This is a guide to installation and administration for R.
This manual is for R, version 4.4.1 Patched (2024-08-05).
Copyright © 2001–2024 R Core Team
Permission is granted to make and distribute verbatim copies of this manual provided the copyright notice and this permission notice are preserved on all copies.
Permission is granted to copy and distribute modified versions of this manual under the conditions for verbatim copying, provided that the entire resulting derived work is distributed under the terms of a permission notice identical to this one.
Permission is granted to copy and distribute translations of this manual into another language, under the above conditions for modified versions, except that this permission notice may be stated in a translation approved by the R Core Team.
Sources, binaries and documentation for R can be obtained via CRAN, the “Comprehensive R Archive Network” whose current members are listed at https://CRAN.R-project.org/mirrors.html.
The simplest way is to download the most recent R-x.y.z.tar.gz file, and unpack it with
tar -xf R-x.y.z.tar.gz
on systems that have a suitable1
tar installed. On other systems you need to have the
gzip program installed, when you can use
gzip -dc R-x.y.z.tar.gz | tar -xf -
The pathname of the directory into which the sources are unpacked should
not contain spaces, as most make programs (and specifically
GNU make) do not expect spaces.
If you want the build to be usable by a group of users, set umask
before unpacking so that the files will be readable by the target group
(e.g., umask 022 to be usable by all users). Keep this
setting of umask whilst building and installing.
If you use a fairly recent GNU version of tar and do
this as a root account (which on Windows includes accounts with
administrator privileges) you may see many warnings about changing
ownership. In which case you can use
tar --no-same-owner -xf R-x.y.z.tar.gz
and perhaps also include the option --no-same-permissions.
(These options can also be set in the TAR_OPTIONS environment
variable: if more than one option is included they should be separated
by spaces.)
A patched version of the current release, ‘r-patched’, and the current development version, ‘r-devel’, are available as daily tarballs and via access to the R Subversion repository. (For the two weeks prior to the release of a minor (4.x.0) version, ‘r-patched’ tarballs may refer to beta/release candidates of the upcoming release, the patched version of the current release being available via Subversion.)
The tarballs are available from https://stat.ethz.ch/R/daily/. Download R-patched.tar.gz or R-devel.tar.gz (or the .tar.bz2 versions) and unpack as described in the previous section. They are built in exactly the same way as distributions of R releases.
Sources are also available via https://svn.R-project.org/R/, the R Subversion repository. If you have a Subversion client (see https://subversion.apache.org/), you can check out and update the current ‘r-devel’ from https://svn.r-project.org/R/trunk/ and the current ‘r-patched’ from ‘https://svn.r-project.org/R/branches/R-x-y-branch/’ (where x and y are the major and minor number of the current released version of R). E.g., use
svn checkout https://svn.r-project.org/R/trunk/ path
to check out ‘r-devel’ into directory path (which will be created if necessary). The alpha, beta and RC versions of an upcoming x.y.0 release are available from ‘https://svn.r-project.org/R/branches/R-x-y-branch/’ in the four-week period prior to the release.
Note that ‘https:’ is required2, and that the SSL certificate for the Subversion server of the R project should be recognized as from a trusted source.
Note that retrieving the sources by e.g. wget -r or
svn export from that URL will not work (and will give a error
early in the make process): the Subversion information is
needed to build R.
The Subversion repository does not contain the current sources for the
recommended packages, which can be obtained by rsync or
downloaded from CRAN. To use rsync to install the
appropriate sources for the recommended packages, run
./tools/rsync-recommended from the top-level directory of the
R sources.
If downloading manually from CRAN, do ensure that you have the
correct versions of the recommended packages: if the number in the file
VERSION is ‘x.y.z’ you need to download
the contents of ‘https://CRAN.R-project.org/src/contrib/dir’,
where dir is ‘x.y.z/Recommended’ for
r-devel or x.y-patched/Recommended for r-patched,
respectively, to directory src/library/Recommended in the sources
you have unpacked. After downloading manually you need to execute
tools/link-recommended from the top level of the sources to
make the requisite links in src/library/Recommended. A suitable
incantation from the top level of the R sources using wget
might be (for the correct value of dir)
wget -r -l1 --no-parent -A\*.gz -nd -P src/library/Recommended \ https://CRAN.R-project.org/src/contrib/dir ./tools/link-recommended
R will configure and build under most common Unix and Unix-alike platforms including ‘cpu-*-linux-gnu’ for the ‘alpha’, ‘arm64’ (also known as ‘aarch64’, ‘ix86’, ‘mips’, ‘mipsel’#, ‘ppc64’, ‘riscv64’, ‘s390x’, ‘sparc64’, and ‘x86_64’ CPUs, ‘aarch64-apple-darwin’3 and ‘x86_64-apple-darwin’ as well as perhaps (it is tested less frequently on these platforms) ‘x86_64-*-freebsd’, ‘x86_64-*-openbsd’ and ‘powerpc-ibm-aix6*’
Only 64-bit platforms are tested regularly, and configure
will warn if it encounters a 32-bit one.
In addition, binary distributions are available for some common Linux distributions (see the FAQ for current details) and for macOS. These are installed in platform-specific ways, so for the rest of this chapter we consider only building from the sources.
Cross-building is not possible: installing R builds a minimal version of R and then runs many R scripts to complete the build.
First review the essential and useful tools and libraries in
Essential and useful other programs under a Unix-alike, and install
those you
want or need. Ensure that either the environment variable TMPDIR
is either unset (and /tmp exists and can be written in and
scripts can be executed from) or points to the absolute path to a valid
temporary directory (one from which execution of scripts is allowed)
which does not contain spaces.
Choose a directory to install the R tree (R is not just a binary, but has additional data sets, help files, font metrics etc). Let us call this place R_HOME. Untar the source code. This should create directories src, doc, and several more under a top-level directory: change to that top-level directory (At this point North American readers should consult Setting paper size.) Issue the following commands:
./configure make
(See Using make if your make is not called ‘make’.) Users of Debian-based 64-bit systems4 may need
./configure LIBnn=lib make
Then check the built system works correctly by
make check
Failures are not necessarily problems as they might be caused by missing
functionality, but you should look carefully at any reported
discrepancies. (Some non-fatal errors are expected in locales that do
not support Latin-1, in particular in true C locales and
non-UTF-8 non-Western-European locales.) A failure in
tests/ok-errors.R may indicate inadequate resource limits
(see Running R).
More comprehensive testing can be done by
make check-devel
or
make check-all
see Testing an Installation
for the possibilities of doing this in parallel. Note that these checks
are only run completely if the recommended packages are installed. If
you have a local CRAN mirror, these checks can be speeded up by either
setting environment variable R_CRAN_WEB to its URL, or having a
file .R/repositories specifying it (see ?setRepositories).
make check-devel checks the included packages’ tests
directories. For grDevices more complete checks will be run if
the environment variable R_GRDEVICES_COMPARE_PS_PDF is set: those
checks are skipped on platforms using musl such as Alpine Linux, and
it knows about differences from transliterations in macOS 14 and later.
Parallel make is supported for building R but not for the ‘check’ targets (as the output is likely to be unreadably interleaved, although where supported5 GNU make’s -O may help).
If the configure and make commands execute
successfully, a shell-script front-end called R will be created
and copied to R_HOME/bin. You can link or copy this script
to a place where users can invoke it, for example to
/usr/local/bin/R. You could also copy the man page R.1 to
a place where your man reader finds it, such as
/usr/local/man/man1. If you want to install the complete R
tree to, e.g., /usr/local/lib/R, see Installation. Note:
you do not need to install R: you can run it from where it was
built.
You do not necessarily have to build R in the top-level source directory (say, TOP_SRCDIR). To build in BUILDDIR, run
cd BUILDDIR TOP_SRCDIR/configure make
and so on, as described further below. This has the advantage of always
keeping your source tree clean and is particularly recommended when you
work with a version of R from Subversion. (You may need
GNU make to allow this, and you will need no spaces
in the path to the build directory. It is unlikely to work if the
source directory has previously been used for a build.)
There are many settings which can be customized when building R and most are described in the file config.site in the top-level source directory. This can be edited, but for an installation using BUILDDIR it is better to put the changed settings in a newly-created file config.site in the build directory.
Now rehash if necessary, type R, and read the R manuals
and the R FAQ (files FAQ or
doc/manual/R-FAQ.html, or
https://CRAN.R-project.org/doc/FAQ/R-FAQ.html which always
has the version for the latest release of R).
Note: if you already have R installed, check that where you installed R replaces or comes earlier in your path than the previous installation. Some systems are set up to have /usr/bin (the standard place for a system installation) ahead of /usr/local/bin (the default place for installation of R) in their default path, and some do not have /usr/local/bin on the default path.
R by default provides help pages as plain text displayed in a pager,
with the options (see the help for help) of displaying help as
HTML or PDF.
By default HTML help pages are created when needed rather than being built at install time.
If you need to disable the server and want HTML help, there is the
option to build HTML pages when packages are installed
(including those installed with R). This is enabled by the
configure option --enable-prebuilt-html. Whether
R CMD INSTALL (and hence install.packages) pre-builds
HTML pages is determined by looking at the R installation and is
reported by R CMD INSTALL --help: it can be overridden by
specifying one of the INSTALL options --html or
--no-html.
The server is disabled by setting the environment variable
R_DISABLE_HTTPD to a non-empty value, either before R is
started or within the R session before HTML help (including
help.start) is used. It is also possible that system security
measures will prevent the server from being started, for example if the
loopback interface has been disabled. See
?tools::startDynamicHelp for more details.
There is a set of manuals that can be built from the sources,
Printed versions of all the help pages for base and recommended packages (around 3750 pages).
Printed versions of the help pages for selected base packages (around 2200 pages)
R FAQ
“An Introduction to R”.
“R Data Import/Export”.
“R Installation and Administration”, this manual.
“Writing R Extensions”.
“The R Language Definition”.
To make these (with ‘fullrefman’ rather than ‘refman’), use
make pdf to create PDF versions make info to create info files (not ‘refman’ nor ‘fullrefman’).
You will not be able to build any of these unless you have
texi2any version 5.1 or later installed, and for PDF you must
have texi2dvi and texinfo.tex installed (which are part
of the GNU texinfo distribution but are, especially
texinfo.tex, often made part of the TeX package in
re-distributions). The path to texi2any can be set by macro
‘TEXI2ANY’ in config.site. NB: texi2any requires
perl.
The PDF versions can be viewed using any recent PDF viewer: they have
hyperlinks that can be followed. The info files are suitable for
reading online with Emacs or the standalone GNU info
program. The PDF versions will be created using the paper size selected
at configuration (default ISO a4): this can be overridden by setting
R_PAPERSIZE
on the make command line, or setting R_PAPERSIZE in the
environment and using make -e. (If re-making the manuals for
a different paper size, you should first delete the file
doc/manual/version.texi. The usual value for North America would
be ‘letter’.)
There are some issues with making the PDF reference manual, fullrefman.pdf or refman.pdf. The help files contain both non-ASCII characters (e.g. in text.Rd) and upright quotes, neither of which are contained in the standard LaTeX Computer Modern fonts. We have provided the following alternatives:
times(The default.) Using standard PostScript fonts, Times Roman, Helvetica
and Courier. This works well both for on-screen viewing and for
printing. One disadvantage is that the Usage and Examples sections may
come out rather wide: this can be overcome by using in addition
either of the options inconsolata (on a Unix-alike only if found
by configure) or beramono, which replace the Courier
monospaced font by Inconsolata or Bera Sans Mono respectively. (You
will need the LaTeX package inconsolata6 or
bera installed.)
Note that in most LaTeX installations this will not actually use the standard fonts for PDF, but rather embed the URW clones NimbusRom, NimbusSans and (for Courier, if used) NimbusMon.
This needs LaTeX packages times, helvetic and (if used) courier installed.
lmUsing the Latin Modern fonts. These are not often installed as
part of a TeX distribution, but can obtained from
https://www.ctan.org/tex-archive/fonts/ps-type1/lm/ and
mirrors. This uses fonts rather similar to Computer Modern, but is not
so good on-screen as times.
The default can be overridden by setting the environment variable
R_RD4PDF. (On Unix-alikes, this will be picked up at install time
and stored in etc/Renviron, but can still be overridden when the
manuals are built, using make -e.) The usual7 default value for R_RD4PDF is
‘times,inconsolata,hyper’: omit ‘inconsolata’ if you do not have
LaTeX package inconsolata installed.
‘hyper’ is always enabled (with a fallback if LaTeX package
hyperref is not installed).
Further options, e.g for hyperref, can be included in a file Rd.cfg somewhere on your LaTeX search path. For example, if you prefer to hyperlink the text and not the page number in the table of contents use
\ifthenelse{\boolean{Rd@use@hyper}}{\hypersetup{linktoc=section}}{}
or
\ifthenelse{\boolean{Rd@use@hyper}}{\hypersetup{linktoc=all}}{}
to hyperlink both text and page number.
Any generated PDF manuals can be compacted by
make compact-pdf
provided qpdf and gs are available (see
?tools::compactPDF for how to specify them if not on the path).
E-book versions of most of the manuals in one or both of .epub and .mobi formats can be made by running in doc/manual one of
make ebooks make epub make mobi
This requires ebook-convert from
Calibre, or from
most Linux distributions. If necessary the path to
ebook-convert can be set as make macro EBOOK by editing
doc/manual/Makefile (which contains a commented value suitable
for macOS) or using make -e.
To ensure that the installed tree is usable by the right group of users,
set umask appropriately (perhaps to ‘022’) before unpacking
the sources and throughout the build process.
After
./configure make make check
(or, when building outside the source,
TOP_SRCDIR/configure, etc) have been completed
successfully, you can install the complete R tree to your system by
typing
make install
A parallel make can be used (but run make before make
install). Those using GNU make 4.0 or later may want to use
make -j n -O to avoid interleaving of output.
This will install to the following directories:
the front-end shell script and other scripts and executables
the man page
all the rest (libraries, on-line help system, …). Here LIBnn is usually ‘lib’, but may be ‘lib64’ on some 64-bit Linux systems. This is known as the R home directory.
where prefix is determined during configuration (typically
/usr/local) and can be set by running configure with
the option --prefix, as in
./configure --prefix=/where/you/want/R/to/go
where the value should be an absolute path. This causes make
install to install the R script to
/where/you/want/R/to/go/bin, and so on. The prefix of the
installation directories can be seen in the status message that is
displayed at the end of configure. The installation may need
to be done by the owner of prefix, often a root account.
There is the option of using make install-strip (see Debugging Symbols).
You can install into another directory tree by using
make prefix=/path/to/here install
at least with GNU make (but not some other Unix
makes).
More precise control is available at configure time via options: see
configure --help for details. (However, most of the ‘Fine
tuning of the installation directories’ options are not used by R.)
Configure options --bindir and --mandir are supported
and govern where a copy of the R script and the man
page are installed.
The configure option --libdir controls where the main R files are installed: the default is ‘eprefix/LIBnn’, where eprefix is the prefix used for installing architecture-dependent files, defaults to prefix, and can be set via the configure option --exec-prefix.
Each of bindir, mandir and libdir can also be
specified on the make install command line (at least for
GNU make).
The configure or make variables rdocdir and
rsharedir can be used to install the system-independent
doc and share directories to somewhere other than
libdir. The C header files can be installed to the value of
rincludedir: note that as the headers are not installed into a
subdirectory you probably want something like
rincludedir=/usr/local/include/R-4.4.1.
If you want the R home to be something other than libdir/R, use rhome: for example
make install rhome=/usr/local/lib64/R-4.4.1
will use a version-specific R home on a non-Debian Linux 64-bit system.
If you have made R as a shared/static library you can install it in your system’s library directory by
make prefix=/path/to/here install-libR
where prefix is optional, and libdir will give more
precise control.8 However, you should not install
to a directory mentioned in LDPATHS (e.g.
/usr/local/lib64) if you intend to work with multiple versions of
R, since that directory may be given precedence over the lib
directory of other R installations.
make install-strip
will install stripped executables, and on platforms where this is supported, stripped libraries in directories lib and modules and in the standard packages.
Note that installing R into a directory whose path contains spaces is not supported, and some aspects (such as installing source packages) will not work.
To install info and PDF versions of the manuals, use one or both of
make install-info make install-pdf
Once again, it is optional to specify prefix, libdir or
rhome (the PDF manuals are installed under the R home
directory).
More precise control is possible. For info, the setting used is that of
infodir (default prefix/info, set by configure
option --infodir). The PDF files are installed into the R
doc tree, set by the make variable rdocdir.
A staged installation is possible, that it is installing R into a
temporary directory in order to move the installed tree to its final
destination. In this case prefix (and so on) should reflect the
final destination, and DESTDIR should be used: see
https://www.gnu.org/prep/standards/html_node/DESTDIR.html.
You can optionally install the run-time tests that are part of
make check-all by
make install-tests
which populates a tests directory in the installation.
You can uninstall R by
make uninstall
optionally specifying prefix etc in the same way as specified for
installation.
This will also uninstall any installed manuals. There are specific targets to uninstall info and PDF manuals in file doc/manual/Makefile.
Target uninstall-tests will uninstall any installed tests, as
well as removing the directory tests containing the test results.
An installed shared/static libR can be uninstalled by
make prefix=/path/to/here uninstall-libR
Now 32-bit builds are unsupported, this section is only of historical interest, although in future the mechanisms could be used for different CPU types on the same OS (e.g. ‘x86_64’ and ‘aarch64’).
Some platforms can support closely related builds of R which can share all but the executables and dynamic objects. Examples include builds under Linux for different CPUs or 32- and 64-bit builds.
R supports the idea of architecture-specific builds, specified by
adding ‘r_arch=name’ to the configure line. Here
name can be anything non-empty, and is used to name subdirectories
of lib, etc, include and the package libs
subdirectories. Example names from other software are the use of
sparcv9 on Sparc Solaris and 32 by gcc on
‘x86_64’ Linux.
