Quadratic Programming with Oracle R Enterprise
- by Jeff Taylor-Oracle
I wanted to use quadprog with ORE on a server running Oracle Solaris
11.2 on a Oracle SPARC T-4 server
For background, see:
Oracle SPARC T4-2
http://docs.oracle.com/cd/E23075_01/
Oracle Solaris 11.2
http://www.oracle.com/technetwork/server-storage/solaris11/overview/index.html
quadprog: Functions to solve Quadratic Programming Problems
http://cran.r-project.org/web/packages/quadprog/index.html
Oracle R Enterprise 1.4 ("ORE") 1.4
http://www.oracle.com/technetwork/database/options/advanced-analytics/r-enterprise/ore-downloads-1502823.html
Problem: path to Solaris Studio doesn't match my installation:
$ ORE CMD INSTALL quadprog_1.5-5.tar.gz
* installing to library
\u2018/u01/app/oracle/product/12.1.0/dbhome_1/R/library\u2019
* installing *source* package \u2018quadprog\u2019 ...
** package \u2018quadprog\u2019 successfully unpacked and MD5
sums checked
** libs
/opt/SunProd/studio12u3/solarisstudio12.3/bin/f95 -m64 -PIC
-g -c aind.f -o aind.o
bash: /opt/SunProd/studio12u3/solarisstudio12.3/bin/f95:
No such file or directory
*** Error code 1
make: Fatal error: Command failed for target `aind.o'
ERROR: compilation failed for package \u2018quadprog\u2019
* removing
\u2018/u01/app/oracle/product/12.1.0/dbhome_1/R/library/quadprog\u2019
$ ls -l /opt/solarisstudio12.3/bin/f95
lrwxrwxrwx 1 root root 15 Aug 19 17:36
/opt/solarisstudio12.3/bin/f95 -> ../prod/bin/f90 Solution: a symbolic link:
$ sudo mkdir -p /opt/SunProd/studio12u3
$ sudo ln -s /opt/solarisstudio12.3 /opt/SunProd/studio12u3/ Now, it is all good:
$ ORE CMD INSTALL quadprog_1.5-5.tar.gz
* installing to library
\u2018/u01/app/oracle/product/12.1.0/dbhome_1/R/library\u2019
* installing *source* package \u2018quadprog\u2019 ...
** package \u2018quadprog\u2019 successfully unpacked and MD5
sums checked
** libs
/opt/SunProd/studio12u3/solarisstudio12.3/bin/f95 -m64 -PIC
-g -c aind.f -o aind.o
/opt/SunProd/studio12u3/solarisstudio12.3/bin/ cc -xc99 -m64
-I/usr/lib/64/R/include -DNDEBUG -KPIC -xlibmieee -c init.c -o
init.o
/opt/SunProd/studio12u3/solarisstudio12.3/bin/f95 -m64 -PIC -g
-c -o solve.QP.compact.o solve.QP.compact.f
/opt/SunProd/studio12u3/solarisstudio12.3/bin/f95 -m64 -PIC -g
-c -o solve.QP.o solve.QP.f
/opt/SunProd/studio12u3/solarisstudio12.3/bin/f95 -m64 -PIC
-g -c util.f -o util.o
/opt/SunProd/studio12u3/solarisstudio12.3/bin/ cc -xc99 -m64 -G
-o quadprog.so aind.o init.o solve.QP.compact.o solve.QP.o
util.o -xlic_lib=sunperf -lsunmath -lifai -lsunimath -lfai
-lfai2 -lfsumai -lfprodai -lfminlai -lfmaxlai -lfminvai
-lfmaxvai -lfui -lfsu -lsunmath -lmtsk -lm -lifai -lsunimath
-lfai -lfai2 -lfsumai -lfprodai -lfminlai -lfmaxlai -lfminvai
-lfmaxvai -lfui -lfsu -lsunmath -lmtsk -lm -L/usr/lib/64/R/lib
-lR
installing to
/u01/app/oracle/product/12.1.0/dbhome_1/R/library/quadprog/libs
** R
** preparing package for lazy loading
** help
*** installing help indices
converting help for package \u2018quadprog\u2019
finding HTML links ... done
solve.QP html
solve.QP.compact html
** building package indices
** testing if installed package can be loaded
* DONE (quadprog) ======
Here is an example from http://cran.r-project.org/web/packages/quadprog/quadprog.pdf > require(quadprog)
> Dmat <- matrix(0,3,3)
> diag(Dmat) <- 1
> dvec <- c(0,5,0)
> Amat <- matrix(c(-4,-3,0,2,1,0,0,-2,1),3,3)
> bvec <- c(-8,2,0)
> solve.QP(Dmat,dvec,Amat,bvec=bvec)
$solution
[1] 0.4761905 1.0476190 2.0952381
$value
[1] -2.380952
$unconstrained.solution
[1] 0 5 0
$iterations
[1] 3 0
$Lagrangian
[1] 0.0000000 0.2380952 2.0952381
$iact
[1] 3 2 Here, the standard example is modified to work with
Oracle R Enterprise
require(ORE)
ore.connect("my-name", "my-sid", "my-host", "my-pass", 1521) ore.doEval(
function () { require(quadprog)
}
) ore.doEval(
function () {
Dmat <- matrix(0,3,3) diag(Dmat)
<- 1 dvec
<- c(0,5,0) Amat
<- matrix(c(-4,-3,0,2,1,0,0,-2,1),3,3) bvec
<- c(-8,2,0) solve.QP(Dmat,dvec,Amat,bvec=bvec)
}
)
$solution
[1] 0.4761905 1.0476190 2.0952381
$value
[1] -2.380952
$unconstrained.solution
[1] 0 5 0
$iterations
[1] 3 0
$Lagrangian
[1] 0.0000000 0.2380952 2.0952381
$iact
[1] 3 2 Now I can combine the quadprog compute algorithms with
the Oracle R Enterprise Database engine functionality:
Scale to large datasets
Access to tables, views, and external tables in the database,
as well as those accessible through database links
Use SQL query parallel execution
Use in-database statistical and data mining functionality
