]

Working Function

Minimum = function(column1,column2){
min(column1,column2)
}

Not Working

a=-10000
Minimum = function(column1,column2){
min(column1,column2,a)
}

How to pass an object/dataframe/function into the RHive-UDF

Thanks
Raja Sekhar

[ssshow16]

If you just pass value into the RHive-UDF, you can do like the following:

Minimum = function(column1,column2_, a_){
min(column1,column2,a)
}

rhive.query("select R('Minimum',col1,col2,-10000,0.0) from table_name")

However, RHive just export the UDF function and RHive-UDF is executed at
each DataNode, so this cannot reference your function/object/dataframe.
If you need to use other function, you have to define a inner function like
the following.

Minimum = function(column1,column2_, a_){
min(column1,column2,a)
sub_func <- function(a,b){
....
}
sub_func(column1, column2)
}

Thanks.

On Tue, Nov 4, 2014 at 3:10 PM, rajasekhariitbbs [email protected]
wrote:

Working Function

Minimum = function(column1,column2){
min(column1,column2)
}
Not Working

a=-10000
Minimum = function(column1,column2){
min(column1,column2,a)
}

How to pass an object/dataframe/function into the RHive-UDF

Thanks
Raja Sekhar

—
Reply to this email directly or view it on GitHub
#73.

[rajasekhariitbbs]

These are the following rhive.evn()
rhive.env()
hadoop home: /home/training/hadoop-2.4.0/
hadoop conf: /home/training/hadoop-2.4.0/etc/hadoop
fs: hdfs://localhost:9000
hive home: /home/training/hive/
user name: training
user home: /home/training
temp dir: /tmp/training

select queries are working good

######### User Define Functions ########
coefficient <- 1.1
scoring <- function(sal) {
coefficient * sal
}
rhive.assign('coefficient',coefficient)
rhive.assign('scoring',scoring)
rhive.export('scoring')
rhive.export('coefficient')

The above rhive.export is saving the files in filesystem /rhive/udf/training, Is that correct?

[ssshow16]

You cannot assign and export a variable 'coefficient'.

Exported file will be saved in HDFS : /rhive/udf/{user}.

On Tue, Nov 4, 2014 at 5:51 PM, rajasekhariitbbs [email protected]
wrote:

These are the following rhive.evn()
rhive.env()
hadoop home: /home/training/hadoop-2.4.0/
hadoop conf: /home/training/hadoop-2.4.0/etc/hadoop
fs: hdfs://localhost:9000
hive home: /home/training/hive/
user name: training
user home: /home/training
temp dir: /tmp/training

select queries are working good

######### User Define Functions ########
coefficient <- 1.1
scoring <- function(sal) {
coefficient * sal
}
rhive.assign('coefficient',coefficient)
rhive.assign('scoring',scoring)
rhive.export('scoring')
rhive.export('coefficient')

The above rhive.export is saving the files in filesystem
/rhive/udf/training, Is that correct?

—
Reply to this email directly or view it on GitHub
#73 (comment).

[rajasekhariitbbs]

In my case the file is being saved in local file system, I'm confused and don't know why it is happening,
For now if I manually copy the .RDA file from local system to HDFS then the code is working.

In the below URL they are exporting the coefficient also
https://github.com/nexr/RHive/wiki/RHive-example-code

[ssshow16]

If RHive-UDF reference R Object( *.RData) as first param of R() in Query.

rhive.query("select R('scoring',col_sal,0.0) from emp")

In your case, two R Object is saved in HDFS because you use rhive.export()
function for each R Object.

• coefficient.RData
• scoring.RData

So, scoring function in scoring.RData cannot reference coefficient value
in coefficient.RData.
In this case, you have to call rhive.exportAll(‘scoring’).
rhive.exportAll() function save all R Object into scoring.RData.

Please, try again.

On Tue, Nov 4, 2014 at 6:03 PM, rajasekhariitbbs [email protected]
wrote:

In my case the file is being saved in local file system, I'm confused and
don't know why it is happening,
For now if I manually copy the .RDA file from local system to HDFS then
the code is working.

In the below URL they are exporting the coefficient also
https://github.com/nexr/RHive/wiki/RHive-example-code

—
Reply to this email directly or view it on GitHub
#73 (comment).