If you have two or more such builds you can install them over each other (and for 32/64-bit builds on one architecture, one build can be done without ‘r_arch’). The space savings can be considerable: on ‘x86_64’ Linux a basic install (without debugging symbols) took 74Mb, and adding a 32-bit build added 6Mb. If you have installed multiple builds you can select which build to run by
R --arch=name
and just running ‘R’ will run the last build that was installed.
R CMD INSTALL will detect if more than one build is installed and
try to install packages with the appropriate library objects for each.
This will not be done if the package has an executable configure
script or a src/Makefile file. In such cases you can install for
extra builds by
R --arch=name CMD INSTALL --libs-only pkg1 pkg2 ...
If you want to mix sub-architectures compiled on different platforms (for example ‘x86_64’ Linux and ‘i686’ Linux), it is wise to use explicit names for each, and you may also need to set libdir to ensure that they install into the same place.
When sub-architectures are used the version of Rscript in
e.g. /usr/bin will be the last installed, but
architecture-specific versions will be available in e.g.
/usr/lib64/R/bin/exec${R_ARCH}. Normally all installed
architectures will run on the platform so the architecture of
Rscript itself does not matter. The executable
Rscript will run the R script, and at that time the
setting of the R_ARCH environment variable determines the
architecture which is run.
When running post-install tests with sub-architectures, use
R --arch=name CMD make check[-devel|all]
to select a sub-architecture to check.
Sub-architectures were also used on Windows, but by selecting executables within the appropriate bin directory such as R_HOME/bin/x64. As from R 4.2.0 only the ‘x64’ subdirectory is used.
There are many other installation options, most of which are listed by
configure --help. Almost all of those not listed elsewhere in
this manual are either standard autoconf options not relevant
to R or intended for specialist uses by the R developers.
One that may be useful when working on R itself is the option
--disable-byte-compiled-packages, which ensures that the base
and recommended packages are not byte-compiled. (Alternatively the
(make or environment) variable R_NO_BASE_COMPILE can be set to a
non-empty value for the duration of the build.)
Option --with-internal-tzcode makes use of R’s own code and
copy of the IANA database for managing timezones. This will be
preferred where there are issues with the system implementation, usually
involving times after 2037 or before 1916. An alternative time-zone
directory9 can be used, pointed
to by environment variable TZDIR: this should contain files such
as Europe/London. On all tested OSes the system timezone was
deduced correctly, but if necessary it can be set as the value of
environment variable TZ.
Options --with-internal-iswxxxxx,
--with-internal-towlower and --with-internal-wcwidth
control the replacement of the system wide-character classification
(such as iswprint), case-changing (wctrans) and width
(wcwidth and wcswidth) functions by ones contained in the
R sources. Replacement of the classification functions has been done
for many years on macOS and AIX (and Windows): option
--with-internal-iswxxxxx allows this to be suppressed on those
platforms or used on others. Replacing the case-changing functions is
the default on macOS and Windows. Replacement of the width functions
has also been done for many years and remains the default. These
options will only matter to those working with non-ASCII character data,
especially in languages written in a non-Western script10 (which includes ‘symbols’ such as emoji). Note
that one of those iswxxxxx is iswprint which is used to
decide whether to output a character as a glyph or as a
‘\U{xxxxxx}’ escape—for example, try ‘"\U1f600"’, an
emoji. The width functions are of most importance in East Asian locale:
their values differ between such locales. (Replacing the system
functions provides a degree of platform-independence (including to OS
updates) but replaces it with a dependence on the R version.)
By default, configure adds a flag (usually -g) to the
compilation flags for C, Fortran and C++ sources. This will slow down
compilation and increase object sizes of both R and packages, so it
may be a good idea to change those flags (set ‘CFLAGS’ etc in
config.site before configuring, or edit files Makeconf
and etc/Makeconf between running configure and
make).
Having debugging symbols available is useful both when running R under a
debugger (e.g., R -d gdb) and when using sanitizers and
valgrind, all things intended for experts.
Debugging symbols (and some others) can be ‘stripped’ on installation by using
make install-strip
How well this is supported depends on the platform: it works best on
those using GNU binutils. On ‘x86_64’ Linux a typical
reduction in overall size was from 92MB to 66MB. On macOS debugging
symbols are not by default included in .dylib and .so
files, so there is negligible difference.
By default configure searches for suitable flags11 for OpenMP support for the C, C++ (default standard)
and Fortran compilers.
Only the C result is currently used for R itself, and only if
MAIN_LD/DYLIB_LD were not specified. This can be
overridden by specifying
R_OPENMP_CFLAGS
Use for packages has similar restrictions (involving SHLIB_LD and
similar: note that as Fortran code is by default linked by the C (or
C++) compiler, both need to support OpenMP) and can be overridden by
specifying some of
SHLIB_OPENMP_CFLAGS SHLIB_OPENMP_CXXFLAGS SHLIB_OPENMP_FFLAGS
Setting these to an empty value will disable OpenMP for that compiler
(and configuring with --disable-openmp will disable all
detection12 of OpenMP). The
configure detection test is to compile and link a standalone
OpenMP program, which is not the same as compiling a shared object and
loading it into the C program of R’s executable. Note that
overridden values are not tested.
C++ is not used by R itself, but support is provided for installing
packages with C++ code via make macros defined in file
etc/Makeconf (and with explanations in file config.site):
CXX CXXFLAGS CXXPICFLAGS CXXSTD CXX11 CXX11STD CXX11FLAGS CXX11PICFLAGS CXX14 CXX14STD CXX14FLAGS CXX14PICFLAGS CXX17 CXX17STD CXX17FLAGS CXX17PICFLAGS CXX20 CXX20STD CXX20FLAGS CXX20PICFLAGS CXX23 CXX23STD CXX23FLAGS CXX23PICFLAGS
The macros CXX etc are those used by default for C++ code.
configure will attempt to set the rest suitably, choosing for
CXXSTD and CXX11STD a suitable flag such as
-std=gnu++17 for C++17 support (which is required if C++ is to be
supported by default). Inferred values can be overridden in file
config.site or on the configure command line:
user-supplied values will be tested by compiling some C++11/14/17/20/23
code.
It may be that there is no suitable flag for C++14/17/20/23 support with
the default compiler, in which case a different compiler could be
selected for CXX14/CXX17/CXX20/CXX23 with its
corresponding flags.
If no suitable compiler/flag is found for the default C++ compiler, one
can be set in file config.site via macros CXX
and CXXSTD. A user-specified compiler does not need to pass the
C++17 tests, so do this at your own risk as some packages may not compile.
The -std flag is supported by the GCC, clang++ and
Intel compilers. Currently accepted values are (plus some synonyms)
g++: c++11 gnu+11 c++14 gnu++14 c++17 gnu++17 c++2a gnu++2a (from 8)
c++20 gnu++20 (from 10) c++23 gnu++23 c++2b gnu++2b (from 11)
Intel: c++11 gnu+11 c++14 gnu++14 c++17 gnu++17
c++20 gnu++20 (from 2021.1) c++2b gnu++2b (from 2022.2)
(Those for LLVM clang++ are documented at
https://clang.llvm.org/cxx_status.html, and follow g++:
-std=c++20 is supported from Clang 10, -std=c++2b from
Clang 13 and -std=c++23 from Clang 17. Apple Clang supports
-std=c++2b from 13.1.6: version 15.0.0 does not support
-std=c++23.)
‘Standards’ for g++ starting with ‘gnu’ enable ‘GNU
extensions’: what those are is hard to track down.
For the use of C++ in R packages see the ‘Writing R Extensions’ manual. Prior to R 3.6.0 the default C++ standard was that of the compiler used: currently it is C++17.
https://en.cppreference.com/w/cpp/compiler_support indicates which versions of common compilers support (parts of) which C++ standards. GCC introduced C++17 support gradually, but version 7 should suffice.
Compiling R requires C99 or later: C11 and C17 are minor updates, but the substantial update planned for ‘C23’ (now expected ca April 2024) will also be supported.
As from R 4.3.0 there is support for packages to indicate their
preferred C version. Macros CC17, C17FLAGS, CC23
and C23FLAGS can be set in config.site (there are examples
there). Those for C17 should support C17 or earlier and not allow C23
additions so for example bool, true and false can
be used as identifiers. Those for C23 should support new types such as
bool.
Some compilers warn enthusiastically about prototypes. For most,
omitting -Wstrict-prototypes in C17FLAGS suffices.
However, versions 15 and later of LLVM clang and 14.0.3 and
later of Apple clang warn by default in all modes if -Wall or
-pedantic is used, and may need
-Wno-strict-prototypes.
There is support for using link-time optimization (LTO) if the toolchain supports it: configure with flag --enable-lto. When LTO is enabled it is used for compiled code in add-on packages unless the flag --enable-lto=R is used13.
The main benefit seen to date from LTO has been detecting long-standing
bugs in the ways packages pass arguments to compiled code and between
compilation units. Benchmarking in 2020 with
gcc/gfortran 10 showed gains of a few percent
in increased performance and reduction in installed size for builds
without debug symbols, but large size reductions for some
packages14 with debug symbols. (Performance and size gains are said to be
most often seen in complex C++ builds.)
Whether toolchains support LTO is often unclear: all of the C compiler, the Fortran compiler15 and linker have to support it, and support it by the same mechanism (so mixing compiler families may not work and a non-default linker may be needed). It has been supported by the GCC and LLVM projects for some years with diverging implementations.
LTO support was added in 2011 for GCC 4.5 on Linux but was little used before 2019: compiler support has steadily improved over those years and --enable-lto=R is nowadays used for some routine CRAN checking.
Unfortunately --enable-lto may be accepted but silently do nothing useful if some of the toolchain does not support LTO: this is less common than it once was.
Various macros can be set in file config.site to customize how LTO is used. If the Fortran compiler is not of the same family as the C/C++ compilers, set macro ‘LTO_FC’ (probably to empty). Macro ‘LTO_LD’ can be used to select an alternative linker should that be needed.
This has been tested on Linux with
gcc/gfortran 8 and later: that needed setting
(e.g. in config.site)
AR=gcc-ar RANLIB=gcc-ranlib
For non-system compilers or if those wrappers have not been installed one may need something like
AR="ar --plugin=/path/to/liblto_plugin.so" RANLIB="ranlib --plugin=/path/to/liblto_plugin.so"
and NM may be needed to be set analogously. (If using an
LTO-enabled build to check packages, set environment variable
UserNM16 to ‘gcc-nm’.)
With GCC 5 and later it is possible to parallelize parts of the LTO linking process: set the make macro ‘LTO’ to something like ‘LTO=-flto=8’ (to use 8 threads), for example in file config.site.
Under some circumstances and for a few packages, the PIC flags have needed overriding on Linux with GCC 9: e.g use in config.site:
CPICFLAGS=-fPIC CXXPICFLAGS=-fPIC CXX11PICFLAGS=-fPIC CXX14PICFLAGS=-fPIC CXX17PICFLAGS=-fPIC CXX20PICFLAGS=-fPIC FPICFLAGS=-fPIC
We suggest only using these if the problem is encountered (it had not been seen on CRAN with GCC 10–14 at the time of writing).
Note that R may need to be re-compiled after even a minor update to the compiler (e.g. from 13.1 to 13.2).
LLVM supports another type of LTO called ‘Thin LTO’ as well as a similar
implementation to GCC, sometimes called ‘Full LTO’. (See
https://clang.llvm.org/docs/ThinLTO.html.) Currently the LLVM
compilers relevant to R are clang and flang for
which this can be selected by setting macro ‘LTO=-flto=thin’. LLVM
has
AR=llvm-ar RANLIB=llvm-ranlib
(but macOS does not, and these are not needed there). Where the linker supports a parallel backend for Thin LTO this can be specified via the macro ‘LTO_LD’: see the URL above for per-linker settings and further linking optimizations.)
For example, on macOS one might use
LTO=-flto=thin LTO_FC= LTO_LD=-Wl,-mllvm,-threads=4
to use Thin LTO with 4 threads for C/C++ code, but skip LTO for Fortran
code compiled with gfortran.
It is said to be particularly beneficial to use -O3 for
clang in conjunction with LTO.
It seems that flang may support LTO, but with no documentation
as yet.
The 2020s versions of Intel’s C/C++ compilers are based on LLVM and as such support LLVM-style LTO, both ‘full’ and ‘thin’. This might use something like
LTO=-flto=thin -flto-jobs=8
LTO effectively compiles all the source code in a package as a single compilation unit and so allows the compiler (with sufficient diagnostic flags such as -Wall) to check consistency between what are normally separate compilation units.
With gcc/gfortran 9.x and later17 LTO will flag inconsistencies in calls to Fortran
subroutines/functions, both between Fortran source files and between
Fortran and C/C++. gfortran 8.4, 9.2 and later can help
understanding these by extracting C prototypes from Fortran source files
with option -fc-prototypes-external, e.g. that (at the time
of writing) Fortran LOGICAL corresponds to int_least32_t *
in C.
Full post-installation testing is possible only if the test files have been installed with
make install-tests
which populates a tests directory in the installation.
If this has been done, two testing routes are available. The first is
to move to the home directory of the R installation (as given by
R RHOME or from R as R.home()) and run
cd tests ## followed by one of ../bin/R CMD make check ../bin/R CMD make check-devel ../bin/R CMD make check-all
and other useful targets are test-BasePackages and
test-Recommended to run tests of the standard and recommended
packages (if installed) respectively.
This re-runs all the tests relevant to the installed R (including for example the code in the package vignettes), but not for example the ones checking the example code in the manuals nor making the standalone Rmath library. This can occasionally be useful when the operating environment has been changed, for example by OS updates or by substituting the BLAS (see Shared BLAS).
Parallel checking of packages may be possible: set the environment
variable TEST_MC_CORES to the maximum number of processes to be
run in parallel. This affects both checking the package examples (part
of make check) and package sources (part of make
check-devel and make check-recommended). It does require a
make command which supports the make -j n
option: most do.
Alternatively, the installed R can be run, preferably with --vanilla. Then
pdf("tests.pdf") ## optional, but prevents flashing graphics windows
Sys.setenv(LC_COLLATE = "C", LC_TIME = "C", LANGUAGE = "en")
tools::testInstalledBasic("both")
tools::testInstalledPackages(scope = "base")
tools::testInstalledPackages(scope = "recommended")
runs the basic tests and then all the tests on the standard and recommended packages. These tests can be run from anywhere: the basic tests write their results in the tests folder of the R home directory and run fewer tests than the first approach: in particular they do not test things which need Internet access—that can be tested by
tools::testInstalledBasic("internet")
It is possible to test the installed packages (but not their
package-specific tests) by testInstalledPackages even if
make install-tests was not run. The outputs are written under the
current directory unless a different one is specified by outDir.
Note that the results may depend on the language set for times and messages: for maximal similarity to reference results you may want to try setting (before starting the R session)
LANGUAGE=en
and use a UTF-8 or Latin-1 locale.
[The rest of this paragraph is only relevant after release.] The bin/windows directory of a CRAN site contains binaries for a base distribution and a large number of add-on packages from CRAN to run on 64-bit ‘x86_64’ Windows.
R is most tested on current versions of Windows 10 and Windows Server 2022 with UTF-8 as the charset encoding. It works also on Windows 11. It runs on older versions of Windows, but normally with other charset encoding and may require manual installation of the Universal C Runtime (UCRT).
Your file system must allow long file names (as is likely except perhaps for some network-mounted systems). If it does not also support conversion to short name equivalents (a.k.a. DOS 8.3 names), then R must be installed in a path that does not contain spaces.
Installation is via the installer R-4.4.1patched-win.exe. Just double-click on the icon and follow the instructions. You can uninstall R from the Control Panel.
You will be asked to choose a language for installation: that choice applies to both installation and un-installation but not to running R itself.
See the R Windows FAQ for more details on the binary installer and for information on use on older Windows systems.
It is possible to use other 64-bit toolchains (including ‘MSYS2’) with UCRT support to build R, but this manual only documents that used for recent binary distributions of R. When using other toolchains, makefiles of R and packages may need to be adapted.
The binary distribution of R is currently built with tools from Rtools44 for Windows. See Building R and packages for more details on how to use it.
The toolset includes compilers (currently GCC version 13.2.0 with selected additional patches) and runtime libraries from the ‘MinGW-w64’ project and a number of pre-compiled static libraries and headers used by R and R packages, compiled by ’MXE’ (M cross environment, with updates maintained by Tomas Kalibera). The toolset also includes build tools from the the ’MSYS2’ project. Additional build tools packaged by ’MSYS2’ may be installed via a package manager (‘pacman’).
There is also an experimental variant of Rtools44 with support for 64-bit
ARM CPUs (aarch64) via LLVM 17 toolchain using clang/flang-new compilers,
lld linker, and libc++.
The toolsets used for 64-bit Windows from 2008 to 2022 were based on MinGW-w64. The assistance of Yu Gong at a crucial step in porting R to MinGW-w64 is gratefully acknowledged, as well as help from Kai Tietz, the lead developer of the MinGW-w64 project and from Martin Storsjo.
Both building R and checking packages need a distribution of LaTeX
installed, with the directory containing pdflatex on the path.
The ‘MiKTeX’ (https://miktex.org/) distribution of LaTeX is that used on CRAN. This can be set up to install extra packages ‘on the fly’ (without asking), which is the simplest way to use it. The ‘basic’ version of ‘MiKTeX’ will need to add some packages.18 In any case ensure that the inconsolata package is installed—you can check with the ‘MiKTeX’ Package Manager.
It is also possible to use the TeX Live distribution from https://www.tug.org/texlive/. (The CRAN package tinytex can install and manage a subset of TeX Live.)
You can test a build by running
make check
The recommended packages can be checked by
make check-recommended
Other levels of checking are
make check-devel
for a more thorough check of the R functionality, and
make check-all
for both check-devel and check-recommended.
If a test fails, there will almost always be a .Rout.fail file in the directory being checked (often tests/Examples or tests): examine the file to help pinpoint the problem.
Parallel checking of package sources (part of make check-devel
and make check-recommended) is possible: see the environment
variable TEST_MC_CORES to the maximum number of processes to be
run in parallel.
The Windows installer contains a set of test files used when building R.
