Hi All,
I'm still having significantly difficulty getting the quantreg library
running in R. I'm running R on MEPIS using the debs created by Dirk
Eddelbuettel and placed in apt testing. When I try to install quantreg using
the install.packages() function it fails with: /usr/bin/ld: cannot find
-lblas-3
Dirk was nice enough to send me a .deb for quantreg which installs without
error. However, when I try to run the rq command after entering
library(quantreg), I get: Error: Object "rq" not found
I've tried reinstalling atlas3-base (as Douglas Bates suggested), every R
package on my machine and ripping out every config file I can find and
re-installing to no avail. I am a true beginner at R and could certainly
be doing something or just be missing a package. If anyone has anymore
ideas, I'd love to hear them. Thanks for the help.
-Mark
engle at unr dot nevada dot edu
Having problems with quantreg
4 messages · mark engle, Dirk Eddelbuettel
On 21 February 2005 at 22:42, mark engle wrote:
| Hi All,
| I'm still having significantly difficulty getting the quantreg library
| running in R. I'm running R on MEPIS using the debs created by Dirk
| Eddelbuettel and placed in apt testing. When I try to install quantreg using
| the install.packages() function it fails with: /usr/bin/ld: cannot find
| -lblas-3
|
| Dirk was nice enough to send me a .deb for quantreg which installs without
| error. However, when I try to run the rq command after entering
| library(quantreg), I get: Error: Object "rq" not found
|
| I've tried reinstalling atlas3-base (as Douglas Bates suggested), every R
| package on my machine and ripping out every config file I can find and
| re-installing to no avail. I am a true beginner at R and could certainly
| be doing something or just be missing a package. If anyone has anymore
| ideas, I'd love to hear them. Thanks for the help.
i) I know little about MEPIS, and in particular R on MEPIS. Take everything
with a grain of salt, and if in doubt, test on a proper Debian machine.
Try to find one somewhere at nevada.edu. It will easier for us to get it
to work there -- and let's then try to apply what we learned on your
MEPIS box.
ii) The (autogenerated) quantreg package I sent you has Depends as follows:
edd@chibud:~> dpkg -f src/progs/perl/cran2deb/builds/r-cran-quantreg_3.50-0.r2d.1_i386.deb
Package: r-cran-quantreg
Version: 3.50-0.r2d.1
Section: math
Priority: optional
Architecture: i386
Depends: atlas3-base | refblas3 | libblas.so.3, libc6 (>= 2.3.2.ds1-4), libg2c0(>= 1:3.3.4-1), libgcc1 (>= 1:3.4.1-3), r-base-core
Installed-Size: 848
Maintainer: Dirk Eddelbuettel <edd@debian.org>
Source: quantreg
Description: GNU R package "Quantile Regression"
Quantile regression and related methods.
.
Author(s): Roger Koenker <rkoenker@uiuc.edu>
Initial R port from Splus by Brian Ripley <ripley@stats.ox.ac.uk>
Date:
.
URL: http://www.econ.uiuc.edu/~roger/research/rq/rq.html
Notice how it says "atlas3-base | refblas3 | libblas.so.3" What you
need is a concrete package that provides the virtual package
libblas.so.3.
You tried atlas3-base, and it didn't work. The next thing to try would
be refblas3. Install that and see what it does.
Afterwards there are the others:
edd@chibud:~> apt-cache showpkg libblas.so.3
Package: libblas.so.3
Versions:
Reverse Depends:
r-cran-matrix,libblas.so.3
r-cran-rscalapack,libblas.so.3
[...]
Dependencies:
Provides:
Reverse Provides:
refblas3 1.2-6
atlas3-sse2 3.6.0-19
atlas3-sse 3.6.0-19
atlas3-base 3.6.0-19
atlas3-3dnow 3.6.0-19
edd@chibud:~>
Anyone of those after the "Reverse Provides" will do, but the
hardware-optimised ones (atlas3-sse2, -sse2, -3dnow) only makes
sense on the matching hardware.