The toolset is not needed to run these tests, but more comprehensive
analysis of errors will be given if diff is in the path.
Launch either Rgui or Rterm (preferred), preferably with
--vanilla. Then run
Sys.setenv(LC_COLLATE = "C", LC_TIME="C", LANGUAGE = "en")
tools::testInstalledBasic("both")
tools::testInstalledPackages(scope = "base")
tools::testInstalledPackages(scope = "recommended")
runs the basic tests and then all the tests on the standard and recommended
packages. These tests can be run from anywhere: testInstalledBasic
writes results in the tests folder of the R home directory (as
given by R.home()) and testInstalledPackages under the current
directory unless a different one is specified by outDir.
For the tests folder to be writeable, one normally needs to install R to a directory other than the default C:\Program Files. The installer also allows to install R without Administrator privileges, see the R Windows FAQ for more details.
The results of example(md5sums) when testing tools may
differ from the reference output as some files are installed with
Windows’ CRLF line endings. Also, expect differences in
reg-plot-latin1.pdf.
One can also run tests from the toolset shell (e.g. bash) similarly
to a Unix-like installation. Move to the home directory of the R
installation (as given by R RHOME or from R as R.home())
and run
cd tests ## followed by one of ../bin/R CMD make check ../bin/R CMD make check-devel ../bin/R CMD make check-all
Remember that LaTeX needs to be on the path.
[The rest of this paragraph is only relevant after release.] The front page of a CRAN site has a link ‘Download R for (Mac) OS X’ which takes you to a new page. Two files are offered for download, R-4.4.1-arm64.pkg and R-4.4.1-x86_64.pkg. Both are for macOS 11 or later (Big Sur, Monterey, Ventura, Sonoma, …).
The first is for ‘Apple Silicon’ (aka ‘M1’, ‘M2’, …) Macs, the second for older Macs with an ‘x86_64’ (Intel) CPU.
It is important that if you use a binary installer package that your OS is fully updated: look at ‘Software Update’ in ’System Preferences’ to be sure.
To install, just double-click on the icon of the file you downloaded. At the ‘Installation Type’ stage, note the option to ‘Customize’. This currently shows four components: everyone will need the ‘R Framework’ component: the remaining components are optional. (The ‘Tcl/Tk’ component is needed to use package tcltk. The ‘Texinfo’ component is only needed by those installing source packages or R from its sources.)
Note for Ventura users: installation from the Downloads folder may not be allowed or may require additional authorization, so we suggest you download somewhere else such as your desktop or home folder.
These are Apple Installer packages. If you encounter any problem during the installation, please check the Installer log by clicking on the “Window” menu and item “Installer Log”. The full output (select “Show All Log”) is useful for tracking down problems. Note that the installer is clever enough to try to upgrade the last-installed version of the application where you installed it (which may not be where you want this time …).
Various parts of the build require XQuartz to be installed: see
https://www.xquartz.org/releases/.19 These include the tcltk package
and the X11 graphics device: attempting to use these without
XQuartz will remind you. This is also needed for some
builds of the cairographics-based devices (which are not often used on
macOS) such as png(type = "cairo") and svg() and some
third-party packages (e.g. rgl).
If you update your macOS version, you should re-install R (and perhaps XQuartz): the installer may tailor the installation to the current version of the OS.
Installers for R-patched and R-devel are usually available from https://mac.R-project.org. (Some of these packages may be unsigned/not notarized: to install those Control/right/two-finger click, select ‘Open With’ and ‘Installer’.)
For building R from source, see macOS.
There are two ways to run R on macOS from a CRAN binary distribution.
There is a GUI console normally installed with the R icon in /Applications which you can run by double-clicking (e.g. from Launchpad or Finder). (If you cannot find it there it was possibly installed elsewhere so try searching for it in Spotlight.) This is usually referred to as R.APP to distinguish it from command-line R: its user manual is currently part of the macOS FAQ at https://cran.r-project.org/bin/macosx/RMacOSX-FAQ.html and can be viewed from R.APP’s ‘Help’ menu.
You can run command-line R and Rscript from a
Terminal20 so these can be typed as commands
as on any other Unix-alike: see the next chapter of this manual. There
are some small differences which may surprise users of R on other
platforms, notably the default location of the personal library
directory (under ~/Library/R, e.g.
~/Library/R/arm64/4.4/library), and that warnings, messages and
other output to stderr are highlighted in bold.
Those using the zsh shell (the default for new user accounts)
might find the command R being masked by the zsh
builtin r (which recalls commands). One can use a full path
to R in an alias, or add disable r to ~/.zshrc.
It has been reported that running R.APP may fail if no preferences are stored, so if it fails when launched for the very first time, try it again (the first attempt will store some preferences).
Users of R.APP need to be aware of the ‘App Nap’ feature (https://developer.apple.com/library/archive/releasenotes/MacOSX/WhatsNewInOSX/Articles/MacOSX10_9.html) which can cause R tasks to appear to run very slowly when not producing output in the console. Here are ways to avoid it:
defaults write org.R-project.R NSAppSleepDisabled -bool YES
Using the X11 graphics device or the X11-based versions of View()
and edit() for data frames and matrices (the latter are the
default for command-line R but not R.APP) requires
XQuartz to be installed.
Under some rather nebulous circumstances messages have been seen from
fontconfig about missing/unreadable configuration files when
using cairo-based devices, especially X11(type = "cairo"). With
XQuartz installed there are two fontconfig areas from different
versions and it can help to set
setenv FONTCONFIG_PATH /opt/X11/lib/X11/fontconfig
Another symptom has been that italic/oblique fonts are replaced by upright ones.
R for macOS consists of two parts: the GUI (R.APP) and the R framework. Un-installation is as simple as removing those folders (e.g. by dragging them onto the Bin21). The typical installation will install the GUI into the /Applications/R.app folder and the R framework into the /Library/Frameworks/R.framework folder. The links to R and Rscript in /usr/local/bin should also be removed.
If you want to get rid of R more completely using a Terminal, simply run:
sudo rm -Rf /Library/Frameworks/R.framework /Applications/R.app \ /usr/local/bin/R /usr/local/bin/Rscript
The installation consists of up to four Apple packages:22 for the ‘Apple Silicon’ build, org.R-project.arm64.R.fw.pkg,
org.R-project.arm64.R.GUI.pkg, org.r-project.arm64.tcltk
and org.r-project.arm64.texinfo. You can use sudo pkgutil
--forget if you want the Apple Installer to forget about the package
without deleting its files (useful for the R framework when
installing multiple R versions in parallel), or after you have
deleted the files. NB: the package names are case-sensitive and
the R domain is named inconsistently.
Uninstalling the Tcl/Tk and Texinfo components (which are installed under /opt/R/arm64 on a ‘arm64’ build and /opt/R/x86_64 for an ‘x86_64’ one) is not as simple. You can list the files they installed in a Terminal by e.g.
pkgutil --files org.r-project.arm64.tcltk pkgutil --files org.r-project.arm64.texinfo
(For the ‘Intel build, replace arm64 by x86_64.)
These are paths relative to /, the root of the file system.
If you are not compiling R nor installing packages from source you could remove all of /opt/R/arm64 or /opt/R/x86_64.
The installer will remove any previous version23 of the R framework which it
finds installed. This can be avoided by using pkgutil
--forget (see the previous section). However, note that different
versions are installed under
/Library/Frameworks/R.framework/Versions as 4.4-arm64 (or
4.4-x86_64), 4.3 and so on, so it is not possible to have
different ‘4.x.y’ versions installed for the same ‘x’ and CPU
type.
R.APP will always run the ‘current’ version of R, that is the last installed version.
How to start R and what command-line options are available is discussed in Invoking R in An Introduction to R.
You should ensure that the shell has set adequate resource limits: R
expects a stack size of at least 8MB and to be able to open at least 256
file descriptors. (Any modern OS should have default limits at least as
large as these, but apparently NetBSD may not. Use the shell command
ulimit (sh/bash) or limit
(csh/tcsh) to check.) For some
compilers24 and packages a larger
stack size has been needed: 20-25MB has sufficed to date.
R makes use of a number of environment variables, the default values
of many of which are set in file R_HOME/etc/Renviron (there
are none set by default on Windows and hence no such file). These are
set at configure time, and you would not normally want to
change them – a possible exception is R_PAPERSIZE (see Setting paper size). The paper size will be deduced from the ‘LC_PAPER’
locale category if it exists and R_PAPERSIZE is unset, and this
will normally produce the right choice from ‘a4’ and ‘letter’
on modern Unix-alikes (but can always be overridden by setting
R_PAPERSIZE).
Various environment variables can be set to determine where R creates
its per-session temporary directory. The environment variables
TMPDIR, TMP and TEMP are searched in turn and the
first one which is set and points to a writable area is used. If none
do, the final default is /tmp on Unix-alikes and the value of
R_USER on Windows. The path should be an absolute path not
containing spaces25
(and it is best to avoid non-alphanumeric characters such as + or
quotes).
Some Unix-alike systems are set up to remove files and directories
periodically from /tmp, for example by a cron job
running tmpwatch. Set TMPDIR to another directory
before starting long-running jobs on such a system.
Note that TMPDIR will be used to execute configure
scripts when installing packages, so if /tmp has been mounted as
‘noexec’, TMPDIR needs to be set to a directory from which
execution is allowed.
It is helpful to use the correct terminology. A package is
loaded from a library by the function library(). Thus a
library is a directory containing installed packages; the main library
is R_HOME/library, but others can be used, for example by
setting the environment variable R_LIBS or using the R function
.libPaths(). To avoid any confusion you will often see a library
directory referred to as a ‘library tree’.
The set of packages loaded on startup is by default
> getOption("defaultPackages")
[1] "datasets" "utils" "grDevices" "graphics" "stats" "methods"
(plus, of course, base) and this can be changed by setting the
option in startup code (e.g. in ~/.Rprofile). It is initially
set to the value of the environment variable R_DEFAULT_PACKAGES if
set (as a comma-separated list). Setting R_DEFAULT_PACKAGES=NULL
ensures that only package base is loaded.
Changing the set of default packages is normally used to reduce the set
for speed when scripting: in particular not using methods will
reduce the start-up time by a factor of up to two. But it can also be
used to customize R, e.g. for class use. Rscript
also checks the environment variable R_SCRIPT_DEFAULT_PACKAGES;
if set, this takes precedence over R_DEFAULT_PACKAGES.
R packages are installed into libraries, which are directories in the file system containing a subdirectory for each package installed there.
R comes with a single library, R_HOME/library which is the value of the R object ‘.Library’ containing the standard and recommended26 packages. Both sites and users can create others and make use of them (or not) in an R session. At the lowest level ‘.libPaths()’ can be used to add paths to the collection of libraries or to report the current collection.
R will automatically make use of a site-specific library
R_HOME/site-library if this exists (it does not in a
vanilla R installation). This location can be overridden by
setting27 ‘.Library.site’ in
R_HOME/etc/Rprofile.site, or (not recommended) by setting
the
environment variable R_LIBS_SITE.
Users can have one or more libraries, normally specified by the
environment variable R_LIBS_USER. This has a default value (to
see it, use ‘Sys.getenv("R_LIBS_USER")’ within an R session),
but that is only used if the corresponding directory actually exists
(which by default it will not).
Both R_LIBS_USER and R_LIBS_SITE can specify multiple
library paths, separated by colons (semicolons on Windows).
Packages may be distributed in source form or compiled binary form. Installing source packages which contain C/C++/Fortran code requires that compilers and related tools be installed. Binary packages are platform-specific and generally need no special tools to install, but see the documentation for your platform for details.
Note that you may need to specify implicitly or explicitly the library to which the package is to be installed. This is only an issue if you have more than one library, of course.
Ensure that the environment variable TMPDIR is either unset (and
/tmp exists and can be written in and executed from) or is the
absolute path to a valid temporary directory, not containing spaces.
For most users it suffices to call ‘install.packages(pkgname)’ or its GUI equivalent if the intention is to install a CRAN package and internet access is available.28 On most systems ‘install.packages()’ will allow packages to be selected from a list box (typically with thousands of items).
To install packages from source on a Unix-alike use in a terminal
R CMD INSTALL -l /path/to/library pkg1 pkg2 ...
The part ‘-l /path/to/library’ can be omitted, in which case the
first library of a normal R session is used (that shown by
.libPaths()[1]).
There are a number of options available: use R CMD INSTALL --help
to see the current list.
Alternatively, packages can be downloaded and installed from within
R. First choose your nearest CRAN mirror using
chooseCRANmirror(). Then download and install packages
pkg1 and pkg2 by
> install.packages(c("pkg1", "pkg2"))
The essential dependencies of the specified packages will also be fetched.
Unless the library is specified (argument lib) the first library
in the library search path is used: if this is not writable, R will
ask the user (in an interactive session) if the default personal library
should be created, and if allowed to will install the packages there.
If you want to fetch a package and all those it depends on (in any way) that are not already installed, use e.g.
> install.packages("Rcmdr", dependencies = TRUE)
install.packages can install a source package from a local
.tar.gz file (or a URL to such a file) by setting argument
repos to NULL: this will be selected automatically if the
name given is a single .tar.gz file.
install.packages can look in several repositories, specified as a
character vector by the argument repos: these can include a
CRAN mirror, Bioconductor, R-forge, rforge.net,
local archives, local files, …). Function
setRepositories() can select amongst those repositories that the
R installation is aware of.
Something which sometimes puzzles users is that install.packages()
may report that a package which they believe should be available is not
found. Some possible reasons:
getOption("repos")
pkg, use
av <- available.packages(filters=list()) av[av[, "Package"] == pkg, ]
in your R session, and look at the ‘Depends’ and ‘OS_type’ fields (there may be more than one matching entry). If the package depends on a version of R later than the one in use, it is possible that an earlier version is available which will work with your version of R: for CRAN look for ‘Old sources’ on the package’s CRAN landing page and manually retrieve an appropriate version (of comparable age to your version of R).
Naive users sometimes forget that as well as installing a package, they
have to use library to make its functionality available.
What install.packages does by default is different on Unix-alikes
(except macOS) and Windows. On Unix-alikes it consults the list of
available source packages on CRAN (or other
repositories), downloads the latest version of the package sources, and
installs them (via R CMD INSTALL). On ‘x86_64’ Windows
it looks (by default) first at the list of binary versions of
packages available for your version of R and downloads the latest
versions (if any). If no binary version is available or the source
version is newer, it will install the source versions of packages
without compiled C/C++/Fortran code, and offer to do so for those with,
if make is available (and this can be tuned by option
"install.packages.compile.from.source").
[At present binary packages are not distributed for ‘aarch64’ Windows, so most of this subsection only applies to ‘x86_64’.]
On Windows install.packages can also install a binary package
from a local zip file (or the URL of such a file) by setting
argument repos to NULL. Rgui.exe has a menu
Packages with a GUI interface to install.packages,
update.packages and library.
Windows binary packages for R were distributed as a single binary containing either or both architectures (32- and 64-bit). Prior to R 4.2.0, they might contain only the 32-bit architecture.
R CMD INSTALL works in Windows to install source packages. No
additional tools are needed if the package does not contain compiled
code, and install.packages(type="source") will work for such
packages. Those with compiled code need the tools (see The Windows toolset). The tools are found automatically by R when installed by
the toolset installer. See
Building
R and packages for more details.
Occasional permission problems after unpacking source packages have been
seen on some systems: these have been circumvented by setting the
environment variable R_INSTALL_TAR to ‘tar.exe’.
If you have only a source package that is known to work with current R and just want a binary Windows build of it, you could make use of the building service offered at https://win-builder.r-project.org/.
For almost all packages R CMD INSTALL will attempt to install
both 32- and 64-bit builds of a package if run from a 32/64-bit install
of R (only 64-bit builds and installs are supported since R 4.2.0).
It will report success if the installation of the architecture
of the running R succeeded, whether or not the other
architecture was successfully installed. The exceptions are packages
with a non-empty configure.win script or which make use of
src/Makefile.win. If configure.win does something
appropriate to both architectures use29 option
--force-biarch: otherwise R CMD INSTALL
--merge-multiarch can be applied to a source tarball to merge separate
32- and 64-bit installs. (This can only be applied to a tarball, and
will only succeed if both installs succeed.)
If you have a package without compiled code and no Windows-specific help, you can zip up an installation on another OS and install from that zip file on Windows. However, such a package can be installed from the sources on Windows without any additional tools.
On macOS install.packages works as it does on other Unix-alike
systems, but there is an additional type mac.binary (available
for the CRAN distribution but not when compiling R from
source) which can be passed to install.packages in order to
download and install binary packages from a suitable repository. These
binary package files for macOS have the extension ‘.tgz’. The
R.APP GUI provides menus for installation of either binary or source
packages, from CRAN, other repositories or local files.
On R builds using binary packages, the default is type both:
this looks first at the list of binary packages available for your
version of R and installs the latest versions (if any). If no binary
version is available or the source version is newer, it will install the
source versions of packages without compiled C/C++/Fortran code and offer
to do so for those with, if make is available.
Note that most binary packages which include compiled code are tied to a particular series (e.g. R 4.4.x or 4.3.x) of R.
Installing source packages which do not contain compiled code should
work with no additional tools. For others you will need the ‘Command
Line Tools’ for Xcode and compilers which match those used to
build R, plus a Fortran compiler for packages which contain Fortran
code: see macOS. Packages with C/C++ source code which link to
Fortran libraries (which include the BLAS and LAPACK
libraries) will need either the Fortran compiler or, for CRAN
binary distributions of R, to specify
FLIBS = -L/Library/Frameworks/R.framework/Resources/lib -lgfortran -lquadmath
in a Makevars file (see the next section) to point to the Fortran libraries in the binary distribution.
Package rJava and those which depend on it need a Java runtime installed and several packages need X11 installed, including those using Tk. See macOS and Java. Package rjags needs a build of JAGS installed under /usr/local, such as those at https://sourceforge.net/projects/mcmc-jags/files/JAGS/4.x/Mac%20OS%20X/.
Tcl/Tk extension BWidget used to be
distributed with R but no longer is; Tktable has been
distributed with recent versions of R.
The default compilers specified are shown in file /Library/Frameworks/R.framework/Resources/etc/Makeconf. At the time of writing those settings assumed that the C, Fortran and C++ compilers were on the path (see macOS). The settings can be changed, either by editing that file or in a file such as ~/.R/Makevars (see the next section). Entries which may need to be changed include ‘CC’, ‘CXX’, ‘FC’, ‘FLIBS’ and the corresponding flags, and perhaps ‘CXXCPP’, ‘DYLIB_LD’, ‘MAIN_LD’, ‘SHLIB_CXXLD’ and ‘SHLIB_LD’, as well as their ‘CXX11’, ‘CXX14’, ‘CXX17’ and ‘CXX20’ variants.
So for example you could select a specific LLVM
clang for both C and C++ with extensive checking by having in
~/.R/Makevars
SDK=/Library/Developer/CommandLineTools/SDKs/MacOSX.sdk CC = /usr/local/clang/bin/clang -isysroot $(SDK) CXX = /usr/local/clang/bin/clang++ -isysroot $(SDK) CXX11 = $CXX CXX14 = $CXX CXX17 = $CXX CXX20 = $CXX CXX23 = $CXX CFLAGS = -g -O2 -Wall -pedantic -Wconversion -Wno-sign-conversion CXXFLAGS = -g -O2 -Wall -pedantic -Wconversion -Wno-sign-conversion CXX11FLAGS = $CXXFLAGS CXX14FLAGS = $CXXFLAGS CXX17FLAGS = $CXXFLAGS CXX20FLAGS = $CXXFLAGS CXX23FLAGS = $CXXFLAGS
(the current SDK can be found by running xcrun -show-sdk-path)
and for the current macOS distribution of gfortran at
https://mac.r-project.org/tools/
FC = /opt/gfortran/bin/gfortran (arm64) FLIBS = -L/opt/gfortran/lib/gcc/aarch64-apple-darwin20.0/12.2.0 -L/opt/gfortran/lib -lgfortran -lemutls_w -lquadmath (Intel) FLIBS = -L/opt/gfortran/lib/gcc/x86_64-apple-darwin20.0/12.2.0 -L/opt/gfortran/lib -lgfortran -lquadmath
(line broken here for the manual only).
If that clang build
supports OpenMP, you can add
SHLIB_OPENMP_CFLAGS = -fopenmp SHLIB_OPENMP_CXXFLAGS = -fopenmp
to compile OpenMP-using packages. It will also be necessary to arrange
for the libomp.dylib library to be found at both install time and run
time, for example by copying/linking it somewhere that is searched such
as /usr/local/lib.
Apple includes many Open Source libraries in macOS but increasingly
without the corresponding headers (not even in Xcode nor the Command
Line Tools): they are often rather old versions. If installing packages
from source using them it is usually easiest to install a
statically-linked up-to-date copy of the Open Source package from its
sources or from https://mac.r-project.org/bin/.
But sometimes it is desirable/necessary to use Apple’s dynamically
linked library, in which case appropriate headers could be extracted
from the sources30 available via
https://opensource.apple.com/releases – this has been used for
iodbc.
Some care may be needed with selecting compilers when installing external
software for use with packages. The ‘system’ compilers as used when
building R are clang and clang++, but the Apple
toolchain also provides compilers called gcc and g++
which despite their names are based on LLVM and libc++ like the
system ones and which behave in almost the same way as the system ones.
Most Open Source software has a configure script developed
using GNU autoconf and hence will select gcc and
g++ as the default compilers: this usually works fine. For
consistency one can use
./configure CC=clang CFLAGS=-O2 CXX=clang++ CXXFLAGS=-O2
(avoiding autoconf’s default -g).
As from R 4.3.0, R CMD INSTALL and
install.packages() try to invoke configure with the
same compilers and flags used to build R.
The R system and package-specific compilation flags can be overridden
or added to by setting the appropriate Make variables in the personal
file HOME/.R/Makevars-R_PLATFORM (but
HOME/.R/Makevars.win or HOME/.R/Makevars.win64
on Windows), or if that does not exist, HOME/.R/Makevars,
where ‘R_PLATFORM’ is the platform for which R was built, as
available in the platform component of the R variable
R.version. The full path to an alternative personal
file31 can be specified via the environment variable
R_MAKEVARS_USER.
Package developers are encouraged to use this mechanism to enable a
reasonable amount of diagnostic messaging (“warnings”) when compiling,
such as e.g. -Wall -pedantic for tools from GCC, the GNU
Compiler Collection, and for LLVM (clang and flang-new).
Note that this mechanism can also be used when it is necessary to change the optimization level whilst installing a particular package. For example
## for C code CFLAGS = -g -O -mtune=native ## for C++ code CXXFLAGS = -g -O -mtune=native ## for C++11 code CXX11FLAGS = -g -O -mtune=native ## for fixed-form Fortran code FFLAGS = -g -O -mtune=native ## for C17 code C17FLAGS = -g -O -mtune=native -Wno-strict-prototypes
Note that if you have specified a non-default C++ or C standard, you need to set the flag(s) appropriate to that standard.
Another use is to override the settings in a binary installation of R.
For example, for the current distribution of gfortran at
https://mac.r-project.org/tools/
FC = /opt/gfortran/bin/gfortran (arm64) FLIBS = -L/opt/gfortran/lib/gcc/aarch64-apple-darwin20.0/12.2.0 -L/opt/gfortran/lib -lgfortran -lemutls_w -lquadmath (Intel) FLIBS = -L/opt/gfortran/lib/gcc/x86_64-apple-darwin20.0/12.2.0 -L/opt/gfortran/lib -lgfortran -lquadmath
(line broken here for the manual only).
There is also provision for a site-wide Makevars.site file under
R_HOME/etc (in a sub-architecture-specific directory if
appropriate). This is read immediately after Makeconf, and the
path to an alternative file can be specified by environment variable
R_MAKEVARS_SITE.
Note that these mechanisms do not work with packages which fail to pass settings down to sub-makes, perhaps reading etc/Makeconf in makefiles in subdirectories. Fortunately such packages are unusual.
When installing packages from their sources, there are some extra considerations on installations which use sub-architectures. These were commonly used on Windows prior to R 4.2.0 but can in principle be used on other platforms.
When a source package is installed by a build of R which supports multiple sub-architectures, the normal installation process installs the packages for all sub-architectures. The exceptions are
where there is an configure script, or a file src/Makefile.
where there is a non-empty configure.win script, or a file src/Makefile.win (with some exceptions where the package is known to have an architecture-independent configure.win, or if --force-biarch or field ‘Biarch’ in the DESCRIPTION file is used to assert so).
In those cases only the current architecture is installed. Further sub-architectures can be installed by
R CMD INSTALL --libs-only pkg
using the path to R or R --arch to select the
additional sub-architecture. There is also R CMD INSTALL
--merge-multiarch to build and merge the two architectures, starting
with a source tarball.
Packages are by default byte-compiled on installation. Byte-compilation can be controlled on a per-package basis by the ‘ByteCompile’ field in the DESCRIPTION file.
Some R packages contain compiled code which links to external software libraries. Unless the external library is statically linked (which is done as much as possible for binary packages on Windows and macOS), the libraries have to be found when the package is loaded and not just when it is installed. How this should be done depends on the OS (and in some cases the version).
For Unix-alikes except macOS the primary mechanism is the ld.so
cache controlled by ldconfig: external dynamic libraries
recorded in that cache will be found. Standard library locations will
be covered by the cache, and well-designed software will add its
locations (as for example openmpi does on Fedora). The secondary
mechanism is to consult the environment variable LD_LIBRARY_PATH.
The R script controls that variable, and sets it to the concatenation
of R_LD_LIBRARY_PATH, R_JAVA_LD_LIBRARY_PATH and the
environment value of LD_LIBRARY_PATH. The first two have defaults
which are normally set when R is installed (but can be overridden in
the environment) so LD_LIBRARY_PATH is the best choice for a user
to set.
On macOS the primary mechanism is to embed the absolute path to
dependent dynamic libraries into an object when it is compiled. Few
R packages arrange to do so, but it can be edited32 via
install_name_tool — that only deals with direct dependencies
and those would also need to be compiled to include the absolute paths
of their dependencies. If the choice of absolute path is to be deferred
to load time, how they are resolved is described in man dyld:
the role of LD_LIBRARY_PATH is replaced on macOS by
DYLD_LIBRARY_PATH and DYLD_FALLBACK_LIBRARY_PATH. Running
R CMD otool -L on the package shared object will show where
(if anywhere) its dependencies are
resolved. DYLD_FALLBACK_LIBRARY_PATH is preferred (and it is that
which is manipulated by the R script), but as from 10.11 (‘El
Capitan’) the default behaviour had been changed for security reasons
to discard these environment variables when invoking a shell script (and
R is a shell script). That makes the only portable option to set
R_LD_LIBRARY_PATH in the environment, something like
export R_LD_LIBRARY_PATH="`R RHOME`/lib:/opt/local/lib"
The precise rules for where Windows looks for DLLs are complex and
depend on the version of Windows. But for present purposes the main
solution is to put the directories containing the DLLs the package
links to (and any those DLLs link to) on the PATH.
The danger with any of the methods which involve setting environment
variables is of inadvertently masking a system library. This is less
for DYLD_FALLBACK_LIBRARY_PATH and for appending to
PATH on Windows (as it should already contain the system library
paths).
The command update.packages() is the simplest way to ensure that
all the packages on your system are up to date. It downloads the list
of available packages and their current versions, compares it with those
installed and offers to fetch and install any that have later versions
on the repositories.
An alternative interface to keeping packages up-to-date is provided by
the command packageStatus(), which returns an object with
information on all installed packages and packages available at multiple
repositories. The print and summary methods give an
overview of installed and available packages, the upgrade method
offers to fetch and install the latest versions of outdated packages.
One sometimes-useful additional piece of information that
packageStatus() returns is the status of a package, as
"ok", "upgrade" or "unavailable" (in the currently
selected repositories). For example
> inst <- packageStatus()$inst
> inst[inst$Status != "ok", c("Package", "Version", "Status")]
Package Version Status
Biobase Biobase 2.8.0 unavailable
RCurl RCurl 1.4-2 upgrade
Rgraphviz Rgraphviz 1.26.0 unavailable
rgdal rgdal 0.6-27 upgrade
Packages can be removed in a number of ways. From a command prompt they can be removed by
R CMD REMOVE -l /path/to/library pkg1 pkg2 ...
From a running R process they can be removed by
> remove.packages(c("pkg1", "pkg2"),
lib = file.path("path", "to", "library"))
Finally, one can just remove the package directory from the library.
Utilities such as install.packages can be pointed at any
CRAN-style repository, and R users may want to set up their
own. The ‘base’ of a repository is a URL such as
https://www.stats.ox.ac.uk/pub/RWin/: this must be an URL scheme
that download.packages supports (which also includes
‘https://’, ‘ftp://’ and ‘file://’). Under that base URL
there should be directory trees for one or more of the following types
of package distributions:
"source": located at src/contrib and containing
.tar.gz files. Other forms of compression can be used, e.g.
.tar.bz2 or .tar.xz files. Complete repositories contain
the sources corresponding to any binary packages, and in any case it is
wise to have a src/contrib area with a possibly empty
PACKAGES file.
"win.binary": located at bin/windows/contrib/x.y for
R versions x.y.z and containing .zip files for Windows.
"mac.binary": located at
bin/macosx/big-sur-arm64/contrib/4.y or
bin/macosx/big-sur-x86_64/contrib/4.y for the CRAN
builds for macOS for R versions 4.y.z, containing .tgz
files. (bin/macosx/contrib/4.y for y = 0, 1 or 2.)
Each terminal directory must also contain a PACKAGES file. This
can be a concatenation of the DESCRIPTION files of the packages
separated by blank lines, but only a few of the fields are needed. The
simplest way to set up such a file is to use function
write_PACKAGES in the tools package, and its help explains
which fields are needed. Optionally there can also be
PACKAGES.rds and PACKAGES.gz files, downloaded in
preference to PACKAGES. (If you have a mis-configured server
that does not report correctly non-existent files you may need these
files.)
To add your repository to the list offered by setRepositories(),
see the help file for that function.
Incomplete repositories are better specified via a
contriburl argument than via being set as a repository.
A repository can contain subdirectories, when the descriptions in the PACKAGES file of packages in subdirectories must include a line of the form
Path: path/to/subdirectory
—once again write_PACKAGES is the simplest way to set this up.
It can be convenient to run R CMD check on an installed
package, particularly on a platform which uses sub-architectures. The
outline of how to do this is, with the source package in directory
pkg (or a tarball filename):
R CMD INSTALL -l libdir pkg > pkg.log 2>&1 R CMD check -l libdir --install=check:pkg.log pkg
Where sub-architectures are in use the R CMD check line can be
repeated with additional architectures by
R --arch arch CMD check -l libdir --extra-arch --install=check:pkg.log pkg
where --extra-arch selects only those checks which depend on
the installed code and not those which analyse the sources. (If
multiple sub-architectures fail only because they need different
settings, e.g. environment variables, --no-multiarch may need
to be added to the INSTALL lines.) On Unix-alikes the
architecture to run is selected by --arch: this can also be
used on Windows with R_HOME/bin/R.exe, but it is more usual
to select the path to the Rcmd.exe of the desired
architecture.
So on Windows to install, check and package for distribution a source package from a tarball which has been tested on another platform one might use
.../bin/x64/Rcmd INSTALL -l libdir tarball --build > pkg.log 2>&1
Internationalization refers to the process of enabling support for many human languages, and localization to adapting to a specific country and language.
Current builds of R support all the character sets that the
underlying OS can handle. These are interpreted according to the
current locale, a sufficiently complicated topic to merit a
separate section. Note though that R has no built-in support for
right-to-left languages and bidirectional output, relying on the OS
services. For example, how character vectors in UTF-8 containing both
English digits and Hebrew characters are printed is OS-dependent (and
perhaps locale-dependent).
The other aspect of the internationalization is support for the translation of messages. This is enabled in almost all builds of R.
A locale is a description of the local environment of the user,
including the preferred language, the encoding of characters, the
currency used and its conventions, and so on. Aspects of the locale are
accessed by the R functions Sys.getlocale and
Sys.localeconv.
The system of naming locales is OS-specific. There is quite wide agreement on schemes, but not on the details of their implementation. A locale needs to specify
@latin, @cyrillic for Serbian, @iqtelif)
or language dialect (e.g. @saaho, a dialect of Afar, and
@bokmal and @nynorsk, dialects of Norwegian regarded by
some OSes as separate languages, no and nn).
R is principally concerned with the first (for translations) and third. Note that the charset may be deducible from the language, as some OSes offer only one charset per language.
Modern Linux uses the XPG33 locale specifications which have the form
‘en_GB’, ‘en_GB.UTF-8’, ‘aa_ER.UTF-8@saaho’,
‘de_AT.iso885915@euro’, the components being in the order listed
above. (See man locale and locale -a for more
details.) Similar schemes are used by most Unix-alikes: some (including
some distributions of Linux) use ‘.utf8’ rather than ‘.UTF-8’.
Note that whereas UTF-8 locales are nowadays almost universally used, locales such as ‘en_GB’ use 8-bit encodings for backwards compatibility.
Windows also uses locales, but specified in a rather less concise way. Most users will encounter locales only via drop-down menus, but more information and lists can be found by searching for ‘Windows language country strings’).
It offers only one encoding per language.
Some care is needed with Windows’ locale names. For example,
chinese is Traditional Chinese and not Simplified Chinese as used
in most of the Chinese-speaking world.
macOS supports locales in its own particular way, but the R GUI tries to
make this easier for users. See
https://developer.apple.com/library/archive/documentation/MacOSX/Conceptual/BPInternational/
for how users can set their locales. End users will
generally only see lists of languages/territories. Users of R in a
terminal may need to set the locale to something like ‘en_GB.UTF-8’
if it defaults to ‘C’ (as it sometimes does when logging in
remotely and for batch jobs: note whether Terminal sets the
LANG environment variable is an (advanced) preference, but does so
by default).
Internally macOS uses a form similar to Linux: the main difference from
other Unix-alikes is that where a character set is not specified it is
assumed to be UTF-8.
The preferred language for messages is by default taken from the locale.
This can be overridden first by the setting of the environment variable
LANGUAGE and then34
by the environment variables LC_ALL, LC_MESSAGES and
LANG. (The last three are normally used to set the locale and so
should not be needed, but the first is only used to select the language
for messages.) The code tries hard to map locales to languages, but on
some systems (notably Windows) the locale names needed for the
environment variable LC_ALL do not all correspond to XPG language
names and so LANGUAGE may need to be set. (One example is
‘LC_ALL=es’ on Windows which sets the locale to Estonian and the
language to Spanish.)
It is usually possible to change the language once R is running
via (not Windows) Sys.setlocale("LC_MESSAGES",
"new_locale"), or by setting an environment variable such as
LANGUAGE, provided35 the language you are changing to can be output in the current
character set. But this is OS-specific, and has been known to stop
working on an OS upgrade. Note that translated messages may be cached,
so attempting to change the language of an error that has already been
output in another language may not work.
Messages are divided into domains, and translations may be available for some or all messages in a domain. R makes use of the following domains.
R for the C-level error and warning messages from the R
interpreter.
R-pkg for the R stop, warning and
message messages in each package, including R-base for the
base package.
pkg for the C-level messages in each package.
RGui for the menus etc of the R for Windows GUI front-end.
Dividing up the messages in this way allows R to be extensible: as packages are loaded, their message translation catalogues can be loaded too.
R can be built without support for translations, but it is enabled by default.
R-level and C-level domains are subtly different, for example in the way strings are canonicalized before being passed for translation.
Translations are looked for by domain according to the currently specified language, as specifically as possible, so for example an Austrian (‘de_AT’) translation catalogue will be used in preference to a generic German one (‘de’) for an Austrian user. However, if a specific translation catalogue exists but does not contain a translation, the less specific catalogues are consulted. For example, R has catalogues for ‘en_GB’ that translate the Americanisms (e.g., ‘gray’) in the standard messages into English.36 Two other examples: there are catalogues for ‘es’, which is Spanish as written in Spain and these will by default also be used in Spanish-speaking Latin American countries, and also for ‘pt_BR’, which are used for Brazilian locales but not for locales specifying Portugal.