Hope this helps, Dirk
| -Mark
|
| engle at unr dot nevada dot edu
|
| _______________________________________________
| R-SIG-Debian mailing list
| R-SIG-Debian@r-project.org
| https://stat.ethz.ch/mailman/listinfo/r-sig-debian
Better to have an approximate answer to the right question than a precise answer to the wrong question. -- John Tukey as quoted by John Chambers
Thanks for all of the help. I had already installed refblas3. I installed atlas3-sse, which didn't seem to change anything. I then added atlas3-base-dev (seems kind of obvious now) and the quantreg package compiled from source and commands from the package are now recognized. I'm going to give it a quick run around the block but I think that will solve everything for me. Thanks again. -Mark
On Tuesday 22 February 2005 5:16, Dirk Eddelbuettel wrote:
On 21 February 2005 at 22:42, mark engle wrote: | Hi All, | I'm still having significantly difficulty getting the quantreg | library running in R. I'm running R on MEPIS using the debs created by | Dirk Eddelbuettel and placed in apt testing. When I try to install | quantreg using the install.packages() function it fails with: | /usr/bin/ld: cannot find -lblas-3 | | Dirk was nice enough to send me a .deb for quantreg which installs | without error. However, when I try to run the rq command after entering | library(quantreg), I get: Error: Object "rq" not found | | I've tried reinstalling atlas3-base (as Douglas Bates suggested), every R | package on my machine and ripping out every config file I can find and | re-installing to no avail. I am a true beginner at R and could | certainly be doing something or just be missing a package. If anyone has | anymore ideas, I'd love to hear them. Thanks for the help. i) I know little about MEPIS, and in particular R on MEPIS. Take everything with a grain of salt, and if in doubt, test on a proper Debian machine. Try to find one somewhere at nevada.edu. It will easier for us to get it to work there -- and let's then try to apply what we learned on your MEPIS box. ii) The (autogenerated) quantreg package I sent you has Depends as follows: edd@chibud:~> dpkg -f src/progs/perl/cran2deb/builds/r-cran-quantreg_3.50-0.r2d.1_i386.deb Package: r-cran-quantreg Version: 3.50-0.r2d.1 Section: math Priority: optional Architecture: i386 Depends: atlas3-base | refblas3 | libblas.so.3, libc6 (>= 2.3.2.ds1-4), libg2c0(>= 1:3.3.4-1), libgcc1 (>= 1:3.4.1-3), r-base-core Installed-Size: 848 Maintainer: Dirk Eddelbuettel <edd@debian.org> Source: quantreg Description: GNU R package "Quantile Regression" Quantile regression and related methods. . Author(s): Roger Koenker <rkoenker@uiuc.edu> Initial R port from Splus by Brian Ripley <ripley@stats.ox.ac.uk> Date: . URL: http://www.econ.uiuc.edu/~roger/research/rq/rq.html Notice how it says "atlas3-base | refblas3 | libblas.so.3" What you need is a concrete package that provides the virtual package libblas.so.3. You tried atlas3-base, and it didn't work. The next thing to try would be refblas3. Install that and see what it does. Afterwards there are the others: edd@chibud:~> apt-cache showpkg libblas.so.3 Package: libblas.so.3 Versions: Reverse Depends: r-cran-matrix,libblas.so.3 r-cran-rscalapack,libblas.so.3 [...] Dependencies: Provides: Reverse Provides: refblas3 1.2-6 atlas3-sse2 3.6.0-19 atlas3-sse 3.6.0-19 atlas3-base 3.6.0-19 atlas3-3dnow 3.6.0-19 edd@chibud:~> Anyone of those after the "Reverse Provides" will do, but the hardware-optimised ones (atlas3-sse2, -sse2, -3dnow) only makes sense on the matching hardware. Hope this helps, Dirk | -Mark | | engle at unr dot nevada dot edu | | _______________________________________________ | R-SIG-Debian mailing list | R-SIG-Debian@r-project.org | https://stat.ethz.ch/mailman/listinfo/r-sig-debian
Mark Engle Dept. of Nat. Res. and Env. Sci. Mail Stop 370 University of Nevada Reno, NV 89557 engle@unr.nevada.edu (775)784-4966
On 22 February 2005 at 15:01, mark engle wrote:
| Thanks for all of the help. I had already installed refblas3. I installed | atlas3-sse, which didn't seem to change anything. I then added | atlas3-base-dev (seems kind of obvious now) and the quantreg package compiled | from source and commands from the package are now recognized. I'm going to | give it a quick run around the block but I think that will solve everything | for me. Thanks again. So I suspect you never had the r-base-dev package installed which would have lead you to having a -dev package providing blas-3 development headers and libs. Hm, Doug and I should have checked that first ... Anyway, glad to hear you have it working now. Regards, Dirk
Better to have an approximate answer to the right question than a precise answer to the wrong question. -- John Tukey as quoted by John Chambers