Translations in the right language but the wrong charset are made use of
by on-the-fly re-encoding. The LANGUAGE variable (only) can be a
colon-separated list, for example ‘se:de’, giving a set of
languages in decreasing order of preference. One special value is
‘en@quot’, which can be used in a UTF-8 locale to have American
error messages with pairs of single quotes translated to Unicode directional
quotes.
If no suitable translation catalogue is found or a particular message is not translated in any suitable catalogue, ‘English’37 is used.
See https://developer.r-project.org/Translations30.html for how to prepare and install translation catalogues.
The routines supporting the distribution and special38 functions in R and a few others are declared in C header file Rmath.h. These can be compiled into a standalone library for linking to other applications. (Note that they are not a separate library when R is built, and the standalone version differs in several ways.)
The makefiles and other sources needed are in directory src/nmath/standalone, so the following instructions assume that is the current working directory (in the build directory tree on a Unix-alike if that is separate from the sources).
Rmath.h contains ‘R_VERSION_STRING’, which is a character
string containing the current R version, for example "4.4.0".
There is full access to R’s handling of NaN, Inf and
-Inf via special versions of the macros and functions
ISNAN, R_FINITE, R_log, R_pow and R_pow_di
and (extern) constants R_PosInf, R_NegInf and NA_REAL.
There is no support for R’s notion of missing values, in particular
not for NA_INTEGER nor the distinction between NA and
NaN for doubles.
A little care is needed to use the random-number routines. You will need to supply the uniform random number generator
double unif_rand(void)
or use the one supplied (and with a shared library or DLL you may have to use the one supplied, which is the Marsaglia-multicarry with an entry point
set_seed(unsigned int, unsigned int)
to set its seeds).
The facilities to change the normal random number generator are
available through the constant N01_kind. This takes values
from the enumeration type
typedef enum {
BUGGY_KINDERMAN_RAMAGE,
AHRENS_DIETER,
BOX_MULLER,
USER_NORM,
INVERSION,
KINDERMAN_RAMAGE
} N01type;
(and ‘USER_NORM’ is not available).
If R has not already been made in the directory tree,
configure must be run as described in the main build
instructions.
Then (in src/nmath/standalone)
make
will make standalone libraries libRmath.a and libRmath.so (libRmath.dylib on macOS): ‘make static’ and ‘make shared’ will create just one of them.
To use the routines in your own C or C++ programs, include
#define MATHLIB_STANDALONE #include <Rmath.h>
and link against ‘-lRmath’ (and ‘-lm’ if needed on your OS).
The example file test.c does nothing useful, but is provided to
test the process (via make test). Note that you will probably
not be able to run it unless you add the directory containing
libRmath.so to the LD_LIBRARY_PATH environment variable
(libRmath.dylib, DYLD_FALLBACK_LIBRARY_PATH on macOS).
The targets
make install make uninstall
will (un)install the header Rmath.h and shared and static
libraries (if built). Both prefix= and DESTDIR are
supported, together with more precise control as described for the main
build.
‘make install’ installs a file for pkg-config to use by
e.g.
$(CC) `pkg-config --cflags libRmath` -c test.c $(CC) `pkg-config --libs libRmath` test.o -o test
On some systems ‘make install-strip’ will install a stripped shared library.
You need to set up39 almost all the tools to make R and then run (in a Unix-like shell)
(cd ../../gnuwin32; make MkRules) (cd ../../include; make -f Makefile.win config.h Rconfig.h Rmath.h) make -f Makefile.win
Alternatively, in a cmd.exe shell use
cd ../../include make -f Makefile.win config.h Rconfig.h Rmath.h cd ../nmath/standalone make -f Makefile.win
This creates a static library libRmath.a and a DLL Rmath.dll. If you want an import library libRmath.dll.a (you don’t need one), use
make -f Makefile.win shared implib
To use the routines in your own C or C++ programs using MinGW-w64, include
#define MATHLIB_STANDALONE #include <Rmath.h>
and link against ‘-lRmath’. This will use the first found of libRmath.dll.a, libRmath.a and Rmath.dll in that order, so the result depends on which files are present. You should be able to force static or dynamic linking via
-Wl,-Bstatic -lRmath -Wl,Bdynamic -Wl,-Bdynamic -lRmath
or by linking to explicit files (as in the ‘test’ target in Makefile.win: this makes two executables, test.exe which is dynamically linked, and test-static.exe, which is statically linked).
It is possible to link to Rmath.dll using other compilers, either directly or via an import library: if you make a MinGW-w64 import library as above, you will create a file Rmath.def which can be used (possibly after editing) to create an import library for other systems such as Visual C++.
If you make use of dynamic linking you should use
#define MATHLIB_STANDALONE #define RMATH_DLL #include <Rmath.h>
to ensure that the constants like NA_REAL are linked correctly.
(Auto-import will probably work with MinGW-w64, but it is better to be
sure. This is likely to also work with VC++, Borland and similar
compilers.)
This appendix gives details of programs you will need to build R on
Unix-like platforms, or which will be used by R if found by
configure.
Remember that some package management systems (such as RPM and Debian/Ubuntu’s) make a distinction between the user version of a package and the development version. The latter usually has the same name but with the extension ‘-devel’ or ‘-dev’: you need both versions installed.
You need a means of compiling C and Fortran 90 (see Using Fortran). Your C compiler should be
ISO/IEC 6005940, POSIX 1003.1 and C99-compliant.41 R tries to choose suitable
flags42 for the C compilers it knows about, but you may have to
set CC or CFLAGS suitably. (Note that options essential
to run the compiler even for linking, such as those to set the
architecture, should be specified as part of CC rather than in
CFLAGS.)
Unless you do not want to view graphs on-screen (or use macOS) you need ‘X11’ installed, including its headers and client libraries. For recent Fedora/RedHat distributions it means (at least) RPMs ‘libX11’, ‘libX11-devel’, ‘libXt’ and ‘libXt-devel’. On Debian/Ubuntu we recommend the meta-package ‘xorg-dev’. If you really do not want these you will need to explicitly configure R without X11, using --with-x=no.
The command-line editing (and command completion) depends on the
GNU readline library (including its headers): version
6.0 or later is needed for all the features to be enabled. Otherwise
you will need to configure with --with-readline=no (or
equivalent).
A suitably comprehensive iconv function is essential. The R
usage requires iconv to be able to translate between
"latin1" and "UTF-8", to recognize "" (as the
current encoding) and "ASCII", and to translate to and from the
Unicode wide-character formats "UCS-[24][BL]E" — this is true
by default for glibc43 but not of most commercial Unixes. However, you
can make use of GNU libiconv (as used on macOS: see
https://www.gnu.org/software/libiconv/).
The OS needs to have enough support44 for wide-character types: this is checked at configuration. Some C99 functions45 are required and checked for at configuration. A small number of POSIX functions46 are essential, and others47 will be used if available.
Installations of zlib (version 1.2.5 or later), libbz2
(version 1.0.6 or later: called bzip2-libs/bzip2-devel or
libbz2-1.0/libbz2-dev by some Linux distributions) and
liblzma48 version 5.0.3 or
later are required.
Either PCRE1 (version 8.32 or later, formerly known as just PCRE) or
PCRE2 is required: PCRE2 is preferred and using PCRE1 requires
configure option --with-pcre1. Only the 8-bit
library and headers are needed if these are packaged separately. JIT
support (optional) is desirable for the best performance. For PCRE2 >=
10.30 (which is desirable as matching has been re-written not to use
recursion and the Unicode tables were updated to version 10)
./configure --enable-jit
suffices. If building PCRE1 for use with R a suitable
configure command might be
./configure --enable-utf --enable-unicode-properties --enable-jit --disable-cpp
The --enable-jit flag is supported for most common CPUs but does not work (well or at all) for ‘arm64’ macOS.
Some packages require the ‘Unicode properties’ which are
optional for PCRE1: support for this and JIT can be checked at run-time
by calling pcre_config().
Library libcurl (version 7.28.0 or later) is required.
Information on libcurl is found from the curl-config
script: if that is missing or needs to be overridden49 there are macros to do so described in file
config.site.
A tar program is needed to unpack the sources and packages
(including the recommended packages). A version50 that can
automagically detect compressed archives is preferred for use with
untar(): the configure script looks for gtar and
gnutar before
tar – use environment variable TAR to override this.
(On NetBSD/OpenBSD systems set this to bsdtar if that is
installed.)
There need to be suitable versions of the tools grep and
sed: the problems are usually with old AT&T and BSD variants.
configure will try to find suitable versions (including
looking in /usr/xpg4/bin which is used on some commercial
Unixes).
You will not be able to build most of the manuals unless you have
texi2any version 5.1 or later installed (which requires
perl), and if not most of the HTML manuals will be linked
to a version on CRAN. To make PDF versions of the manuals you
will also need file texinfo.tex installed (which is part of the
GNU texinfo distribution but is often made part of the
TeX package in re-distributions) as well as
texi2dvi.51
Further, the versions of texi2dvi and texinfo.tex need
to be compatible: we have seen problems with older TeX distributions.
If you want to build from the R Subversion repository then
texi2any is highly recommended as it is used to create files
which are in the tarball but not stored in the Subversion repository.
The PDF documentation (including doc/NEWS.pdf) and building
vignettes needs pdftex and pdflatex. We require
LaTeX version 2005/12/01 or later (for UTF-8 support).
Building PDF package manuals (including the R reference manual) and
vignettes is sensitive to the version of the LaTeX package
hyperref and we recommend that the TeX distribution used is
kept up-to-date. A number of standard LaTeX packages are required
for the PDF manuals
(including url and some of the font packages such as times
and helvetic and also amsfonts) and others such
as hyperref and inconsolata are desirable (and without them
you may need to change R’s defaults: see Making the manuals).
Note that package hyperref (currently) requires packages
kvoptions, ltxcmds and refcount,
and inconsolata requires xkeyval.
Building the base vignettes requires fancyvrb,
natbib, parskip (which currently requires etoolbox)
and listings.
For distributions
based on TeX Live the simplest approach may be to install collections
collection-latex, collection-fontsrecommended,
collection-latexrecommended, collection-fontsextra and
collection-latexextra (assuming they are not installed by
default): Fedora uses names like texlive-collection-fontsextra and
Debian/Ubuntu like texlive-fonts-extra.
Programs qpdf and Ghostscript (gs) are desirable as
these will be used to compact the installed PDF vignettes and any PDF
manuals.
The essential programs should be in your PATH at the time
configure is run: this will capture the full paths.
For date-times to work correctly it is essential that the tables
defining time zones are installed: these are usually in an OS component
named something like tzdata. On most OSes they are required
but installations of Alpine Linux have been seen without them. There is
a configure check that recent date-times to work correctly in
different time zones which catches this when installing from source (but
not for binary distributions).
Those distributing binary versions of R may need to be aware of the
licences of the external libraries it is linked to (including ‘useful’
libraries from the next section). The liblzma library is in the
public domain and X11, libbzip2, libcurl and zlib
have MIT-style licences. PCRE and PCRE2 have a BSD-style licence which
requires distribution of the licence (included in R’s
COPYRIGHTS file) in binary distributions. GNU readline is
licensed under GPL (which version(s) of GPL depends on the
readline version).
The ability to use translated messages makes use of gettext and
most likely needs GNU gettext: you do need this to work
with new translations, but otherwise the version of the gettext
runtime contained in the R sources will be used if no suitable external
gettext is found.
The ‘modern’ version of the X11(), jpeg(), png()
and tiff() graphics devices uses the Cairo and Pango libraries.
Cairo version 1.2.0 or later and Pango version 1.10 or later are
required (but much later versions are current). R checks for
pkg-config, and uses that to check first that the
‘pangocairo’ package is installed (and if not, ‘cairo’) then
if suitable code can be compiled. These tests will fail if
pkg-config is not installed52, and might fail if cairo was built
statically unless configure option
--with-static-cairo is used. Most systems with Gtk+ 2.8
or later installed will have suitable libraries: for Fedora users the
pango-devel RPM and its dependencies suffice.
It is possible (but very unusual on a platform with X11) to build Cairo
without its cairo-xlib module in which case X11(type =
"cairo") will not be available. Pango is optional but highly desirable
as it is likely to give much better text rendering, including kerning.
For the best font experience with these devices you need suitable fonts
installed: Linux users will want the urw-fonts package. On
platforms which have it available, the msttcorefonts
package53 provides
TrueType versions of Monotype fonts such as Arial and Times New Roman.
Another useful set of fonts is the ‘liberation’ TrueType fonts available
at
https://pagure.io/liberation-fonts,54 which cover the Latin, Greek and Cyrillic alphabets
plus a fair range of signs. These share metrics with Arial, Times New
Roman and Courier New, and contain fonts rather similar to the first two
(https://en.wikipedia.org/wiki/Liberation_fonts). Then there
is the ‘Free UCS Outline Fonts’ project
(https://www.gnu.org/software/freefont/) which are
OpenType/TrueType fonts based on the URW fonts but with extended Unicode
coverage. See the R help on X11 on selecting such fonts.
The bitmapped graphics devices jpeg(), png() and
tiff() need the appropriate headers and libraries installed:
jpeg (version 6b or later, or libjpeg-turbo) or
libpng (version 1.2.7 or later) and zlib or libtiff
respectively.
pkg-config is used if available and so needs the appropriate
.pc file (which requires libtiff version 4.x and is not
available on all platforms for jpeg before version 9c). They
also need support for either X11 or cairo (see above).
Should support for these devices not be required or broken
system libraries need to be avoided there are configure
options --without-libpng, --without-jpeglib and
--without-libtiff. The TIFF library has many optional features
such as jpeg, libz, zstd, lzma, webp,
jbig and jpeg12, none of which is required for the
tiff() devices but may need to be present to link the library
(usually only an issue for static linking). pkg-config can
tell you what other libraries are required for linking, for example by
pkg-config libtiff-4 --static --libs.
Option --with-system-tre is also available: it needs a recent
version of TRE. (The latest sources are in the git repository
at https://github.com/laurikari/tre/, but at the time of writing
the resulting build did not complete its checks, nor did R built
against the version supplied by Fedora.)
An implementation of XDR is required, and the R sources
contain one which is likely to suffice (although a system version may
have higher performance). XDR is part of RPC and
historically has been part of libc on a Unix-alike. (In
principle man xdr_string should tell you which library is
needed, but it often does not: on some OSes it is provided by
libnsl.) However some builds55 of glibc omit or hide it with the intention
that the TI-RPC library be used, in which case libtirpc
(and its development version) should be installed, and its
headers56 need to be on
the C include path or under /usr/include/tirpc.
Library libdeflate (https://github.com/ebiggers/libdeflate)
is used by memCompress() and memDecompress() if available.
Use of the X11 clipboard selection requires the Xmu headers and
libraries. These are normally part of an X11 installation (e.g. the
Debian meta-package ‘xorg-dev’), but some distributions have split
this into smaller parts, so for example recent versions of Fedora
require the ‘libXmu’ and ‘libXmu-devel’ RPMs.
Some systems (notably macOS and at least some FreeBSD systems) have
inadequate support for collation in multibyte locales. It is possible
to replace the OS’s collation support by that from ICU (International
Components for Unicode, https://icu.unicode.org/), and this
provides much more precise control over collation on all systems. ICU
is available as sources and as binary distributions for (at least) most
Linux distributions, FreeBSD, macOS and AIX, usually as libicu or
icu4c. It will be used by default where available: should a very
old or broken version of ICU be found this can be suppressed by
--without-ICU.
The bitmap and dev2bitmap devices and function
embedFonts() use Ghostscript
(https://www.ghostscript.com/). This should either be in your
path when the command is run, or its full path specified by the
environment variable R_GSCMD at that time.
At the time of writing a full installation on Fedora Linux used the following packages and their development versions, and this may provide a useful checklist for other systems:
bzip2 cairo fontconfig freetype fribidi gcc gcc-gfortran gcc-c++ glib2 glibc harfbuzz lapack libX11 libXext libXt libcurl libdeflate libicu libjpeg libpng libtiff libtirpc libxcrypt ncurses pango pkgconf-pkg-config pcre2 readline tcl tk xz zlib
plus, preferably a TeX installation and Java.
The tcltk package needs Tcl/Tk ≥ 8.4 installed: the sources are available at https://www.tcl.tk/. To specify the locations of the Tcl/Tk files you may need the configuration options
use Tcl/Tk, or specify its library directory
specify location of tclConfig.sh
specify location of tkConfig.sh
or use the configure variables TCLTK_LIBS and
TCLTK_CPPFLAGS to specify the flags needed for linking against
the Tcl and Tk libraries and for finding the tcl.h and
tk.h headers, respectively. If you have both 32- and 64-bit
versions of Tcl/Tk installed, specifying the paths to the correct config
files may be necessary to avoid confusion between them.
Versions of Tcl/Tk up to 8.5.19 and 8.6.12 have been tested (including most versions of 8.4.x, but not recently).
Note that the tk.h header includes57 X11 headers, so you will need X11 and its development files installed.
The build process looks for Java support on the host system, and if it
finds it sets some settings which are useful for Java-using packages
(such as rJava and JavaGD: installing these from
source requires a full JDK). This check can be suppressed by
configure option --disable-java.
Configure variable JAVA_HOME can be set to point to a specific
JRE/JDK, on the configure command line or in the environment.
Principal amongst these settings are some paths to the Java
libraries and JVM, which are stored in environment variable
R_JAVA_LD_LIBRARY_PATH in file R_HOME/etc/ldpaths (or
a sub-architecture-specific version). A typical setting for
‘x86_64’ Linux is
JAVA_HOME=/usr/lib/jvm/java-1.8.0-openjdk-1.8.0.322.b06-6.fc34.x86_64/jre
R_JAVA_LD_LIBRARY_PATH=${JAVA_HOME}/lib/amd64/server
Unfortunately this depends on the exact version of the JRE/JDK
installed, and so may need updating if the Java installation is updated.
This can be done by running R CMD javareconf which updates
settings in both R_HOME/etc/Makeconf and
R_HOME/etc/ldpaths. See R CMD javareconf --help for
details: note that this needs to be done by the account owning the R
installation.
Another way of overriding those settings is to set the environment variable
R_JAVA_LD_LIBRARY_PATH (before R is started, hence not in
~/.Renviron), which suffices to run already-installed
Java-using packages. For example
R_JAVA_LD_LIBRARY_PATH=/usr/lib/jvm/java-1.8.0/jre/lib/amd64/server
It may be possible to avoid this by specifying an invariant link as the path when configuring. For example, on that system any of
JAVA_HOME=/usr/lib/jvm/java JAVA_HOME=/usr/lib/jvm/java-1.8.0 JAVA_HOME=/usr/lib/jvm/java-1.8.0/jre JAVA_HOME=/usr/lib/jvm/jre-1.8.0
worked (since the ‘auto’ setting of /etc/alternatives chose
Java 8 aka 1.8.0).
‘Non-server’ Oracle distributions of Java as from version 11 are of a full JDK. However, Linux distributions can be confusing: for example Fedora 38 had
java-1.8.0-openjdk java-1.8.0-openjdk-devel java-11-openjdk java-11-openjdk-devel java-17-openjdk java-17-openjdk-devel java-latest-openjdk java-latest-openjdk-devel
where the -devel RPMs are needed to complete the JDK. Debian/Ubuntu use
‘-jre’ and ‘-jdk’, e.g.
sudo apt install default-jdk
Some add-on packages need a C++ compiler. This is specified by the
configure variables CXX, CXXFLAGS and similar.
configure will normally find a suitable compiler. It is
possible to specify an alternative C++17 compiler by the configure
variables CXX17, CXX17STD, CXX17FLAGS and similar
(see C++ Support). Again, configure will normally find a
suitable value for CXX17STD if the compiler given by CXX
is capable of compiling C++17 code, but it is possible that a completely
different compiler will be needed. (Similar macros are provided for
C++20.)
For source files with extension .f90 or .f95 containing
free-form Fortran, the compiler defined by the macro FC is used
by R CMD INSTALL. Note that it is detected by the name of the
command without a test that it can actually compile Fortran 90 code.
Set the configure variable FC to override this if necessary:
variables FCFLAGS and FCLIBS_XTRA might also need to be
set.
See file config.site in the R source for more details about these variables.
The linear algebra routines in R make use of BLAS (Basic Linear Algebra Subprograms, https://netlib.org/blas/faq.html) routines, and most make use of routines from LAPACK (Linear Algebra PACKage, https://netlib.org/lapack/). The R sources contain reference (Fortran) implementations of these, but they can be replaced by external libraries, usually those tuned for speed on specific CPUs. These libraries normally contain all of the BLAS routines and some tuned LAPACK routines and perhaps the rest of LAPACK from the reference implementation. Because of the way linking works, using an external BLAS library may necessitate using the version of LAPACK it contains.
Note that the alternative implementations will not give identical
numeric results. Some differences may be benign (such the signs of SVDs
and eigenvectors), but the optimized routines can be less accurate and
(particularly for LAPACK) can be from older versions with fewer
corrections. However, R relies on
ISO/IEC 60559 compliance. This can be broken
if for example the code assumes that terms with a zero factor are always
zero and do not need to be computed—whereas x*0 can be
NaN. The internal BLAS has been extensively patched to avoid
this whereas MKL’s documentation has warned
LAPACK routines assume that input matrices do not contain IEEE 754 special values such as INF or NaN values. Using these special values may cause LAPACK to return unexpected results or become unstable.
Some of the external libraries are multi-threaded. One issue is
that R profiling (which uses the SIGPROF signal) may cause
problems, and you may want to disable profiling if you use a
multi-threaded BLAS. Note that using a multi-threaded
BLAS can result in taking more CPU time and even
more elapsed time (occasionally dramatically so) than using a similar
single-threaded BLAS. On a machine running other tasks, there
can be contention for CPU caches that reduces the effectiveness of the
optimization of cache use by a BLAS implementation: some
people warn that this is especially problematic for hyper-threaded CPUs.
BLAS and LAPACK routines may be used inside threaded code, for example in OpenMP sections in packages such as mgcv. The reference implementations are thread-safe but external ones may not be (even single-threaded ones): this can lead to hard-to-track-down incorrect results or segfaults.
There is a tendency for re-distributors of R to use ‘enhanced’ linear algebra libraries without explaining their downsides.
An external BLAS library has to be explicitly requested at configure time.
You can specify a particular BLAS library via a value
for the configuration option --with-blas. If this is given
with no =, its value is taken from the
environment variable BLAS_LIBS, set for example in
config.site. If neither the option nor the environment variable
supply a value, a search is made for a suitable58 BLAS. If the
value is not obviously a linker command (starting with a dash or giving
the path to a library), it is prefixed by ‘-l’, so
--with-blas="foo"
is an instruction to link against ‘-lfoo’ to find an external BLAS (which needs to be found both at link time and run time).
The configure code checks that the external BLAS is complete
(as of LAPACK 3.9.1: it must include all double precision and double
complex routines, as well as LSAME), and appears to be usable.
However, an external BLAS has to be usable from a shared
object (so must contain position-independent code), and that is not
checked. Also, the BLAS can be switched after configure is run, either
as a symbolic link or by the mechanisms mentioned below, and this can
defeat the completeness check.
Some enhanced BLASes are compiler-system-specific
(Accelerate on macOS, sunperf on Solaris59,
libessl on IBM). The correct incantation for these is often
found via --with-blas with no value on the appropriate
platforms.
Note that under Unix (but not under Windows) if R is compiled against a non-default BLAS and --enable-BLAS-shlib is not used (it is the default on all platforms except AIX), then all BLAS-using packages must also be. So if R is re-built to use an enhanced BLAS then packages such as quantreg will need to be re-installed.
Debian/Ubuntu systems provide a system-specific way to switch the BLAS
in use: Build R with --with-blas to select the OS version of
the reference BLAS, and then use update-alternatives to switch
between the available BLAS libraries. See
https://wiki.debian.org/DebianScience/LinearAlgebraLibraries.
Fedora 33 and later offer ‘FlexiBLAS’, a similar mechanism for switching
the BLAS in use
(https://www.mpi-magdeburg.mpg.de/projects/flexiblas). However,
rather than overriding libblas, this requires configuring R
with option --with-blas=flexiblas. ‘Backend’ wrappers are
available for the reference BLAS, ATLAS and serial, threaded and OpenMP
builds of OpenBLAS and BLIS, and perhaps others60. This can be controlled from a
running R session by package flexiblas.
BLAS implementations which use parallel computations can be non-deterministic: this is known for ATLAS.
ATLAS (https://math-atlas.sourceforge.net/) is a “tuned” BLAS that runs on a wide range of Unix-alike platforms. Unfortunately it is built by default as a static library that on some platforms may not be able to be used with shared objects such as are used in R packages. Be careful when using pre-built versions of ATLAS static libraries (they seem to work on ‘ix86’ platforms, but not always on ‘x86_64’ ones).
ATLAS contains replacements for a small number of LAPACK routines, but can be built to merge these with the reference LAPACK sources to include a full LAPACK library.
Recent versions of ATLAS can be built as a single shared library, either
libsatlas or libtatlas (serial or threaded respectively):
these may even contain a full LAPACK. Such builds can be used by one of
--with-blas=satlas --with-blas=tatlas
or, as on ‘x86_64’ Fedora where a path needs to be specified,
--with-blas="-L/usr/lib64/atlas -lsatlas" --with-blas="-L/usr/lib64/atlas -ltatlas"
Distributed ATLAS libraries cannot be tuned to your machine and so are a compromise: for example Fedora tunes61 ‘x86_64’ RPMs for CPUs with SSE3 extensions, and separate RPMs may be available for specific CPU families.
Note that building R on Linux against distributed shared libraries may need ‘-devel’ or ‘-dev’ packages installed.
Linking against multiple static libraries requires one of
--with-blas="-lf77blas -latlas" --with-blas="-lptf77blas -lpthread -latlas" --with-blas="-L/path/to/ATLAS/libs -lf77blas -latlas" --with-blas="-L/path/to/ATLAS/libs -lptf77blas -lpthread -latlas"
Consult its installation guide62 for how to build ATLAS as a shared library or as a static library with position-independent code (on platforms where that matters).
According to the ATLAS FAQ63 the maximum number of threads used by multi-threaded ATLAS is set at compile time. Also, the author advises against using multi-threaded ATLAS on hyper-threaded CPUs without restricting affinities at compile-time to one virtual core per physical CPU. (For the Fedora libraries the compile-time flag specifies 4 threads.)
Dr Kazushige Goto wrote a tuned BLAS for several processors and OSes, which was frozen in 2010. OpenBLAS (https://www.openblas.net/) is a descendant project with support for some later CPUs.
This can be used by configuring R with something like
--with-blas="openblas"
See see Shared BLAS for an alternative (and in many ways preferable) way to use them.
Some platforms provide multiple builds of OpenBLAS: for example Fedora has RPMs64
openblas openblas-threads openblas-openmp
providing shared libraries
libopenblas.so libopenblasp.so libopenblaso.so
respectively, each of which can be used as a shared BLAS. For the
second and third the number of threads is controlled by
OPENBLAS_NUM_THREADS and OMP_NUM_THREADS (as usual for
OpenMP) respectively.
These and their Debian equivalents contain a complete LAPACK implementation.
Note that building R on Linux against distributed libraries may need ‘-devel’ or ‘-dev’ packages installed.
For ‘ix86’ and ‘x86_64’ CPUs most distributed libraries contain several alternatives for different CPU microarchitectures with the choice being made at run time.
Another descendant project is BLIS (https://github.com/flame/blis). This has (in Fedora) shared libraries
libblis.so libblisp.so libbliso.so
(p for ‘threads’, o for OpenMP as for OpenBLAS) which can
also be used as a shared BLAS. The Fedora builds do not include LAPACK
in the BLIS libraries.
For Intel processors (and perhaps others) and some distributions of Linux, there is Intel’s Math Kernel Library65. You are encouraged to read the documentation which is installed with the library before attempting to link to MKL. This includes a ‘link line advisor’ which will suggest appropriate incantations: its use is recommended. Or see https://www.intel.com/content/www/us/en/developer/tools/oneapi/onemkl-link-line-advisor.html#gs.vpt6qp (which at the time of writing selected the Intel library for linking with GCC).
There are also versions of MKL for macOS66 and Windows, but when these have been tried they did not work with the default compilers used for R on those platforms.
The following examples have been used with MKL versions 10.3 to 2023.2.0, for GCC compilers on ‘x86_64’ CPUs. (See also Intel compilers.)
To use a sequential version of MKL we used
MKL_LIB_PATH=/path/to/intel_mkl/mkl/lib/intel64
export LD_LIBRARY_PATH=$MKL_LIB_PATH
MKL="-L${MKL_LIB_PATH} -lmkl_gf_lp64 -lmkl_core -lmkl_sequential"
./configure --with-blas="$MKL" --with-lapack
The option --with-lapack is used since MKL contains a tuned copy of LAPACK (often older than the current version) as well as the BLAS (see LAPACK), although this can be omitted.
Threaded MKL may be used by replacing the line defining the variable
MKL by
MKL="-L${MKL_LIB_PATH} -lmkl_gf_lp64 -lmkl_core \
-lmkl_gnu_thread -dl -fopenmp"
R can also be linked against a single shared library,
libmkl_rt.so, for both BLAS and LAPACK, but the correct OpenMP and
MKL interface layer then has to be selected via environment variables. With
64-bit builds and the GCC compilers, we used
export MKL_INTERFACE_LAYER=GNU,LP64 export MKL_THREADING_LAYER=GNU
On Debian/Ubuntu, MKL is provided by package intel-mkl-full and one
can set libmkl_rt.so as the system-wide implementation of both BLAS
and LAPACK during installation of the package, so that also R installed
from Debian/Ubuntu package r-base would use it. It is, however,
still essential to set MKL_INTERFACE_LAYER and
MKL_THREADING_LAYER before running R, otherwise MKL computations
will produce incorrect results. R does not have to be rebuilt to use MKL,
but configure includes tests which may discover some errors such as a
failure to set the correct OpenMP and MKL interface layer.
Note that the Debian/Ubuntu distribution can be quite old (for example
2020.4 in mid-2023 when 2023.1 was current): this can be
important for the LAPACK version included.
The default number of threads will be chosen by the OpenMP software, but
can be controlled by setting OMP_NUM_THREADS or
MKL_NUM_THREADS, and in recent versions seems to default to a
sensible value for sole use of the machine. (Parallel MKL has not
always passed make check-all, but did with MKL 2019.4 and later.)
MKL includes a partial implementation of FFTW3, which causes trouble for applications that require some of the FFTW3 functionality unsupported in MKL. Please see the MKL manuals for description of these limitations and for instructions on how to create a custom version of MKL which excludes the FFTW3 wrappers.
There is Intel documentation for building R with MKL at https://www.intel.com/content/www/us/en/developer/articles/technical/using-onemkl-with-r.html: that includes
-Wl,--no-as-needed
which we have not found necessary.
If when configuring R a system LAPACK library is found of version 3.9.0 or later (and does not contain BLAS routines) it will be used instead of compiling the LAPACK code in the package sources. This can be prevented by configuring R with --without-lapack. Using a static liblapack.a is not supported.
It is assumed that -llapack is the reference LAPACK library but
on Debian/Ubuntu it can be switched, including after R is installed.
On such a platform it is better to use --without-lapack or
--with-blas --with-lapack (see below) explicitly. The known
examples67 of a
non-reference LAPACK library found at installation all contain BLAS
routines so are not used by a default configure run.
Provision is made for specifying an external LAPACK library with option
--with-lapack, principally to cope with BLAS
libraries which contain a copy of LAPACK (such as Accelerate on
macOS and some builds of ATLAS, FlexiBLAS, MKL and OpenBLAS on
‘ix86’/‘x86_64’ Linux). At least LAPACK version 3.2 is
required. This can only be done if --with-blas has been used.
However, the likely performance gains are thought to be small (and may
be negative). The default is not to search for a suitable LAPACK
library, and this is definitely not recommended. You can
specify a specific LAPACK library or a search for a generic library by
the configuration option --with-lapack without a value. The
default for --with-lapack is to check the BLAS
library (for function DPSTRF) and then look for an external
library ‘-llapack’. Sites searching for the fastest possible
linear algebra may want to build a LAPACK library using the
ATLAS-optimized subset of LAPACK. Similarly, OpenBLAS can be built to
contain an optimized subset of LAPACK or a full LAPACK (the latter
seeming to be the default).
A value for --with-lapack can be set via the environment
variable
LAPACK_LIBS, but this will only be used if --with-lapack
is specified and the BLAS library does not contain LAPACK.
Please bear in mind that using --with-lapack is provided only because it is necessary on some platforms and because some users want to experiment with claimed performance improvements. In practice its main uses are without a value,
liblapack which can
be switched by the ‘alternatives’ mechanism.
If building LAPACK from its Netlib sources, be aware that make
with its supplied Makefile will make a static library and
R requires a shared/dynamic one. To get one, use cmake as
documented briefly in README.md. Something like (to build only
the double and double complex subroutines with 32-bit array indices),
mkdir build cd build cmake \ -DCMAKE_INSTALL_PREFIX=/where/you/want/to/install \ -DCMAKE_BUILD_TYPE:STRING=Release \ -DBUILD_DEPRECATED=ON -DBUILD_SHARED_LIBS=ON \ -DBUILD_INDEX64_EXT_API:BOOL=OFF \ -DBUILD_SINGLE:BOOL=OFF -DBUILD_COMPLEX:BOOL=OFF \ -DLAPACKE=OFF -DCBLAS=OFF \ -S .. make -j10
This builds the reference BLAS and the reference LAPACK linked to it.
Note that cmake files do not provide an uninstall
target, but build/install_manifest.txt is a list of the files
installed, so you can remove them via shell commands or from
R.
If using --with-lapack to get a generic LAPACK (or allowing the default to select one), consider also using --with-blas (with a path if an enhanced BLAS is installed).
As with all libraries, you need to ensure that they and R were
compiled with compatible compilers and flags. For example, this has
meant that on Sun Sparc using the Oracle compilers the flag
-dalign is needed if sunperf is to be used.
On some systems it has been necessary that an external BLAS/LAPACK was built with the same Fortran compiler used to build R.
BLAS and LAPACK libraries built with recent versions of
gfortran require calls from C/C++ to handle ‘hidden’ character
lengths — R itself does so but many packages used not to and some
have segfaulted. This was largely circumvented by using the Fortran
flag -fno-optimize-sibling-calls (formerly set by
configure if it detected gfortran 7 or later):
however use of the R headers which include those character-length
arguments is no longer optional in packages.
LAPACK 3.9.0 (and probably earlier) had a bug in which the DCOMBSSQ
subroutine may cause NA to be interpreted as zero. This is fixed in the
R 3.6.3 and later sources, but if you use an external LAPACK, you may
need to fix it there. (The bug was corrected in 3.9.1 and the routine
removed in 3.10.1.)
The code (in dlapack.f) should read
* ..
* .. Executable Statements ..
*
IF( V1( 1 ).GE.V2( 1 ) ) THEN
IF( V1( 1 ).NE.ZERO ) THEN
V1( 2 ) = V1( 2 ) + ( V2( 1 ) / V1( 1 ) )**2 * V2( 2 )
ELSE
V1( 2 ) = V1( 2 ) + V2( 2 )
END IF
ELSE
V1( 2 ) = V2( 2 ) + ( V1( 1 ) / V2( 1 ) )**2 * V1( 2 )
V1( 1 ) = V2( 1 )
END IF
RETURN
(The inner ELSE clause was missing in LAPACK 3.9.0.)
If you do use an external LAPACK, be aware of potential problems with
other bugs in the LAPACK sources (or in the posted corrections to those
sources), seen several times in Linux distributions over the years. We
have even seen distributions with missing LAPACK routines from their
liblapack.
We rely on limited support in LAPACK for matrices with 2^{31} or more elements: it is possible that an external LAPACK will not have that support.
configure has many options: running
./configure --help
will give a list. Probably the most important ones not covered elsewhere are (defaults in brackets)
use the X Window System [yes]
X include files are in DIR
X library files are in DIR
use readline library (if available) [yes]
attempt to compile support for Rprof() [yes]
attempt to compile support for Rprofmem() and tracemem() [no]
build R as a shared/dynamic library [no]
build the BLAS as a shared/dynamic library [yes, except on AIX]
You can use --without-foo or --disable-foo for the negatives.
You will want to use --disable-R-profiling if you are building a profiled executable of R (e.g. with ‘-pg)’. Support for R profiling requires OS support for POSIX threads (aka ‘pthreads’), which are available on all mainstream Unix-alike platforms.
Flag --enable-R-shlib causes the make process to build R as a dynamic (shared) library, typically called libR.so, and link the main R executable R.bin against that library. This can only be done if all the code (including system libraries) can be compiled into a dynamic library, and there may be a performance68 penalty. So you probably only want this if you will be using an application which embeds R. Note that C code in packages installed on an R system linked with --enable-R-shlib is linked against the dynamic library and so such packages cannot be used from an R system built in the default way. Also, because packages are linked against R they are on some OSes also linked against the dynamic libraries R itself is linked against, and this can lead to symbol conflicts.
For maximally effective use of valgrind, R should be
compiled with Valgrind instrumentation. The configure option
is --with-valgrind-instrumentation=level, where
level is 0, 1 or 2. (Level 0 is the default and does not add
anything.) The system headers for valgrind are required: on
Linux they may be in a separate package such as valgrind-devel.
If you need to re-configure R with different options you may need to run
make clean or even make distclean before doing so.
The configure script has other generic options added by
autoconf and which are not supported for R: in particular
building for one architecture on a different host is not possible.
Translation of messages is supported via GNU gettext
unless disabled by the configure option --disable-nls.
The configure report will show NLS as one of the
‘Additional capabilities’ if support has been compiled in, and running
in an English locale (but not the C locale) will include
Natural language support but running in an English locale
in the greeting on starting R.
If you need or want to set certain configure variables to something other than their default, you can do that by either editing the file config.site (which documents many of the variables you might want to set: others can be seen in file etc/Renviron.in) or on the command line as
./configure VAR=value
If you are building in a directory different from the sources, there can be copies of config.site in the source and the build directories, and both will be read (in that order). In addition, if there is a file ~/.R/config, it is read between the config.site files in the source and the build directories.
There is also a general autoconf mechanism for
config.site files, which are read before any of those mentioned
in the previous paragraph. This looks first at a file specified by the
environment variable CONFIG_SITE, and if not is set at files such
as /usr/local/share/config.site and
/usr/local/etc/config.site in the area (exemplified by
/usr/local) where R would be installed.
These variables are precious, implying that they do not have to be exported to the environment, are kept in the cache even if not specified on the command line, checked for consistency between two configure runs (provided that caching is used), and are kept during automatic reconfiguration as if having been passed as command line arguments, even if no cache is used.
See the variable output section of configure --help for a list of
all these variables.
If you find you need to alter configure variables, it is worth noting that some settings may be cached in the file config.cache, and it is a good idea to remove that file (if it exists) before re-configuring. Note that caching is turned off by default: use the command line option --config-cache (or -C) to enable caching.
One common variable to change is R_PAPERSIZE, which defaults to
‘a4’, not ‘letter’. (Valid values are ‘a4’,
‘letter’, ‘legal’ and ‘executive’.)
This is used both when configuring R to set the default, and when running R to override the default. It is also used to set the paper size when making PDF manuals.
The configure default will most often be ‘a4’ if R_PAPERSIZE
is unset. (If the program paperconf is found, present in many
Linux distributions,
or the environment variable PAPERSIZE is set, these are used to
produce the default.)
Another precious variable is R_BROWSER, the default HTML
browser, which should take a value of an executable in the user’s path
or specify a full path.
Its counterpart for PDF files is R_PDFVIEWER.
If you have libraries and header files, e.g., for GNU
readline, in non-system directories, use the variables LDFLAGS
(for libraries, using ‘-L’ flags to be passed to the linker) and
CPPFLAGS (for header files, using ‘-I’ flags to be passed to
the C/C++ preprocessors), respectively, to specify these locations.
These default to ‘-L/usr/local/lib’ (LDFLAGS,
‘-L/usr/local/lib64’ on most 64-bit Linux OSes) and
‘-I/usr/local/include’ (CPPFLAGS, but note that on most
systems /usr/local/include is regarded as a system include
directory and so instances in that macro will be skipped) to catch the
most common cases. If libraries are still not found, then maybe your
compiler/linker does not support re-ordering of -L and
-l flags.
In this case, use a different compiler (or a front-end shell script
which does the re-ordering).
These flags can also be used to build a faster-running version of R.
On most platforms using gcc, having ‘-O3’ in
CFLAGS and FFLAGS produces worthwhile
performance gains with gcc and gfortran, but may
result in a less reliable build (both segfaults and incorrect numeric
computations have been seen). On systems using the GNU linker
(especially those using R as a shared library), it is likely that
including ‘-Wl,-O1’ in LDFLAGS is worthwhile, and
‘'-Bdirect,--hash-style=both,-Wl,-O1'’ is recommended at
https://lwn.net/Articles/192624/. Tuning compilation to a
specific CPU family (e.g. ‘-mtune=native’ for
gcc) can give worthwhile performance gains, especially on
older architectures such as ‘ix86’.
The default settings for making the manuals are controlled by
R_RD4PDF and R_PAPERSIZE.
By default the shell scripts such as R will be ‘#!/bin/sh’
scripts (or using the SHELL chosen by configure). This is
almost always satisfactory, but on a few systems /bin/sh is not a
Bourne shell or clone, and the shell to be used can be changed by
setting the configure variable R_SHELL to a suitable value (a full
path to a shell, e.g. /usr/local/bin/bash).
To build in a separate directory you need a make that supports
the VPATH variable, for example GNU make and
dmake.
If you want to use a make by another name, for example if your
GNU make is called ‘gmake’, you need to set the
variable MAKE at configure time, for example
./configure MAKE=gmake
To compile R, you need a Fortran 90 compiler. The current default
is to search for
gfortran, g95, xlf95 f95,
fort, ifort, ifc, efc,
pgfortran, pgf95 lf95, ftn,
nagfor, xlf90, f90, pgf90,
pghpf, epcf90. (Note that these are searched for by
name, without checking the standard of Fortran they support.) The
command and flags used should support fixed-form Fortran with extension
.f: in the unusual case that a specific flag is needed for
free-form Fortran with extension .f90 or .f95, this can be
specified as part of FCFLAGS.
The search mechanism can be changed using the configure variable
FC which specifies the command that runs the Fortran compiler.
If your Fortran compiler is in a non-standard location, you
should set the environment variable PATH accordingly before
running configure, or use the configure variable FC to
specify its full path.
If your Fortran libraries are in slightly peculiar places, you should
also look at LD_LIBRARY_PATH (or your system’s equivalent) to make
sure that all libraries are on this path.
Note that only Fortran compilers which convert identifiers to lower case are supported.
You must set whatever compilation flags (if any) are needed to ensure
that Fortran integer is equivalent to a C int pointer and
Fortran double precision is equivalent to a C double
pointer. This is checked during the configuration process.
Some of the Fortran code makes use of DOUBLE COMPLEX and
COMPLEX*16 variables. This is checked for at configure time, as
well as its equivalence to the Rcomplex C structure defined in
R_ext/Complex.h.
gfortran 10 by default gives a compilation error for the
previously widespread practice of passing a Fortran array element where
an array is expected, or a scalar instead of a length-one array. See
https://gcc.gnu.org/gcc-10/porting_to.html. gfortran 12
errors in more cases of this.
A wide range of flags can be set in the file config.site or as configure variables on the command line. We have already mentioned
CPPFLAGSheader file search directory (-I) and any other miscellaneous options for the C and C++ preprocessors and compilers
LDFLAGSpath (-L), stripping (-s) and any other miscellaneous options for the linker
and others include
CFLAGSdebugging and optimization flags, C
MAIN_CFLAGSditto, for compiling the main program (e.g. when profiling)
SHLIB_CFLAGSfor shared objects (no known examples)
FFLAGSdebugging and optimization flags, fixed-form Fortran
FCFLAGSdebugging and optimization flags, free-form Fortran
SAFE_FFLAGSditto for source files which need exact floating point behaviour
MAIN_FFLAGSditto, for compiling the main program (e.g. when profiling)
SHLIB_FFLAGSfor shared objects (no known examples)
MAIN_LDFLAGSadditional flags for the main link
SHLIB_LDFLAGSadditional flags for linking the shared objects
LIBnnthe primary library directory, lib or lib64
CPICFLAGSspecial flags for compiling C code to be turned into a shared object
FPICFLAGSspecial flags for compiling Fortran code to be turned into a shared object
CXXPICFLAGSspecial flags for compiling C++ code to be turned into a shared object
DEFSdefines to be used when compiling C code in R itself
Library paths specified as -L/lib/path in LDFLAGS are
collected together and prepended to LD_LIBRARY_PATH (or your
system’s equivalent), so there should be no need for -R or
-rpath flags.
Variables such as CPICFLAGS are determined where possible by
configure. Some systems allows two types of PIC flags, for
example ‘-fpic’ and ‘-fPIC’, and if they differ the first
allows only a limited number of symbols in a shared object. Since R
as a shared library has about 6200 symbols, if in doubt use the larger
version.
Other variables often set by configure include
‘MAIN_LDFLAGS’, ‘SAFE_FFLAGS’, ‘SHLIB_LDFLAGS’ and
‘SHLIB_CXXLDFLAGS’: see file config.site in the sources for
more documentation on these and others.
To compile a profiling version of R, one might for example want to use ‘MAIN_CFLAGS=-pg’, ‘MAIN_FFLAGS=-pg’, ‘MAIN_LDFLAGS=-pg’ on platforms where ‘-pg’ cannot be used with position-independent code.
Beware: it may be necessary to set CFLAGS and
FFLAGS in ways compatible with the libraries to be used: one
possible issue is the alignment of doubles, another is the way
structures are passed.
On some platforms configure will select additional flags for
CFLAGS, CPPFLAGS and LIBS in R_XTRA_CFLAGS
(and so on). These are for options which are always required, for
example to force IEC 60559 compliance.
There are several files that are part of the R sources but can be
re-generated from their own sources by configuring with option
--enable-maintainer-mode and then running make in the
build directory. This requires other tools to be installed, discussed
in the rest of this section.
File configure is created from configure.ac and the files
under m4 by autoconf and aclocal (part of the
automake package). There is a formal version requirement on
autoconf of 2.71 or later, but it is unlikely that anything
other than the most recent versions69
have been thoroughly tested.
File src/include/config.h is created by autoheader
(part of autoconf).
Grammar files *.y are converted to C sources by an implementation
of yacc, usually bison -y: these are found in
src/main and src/library/tools/src. It is known that
earlier versions of bison generate code which reads (and in
some cases writes) outside array bounds: bison 3.8.2 is
currently used.
The ultimate sources for package compiler are in its noweb
directory. To re-create the sources from
src/library/compiler/noweb/compiler.nw, the command
notangle is required. Some Linux distributions include this
command in package noweb. It can also be installed from the
sources at https://www.cs.tufts.edu/~nr/noweb/70. The package
sources are only re-created even in maintainer mode if
src/library/compiler/noweb/compiler.nw has been updated.
This section provides some notes on building R on different Unix-alike platforms. These notes are based on tests run on one or two systems in each case with particular sets of compilers and support libraries. Success in building R depends on the proper installation and functioning of support software; your results may differ if you have other versions of compilers and support libraries.
Older versions of this manual contain notes on platforms such as HP-UX, IRIX, Alpha/OSF1 (for R < 2.10.0, and support has since been removed for all of these) and AIX (for R < = 3.5.x) for which we have had no recent reports.
C macros to select particular platforms can be tricky to track down (there is a fair amount of misinformation on the Web). The Wiki (currently) at https://sourceforge.net/p/predef/wiki/Home/ can be helpful. The R sources have used (often in included software under src/extra)
AIX: _AIX Cygwin: __CYGWIN__ FreeBSD: __FreeBSD__ HP-UX: __hpux__, __hpux IRIX: sgi, __sgi Linux: __linux__ macOS: __APPLE__ NetBSD: __NetBSD__ OpenBSD: __OpenBSD__ Windows: _WIN32, _WIN64 Windows on 64-but ARM: _M_ARM64 or _WIN32 plus __aarch64__
Identifying compilers can be very tricky. GCC defines __GNUC__,
but so do other compilers claiming conformance with it, notably (LLVM
and Apple) clang and Intel compilers. Further, some use the
value of __GNUC__ for their version, not the version of GCC they
claim to be compatible with.71 clang-based
compilers define __clang__. Both LLVM and Apple clang
define __clang_major__ as a string giving their major version,
but for example Apple’s 13.x.y is very different from LLVM’s 13.x.y.
And compilers based on LLVM clang, for example from Intel and
IBM, will define these. Some of the included software uses
__APPLE_CC__ to identify an Apple compiler (which used to include
Apple builds of GCC), but Apple clang is better identified by
the __apple_build_version__ macro.
The ‘X11()’ graphics device is the one started automatically on Unix-alikes (except most macOS builds) when plotting. As its name implies, it displays on a (local or remote) X server, and relies on the services provided by the X server.
The ‘modern’ version of the ‘X11()’ device is based on ‘cairo’ graphics and (in most implementations) uses ‘fontconfig’ to pick and render fonts. This is done on the server, and although there can be selection issues, they are more amenable than the issues with ‘X11()’ discussed in the rest of this section.
When X11 was designed, most displays were around 75dpi, whereas today they are of the order of 100dpi or more. If you find that X11() is reporting72 missing font sizes, especially larger ones, it is likely that you are not using scalable fonts and have not installed the 100dpi versions of the X11 fonts. The names and details differ by system, but will likely have something like Fedora’s
xorg-x11-fonts-75dpi xorg-x11-fonts-100dpi xorg-x11-fonts-ISO8859-2-75dpi xorg-x11-fonts-Type1 xorg-x11-fonts-cyrillic
and you need to ensure that the ‘-100dpi’ versions are installed
and on the X11 font path (check via xset -q). The
‘X11()’ device does try to set a pointsize and not a pixel size:
laptop users may find the default setting of 12 too large (although very
frequently laptop screens are set to a fictitious dpi to appear like a
scaled-down desktop screen).
More complicated problems can occur in non-Western-European locales, so
if you are using one, the first thing to check is that things work in
the C locale. The likely issues are a failure to find any fonts
or glyphs being rendered incorrectly (often as a pair of ASCII
characters). X11 works by being asked for a font specification and
coming up with its idea of a close match. For text (as distinct from
the symbols used by plotmath), the specification is the first element of
the option "X11fonts" which defaults to
"-adobe-helvetica-%s-%s-*-*-%d-*-*-*-*-*-*-*"
If you are using a single-byte encoding, for example ISO 8859-2 in
Eastern Europe or KOI8-R in Russian, use xlsfonts to find an
appropriate family of fonts in your encoding (the last field in the
listing). If you find none, it is likely that you need to install
further font packages, such as ‘xorg-x11-fonts-ISO8859-2-75dpi’ and
‘xorg-x11-fonts-cyrillic’ shown in the listing above.
Multi-byte encodings (most commonly UTF-8) are even more complicated. There are few fonts in ‘iso10646-1’, the Unicode encoding, and they only contain a subset of the available glyphs (and are often fixed-width designed for use in terminals). In such locales fontsets are used, made up of fonts encoded in other encodings. If the locale you are using has an entry in the ‘XLC_LOCALE’ directory (typically /usr/share/X11/locale), it is likely that all you need to do is to pick a suitable font specification that has fonts in the encodings specified there. If not, you may have to get hold of a suitable locale entry for X11. This may mean that, for example, Japanese text can be displayed when running in ‘ja_JP.UTF-8’ but not when running in ‘en_GB.UTF-8’ on the same machine (although on some systems many UTF-8 X11 locales are aliased to ‘en_US.UTF-8’ which covers several character sets, e.g. ISO 8859-1 (Western European), JISX0208 (Kanji), KSC5601 (Korean), GB2312 (Chinese Han) and JISX0201 (Kana)).
On some systems scalable fonts are available covering a wide range of glyphs. One source is TrueType/OpenType fonts, and these can provide high coverage. Another is Type 1 fonts: the URW set of Type 1 fonts provides standard typefaces such as Helvetica with a larger coverage of Unicode glyphs than the standard X11 bitmaps, including Cyrillic. These are generally not part of the default install, and the X server may need to be configured to use them. They might be under the X11 fonts directory or elsewhere, for example,
/usr/share/fonts/default/Type1 /usr/share/fonts/ja/TrueType
Linux is the main development platform for R, so compilation from the sources is normally straightforward with the most common compilers and libraries.73
This section is about the GCC compilers:
gcc/gfortran/g++.
Recall that some package management systems (such as RPM and
deb) make a distinction between the user version of a package and the
developer version. The latter usually has the same name but with the
extension ‘-devel’ or ‘-dev’: you need both versions
installed. So please check the configure output to see if the
expected features are detected: if for example ‘readline’ is
missing add the developer package. (On most systems you will also need
‘ncurses’ and its developer package, although these should be
dependencies of the ‘readline’ package(s).) You should expect to
see in the configure summary
Interfaces supported: X11, tcltk External libraries: pcre2, readline, curl Additional capabilities: PNG, JPEG, TIFF, NLS, cairo, ICU
When R has been installed from a binary distribution there are sometimes problems with missing components such as the Fortran compiler. Searching the ‘R-help’ archives will normally reveal what is needed.
It seems that ‘ix86’ Linux accepts non-PIC code in shared
libraries, but this is not necessarily so on other platforms, in
particular on 64-bit CPUs such as ‘x86_64’. So care
can be needed with BLAS libraries and when building R as a
shared library to ensure that position-independent code is used in any
static libraries (such as the Tcl/Tk libraries, libpng,
libjpeg and zlib) which might be linked against.
Fortunately these are normally built as shared libraries with the
exception of the ATLAS BLAS libraries.
The default optimization settings chosen for CFLAGS etc are
conservative. It is likely that using -mtune will result in
significant performance improvements on recent CPUs: one possibility is
to add -mtune=native for the best possible performance on the
machine on which R is being installed. It is also possible to
increase the optimization levels to -O3: however for many
versions of the compilers this has caused problems in at least one
CRAN package.
Do not use -O3 with gcc 11.0 or 11.1: it mis-compiles
code resulting in plausible but
incorrect results. (This was seen in package MASS but has
been worked around there as from version 3.1-57.)
For comments on ‘ix86’ builds (including 32-bit builds on ‘x86_64’) see the version of this manual for R 4.3.x.
To build a 64-bit version of R on ‘ppc64’ (also known as
‘powerpc64’) with gcc 4.1.1, Ei-ji Nakama used
CC="gcc -m64" CXX="gxx -m64" FC="gfortran -m64" CFLAGS="-mminimal-toc -fno-optimize-sibling-calls -g -O2" FFLAGS="-mminimal-toc -fno-optimize-sibling-calls -g -O2"
the additional flags being needed to resolve problems linking against libnmath.a and when linking R as a shared library.
The setting of the macro ‘SAFE_FFLAGS’ may need some help. It
should not need additional flags on platforms other than ‘68000’
(not likely to be encountered) and ‘ix86’. For the latter, if
the Fortran compiler is GNU (gfortran or possibly
g77) the flags
-msse2 -mfpmath=sse
are added: earlier versions of R added -ffloat-store and this might still be needed if a ‘ix86’ CPU is encountered without SSE2 support. Note that it is a replacement for ‘FFLAGS’, so should include all the flags in that macro (except perhaps the optimization level).
Additional compilation flags can be specified for added safety/security checks. For example Fedora adds
-Werror=format-security -Wp,-D_FORTIFY_SOURCE=3 -Wp,-D_GLIBCXX_ASSERTIONS -Fexceptions -fstack-protector-strong -fasynchronous-unwind-tables -fstack-clash-protection -fcf-protection
to all the C, C++ and Fortran compiler flags (even though
_GLIBCXX_ASSERTIONS is only for C++ in current GCC and
glibc and none of these are documented for gfortran).
Use of _GLIBCXX_ASSERTIONS will link abort and
printf into almost all C++ code, and R CMD check
--as-cran will warn.
R has been built with Linux ‘ix86’ and ‘x86_64’ C and
C++ compilers (https://clang.llvm.org) based on the Clang
front-ends, invoked by CC=clang CXX=clang++, together with
gfortran. These take very similar options to the
corresponding GCC compilers.
This has to be used in conjunction with a Fortran compiler: the
configure code will remove -lgcc from FLIBS,
which is needed for some versions of gfortran.
The current out-of-the-box default for clang++ is to use the
C++ runtime from the installed g++. Using the runtime from
the libc++ project (Fedora RPM
libcxx-devel) via -stdlib=libc++ has also been
tested.
Recent versions have (optional when built) OpenMP support.74
There are problems mixing clang 15.0.0 and later built as
default on Linux to produce PIE code and gfortran 11 or later,
which does not. One symptom is that configure does not detect
FC_LEN_T, which can be overcome by setting
FPIEFLAGS=-fPIE
in config.site. (configure tries that value if it is
unset.)
The name flang has been used for two projects: this is about
the sub-project of LLVM which builds a Fortran compiler and runtime
libraries. The compiler is currently named flang-new but has been
announced to be renamed to flang when more nearly complete (and
at some earlier point in its development was known as f18).
The version in LLVM 16 and later was able to build R on ‘x86_64’ Linux with
FC=/path/to/flang-new
with the matching clang used as the C compiler, and the build
passed make check-all. There is also support for
‘aarch64’ and ‘ppc64le’ Linux, but these have not been
tested with R.
In late 2020 Intel revamped their C/C++ compilers (and later their
Fortran compiler) to use an LLVM back-end (and for the C/C++ compilers,
a modified version of clang as the front-end). Those
compilers are only for ‘x86_64’: the earlier (now called
‘Classic’) C/C++ compilers were discontinued in late 2023 (and are covered in
the version of this manual for R 4.3.x: the Fortran compiler
ifort remains part of the Fortran distribution)..
The compilers are now all under Intel’s ‘oneAPI’ brand. The revamped
ones are icx, icpx and ifx; they are
identified by the C/C++ macro __INTEL_LLVM_COMPILER (and do not
define __INTEL_COMPILER: they also define __clang__ and
__clang_major__).
The C++ compiler uses the system’s libstdc++ as its runtime
library rather than LLVM’s libc++.
Standalone installers (which are free-of-charge) are available from https://www.intel.com/content/www/us/en/developer/articles/tool/oneapi-standalone-components.html: they are also part of the oneAPI Base and HPC (for Fortran) toolkits.
We tried the compilers in oneAPI 2024.2.0 and 2023.x.y using (the paths do differ by compiler version)
IP=/path/to/compilers/bin/ CC=$IP/icx CXX=$IP/icpx FC=$IP/ifx CFLAGS="-O3 -fp-model precise -Wall -Wstrict-prototypes" C17FLAGS="-O3 -fp-model precise -Wall -Wno-strict-prototypes" FFLAGS="-O3 -fp-model precise -warn all,noexternals" FCFLAGS="-free -O3 -fp-model precise -warn all,noexternals" CXXFLAGS="-O3 -fp-model precise -Wall" LDFLAGS="-L/path/to/compilers/compiler/lib -L/usr/local/lib64"
but the build segfaulted in the checks (in complex arithmetic in tests/lapack.R).
Intel document building R with MKL: for the Intel compilers this needed something like
MKL_LIB_PATH=/path/to/intel_mkl/mkl/lib/intel64
export LD_LIBRARY_PATH="$MKL_LIB_PATH"
MKL="-L${MKL_LIB_PATH} -lmkl_intel_lp64 -lmkl_core -lmkl_sequential"
./configure --with-blas="$MKL" --with-lapack
and the build passed its checks with MKL 2023.2.0 (but not 2024.x on the hardware tested). It may also be possible to use a compiler option like -qmkl=sequential.
One quirk is that the Intel Fortran compilers do not accept .f95
files, only .f90, for free-format Fortran. configure
adds -Tf which tells the compiler this is indeed a Fortran file
(and needs to immediately precede the file name), but -free is
needed to say it is free-format. Hence setting the FCFLAGS
macro.
The compilers have many options: as the C/C++ and Fortran compilers have
different origins for their front-ends, there is little consistency in
their options.
(The C/C++ compilers support ‘all’ clang options even if
undocumented for icx/icpc, such as
-Wno-strict-prototypes above, However it is unclear for which
version of clang: the Intel manual suggests checking
icx -help.) The C/C++ compilers support clang-style
LTO: it is not clear if the Fortran one does.
For some earlier versions (including 2023.2.0) all CPU times in e.g.
proc.time() are reported as zero. If you see this, uncomment the
‘INTEL_ICX_FIX’ setting in config.site and re-build.
The preferred Fortran standard for ifx can be set by one of
-std90, -std95, -std03, -std08 or
-std18 (and variants). However, this is documented to only
affect warnings on non-standard features: the default is no such
warnings.
Warning to package maintainers: the Intel Fortran compiler interprets comments intended for Visual Fortran75 like
!DEC$ ATTRIBUTES DLLEXPORT,C,REFERENCE,ALIAS:'kdenestmlcvb' :: kdenestmlcvb
The DLLEXPORT gives a warning but the remainder silently
generates incorrectly named entry points. Such comment lines need to be
removed from code for use with R (even if using Intel Fortran on
Windows).
The instructions here are for ‘Apple Silicon’ (‘arm64’) or Intel 64-bit (‘x86_64’) builds on macOS 11 (Big Sur), 12 (Monterey), 13 (Ventura), 14 (Sonoma) and likely later. (They may well work on Intel macOS 10.14 or 10.15, but are untested there.)
The Apple silicon conmponents install into /opt/R/arm64, the Intel ones into /opt/R/x86_64. That may not exist76 so it is simplest to first create the directory and adjust its ownership if desired: for example by
sudo mkdir -p /opt/R/arm64 sudo chown -R $USER /opt/R
Also, add /opt/R/arm64/bin or /opt/R/x86_64/bin to your path.
Define an appropriate variable in your Terminal:
set LOCAL=/opt/R/arm64 # Apple Silicon set LOCAL=/opt/R/x86_64 # Intel
to use the code snippets here.
The following are essential to build R:
xcode-select --install in a terminal.
If you have a fresh OS installation, running e.g. make in a
terminal will offer the installation of the command-line tools. If you
have installed Xcode, this provides the command-line tools. The tools
may need to be reinstalled when macOS is upgraded, as upgrading may
partially or completely remove them.
The Command Line Tools provide C and C++ compilers derived from LLVM’s
clang but nowadays known as ‘Apple clang’ with different
versioning (so Apple clang 15 is unrelated to LLVM clang 15).
pcre277 and
xz (for liblzma) from
https://mac.r-project.org/bin/. There is an R script there to
help with installing all the needed components. (At the time of writing
install.libs("r-base-dev") installed neither readline5 nor
those needed to support Pango.)
Intel users want the darwin20 components: the darwin17
ones are for macOS 10.13–10.15.
Or this can be done manually, by for example
curl -OL https://mac.r-project.org/bin/darwin20/arm64/pcre2-10.42-darwin.20-arm64.tar.xz sudo tar -xvzf pcre2-10.42-darwin.20-arm64.tar.gz -C / curl -OL https://mac.r-project.org/bin/darwin20/arm64/xz-5.4.2-darwin.20-arm64.tar.xz sudo tar -xvzf xz-5.4.2-darwin.20-arm64.tar.xz -C /
(sudo is not needed if your account owns /opt/R/arm64 or
/opt/R/x86_64 as appropriate.)
Messages like ‘opt/R/: Can't restore time’ should be ignored.
and desirable
readline5.78 If readline is
not present, the emulation in Apple’s version of libedit (aka
editline) will be used: if you wish to avoid that, configure with
--without-readline.
jpeg, libpng, pkgconfig, tiff and
zlib-system-stub from https://mac.r-project.org/bin// for
the full range of bitmapped graphics devices. (Some builds of
tiff may require libwebp and/or openjpeg.)
configure
script can be told to look for X11 in XQuartz’s main
location of /opt/X11, e.g. by
--x-includes=/opt/X11/include --x-libraries=/opt/X11/lib
Be wary of pre-release versions of XQuartz, which may be offered as an update.
clang in the Command
Line Tools: this is needed for the quartz() graphics device.
Use --without-aqua if you want a standard Unix-alike build:
apart from disabling quartz() and the ability to use the build
with R.APP, it also changes the default location of the personal
library (see ?.libPaths).
texi2any from a ‘texinfo’ distribution, which requires
perl (currently a default part of macOS but it has been
announced that it may not be in future).
A version of texi2any has been included in the binary
distribution of R and there is a texinfo component at
https://mac.r-project.org/bin/.
To build R itself from the sources with the C/C++ compilers in the
Command Line Tools (or Xcode) and gfortran from the installer
mentioned below, use a file config.site containing
CC=clang OBJC=$CC FC="/opt/gfortran/bin/gfortran -mtune=native" CPPFLAGS='-isystem $LOCAL/include' CXX=clang++
and configure by something like
./configure -C \ --enable-R-shlib --enable-memory-profiling \ --x-includes=/opt/X11/include --x-libraries=/opt/X11/lib \ --with-tcl-config=$LOCAL/lib/tclConfig.sh \ --with-tk-config=$LOCAL/lib/tkConfig.sh \ PKG_CONFIG_PATH=$LOCAL/lib/pkgconfig:/usr/lib/pkgconfig
(See below for other options for Tcl/Tk.) For an ‘arm64’ build further flags are desirable in config.site:
CFLAGS="-falign-functions=8 -g -O2"
is needed to inter-work with gfortran without segfaulting in
some packages. Some builds of gfortran have targetted the
current version of macOS (unlike clang), causing linker
warnings: to avoid these use
FFLAGS="-g -O2 -mmacosx-version-min=11.0" FCFLAGS="-g -O2 -mmacosx-version-min=11.0"
or perhaps
FFLAGS="-g -O2 -mmacos-version-min=11.0" FCFLAGS="-g -O2 -mmacos-version-min=11.0"
where 11.0 can be replaced by 12.0, 13.0 or
14.0 for macOS 12.x, 13.x and 14.x.
To install packages using compiled code one needs the Command Line Tools
(or Xcode) and appropriate compilers, e.g. the C/C++ compilers from
those tools and/or gfortran. Some packages have further
requirements such as component pkgconfig (and to set
PKG_CONFIG_PATH= as above).
A subversion client can be obtained from https://mac.r-project.org/tools/, for example by (Apple Silicon)
curl -OL https://mac.r-project.org/tools/subversion-1.14.1-darwin.20-arm64.tar.gz tar xf subversion-1.14.1-darwin.20-arm64.tar.gz sudo cp subversion-1.14.1-darwin-20-arm64/svn $LOCAL/bin
or (Intel)
curl -OL https://mac.r-project.org/tools/subversion-1.14.0-darwin15.6.tar.gz tar xf subversion-1.14.0-darwin15.6.tar.gz sudo cp subversion-1.14.0-darwin15.6/svn $LOCAL/bin
If building software or installing source packages with cmake
(or a non-Apple make) for ‘Apple Silicon’ ensure it contains
the ‘arm64’ architecture (use file to be sure).
Running Apple compilers from an ‘x86_64’ executable will
generate ‘x86_64’ code ….
Updating an ‘arm64’ build may fail because of the bug described at https://openradar.appspot.com/FB8914243 but ab initio builds work. This has been far rarer since macOS 13.
If you are using the macOS 13 SDK79,
you may need to add something like -mmacos-version-min=12.0 to
‘CFLAGS’.
Linker warnings like
ld: warning: could not create compact unwind for _sort_: register 26 saved somewhere other than in frame ld: warning: ld: warning: could not create compact unwind for _arcoef_: registers 23 and 24 not saved contiguously in frame ld: warning: could not create compact unwind for ___emutls_get_address: registers 23 and 24 not saved contiguously in frame
can be ignored. These stem from compiled Fortran code, including its run-time libraries.
The default security settings can make it difficult to install Apple
packages which have not been ‘notarized’80
by Apple. And not just packages, as this has been seen for executables
contained in tarballs/zipfiles (for example, for pandoc).
Usually one can use ‘Open With’ (Control/right/two-finger-click in
Finder), then select ‘Installer’ and ‘Open’ if you get a
further warning message.
If you run into problems with ‘quarantine’ for tarballs downloaded in a
browser, consider using curl -OL to download (as illustrated
above) or xattr -c to remove extended attributes.
configure defaults to --with-internal-tzcode on
macOS. The native implementation used to be unusable on earlier
versions (with a 32-bit time_t and/or timezone tables missing
information beyond the 32-bit range). As from macOS 12.6, option
--without-internal-tzcode can be used to override this and R
contains sufficient workarounds (for example, the native code fails to
recognize dates with a negative tm_year, that is dates before
1900) for R to pass its checks. However, there are discrepancies,
notably in Europe in the 1900s and 1940s, even though the Olson database
contains the correct information.
There is a ‘universal’ (arm64 and Intel) build of
gfortran 12.2 at
https://mac.r-project.org/tools/gfortran-12.2-universal.pkg.
This installs into /opt/gfortran.
The /opt/gfortran/SDK symlink should point to the desired path to
the SDK (defaults to the command line tools SDK). This can be updated by
running /opt/gfortran/bin/gfortran-update-sdk or manually. If the
symlink is broken, the driver will issue a warning and use xcrun
-show-sdk-path to try to find an SDK and use its path. (The SDK path
is used when using gfortran to link, so not when building R
but when installing a few packages.)
Builds of gfortran 13.2 and 14.1 for arm64 macOS 14 are
available at
https://github.com/fxcoudert/gfortran-for-macOS/releases. These
can be built for Intel and older OSes from the sources at
https://github.com/iains/gcc-13-branch/ and
gcc-14-branch/. To use one of the pre-built compilers with Apple
clang needs something like
LDFLAGS="-L/opt/R/arm64/lib -rpath /usr/local/gfortran/lib"
in config.site to ensure the Fortran run-time libraries are found.
Cairo-based graphics devices such as cairo_ps, cairo_pdf,
X11(type = "cairo") and the Cairo-based types of devices
bmp jpeg, png and tiff are not the default
on macOS, and much less used than the Quartz-based devices. However,
the only SVG device in the R distribution, svg, is based on
Cairo.
Support for Cairo is optional and can be added in several ways, all of
which need pkg-config. configure will add Cairo support
if pkg-config finds package cairo unless
--without-cairo is used.
A way to statically link Cairo is by downloading and unpacking
components cairo, fontconfig, freetype,
pixman and zlib-system-stub (and do not have
/opt/X11/lib/pkgconfig in PKG_CONFIG_PATH). Some static
builds of fontconfig need libxml2 (from component
xml2) and others expat, supplied by macOS but needing a
file $LOCAL/lib/pkgconfig/expat.pc along the lines of
Name: expat Version: 2.2.8 Description: expat XML parser URL: http://www.libexpat.org Libs: -lexpat Cflags:
Note that the list of components is liable to change: running
pkg-config cairo --exists --print-errors should tell you if
any others are required.
The best font experience of Cairo graphics will be to use it in
combination with Pango which will match that supported on most other
Unix-alikes. configure uses pkg-config to determine
if all the external software required by both Pango and Cairo is
available: running pkg-config pangocairo --exists
--print-errors should show if the installation suffices and if not,
what is missing. At the time of writing using pre-built components
cairo, fontconfig, freetype, ffi,
fribidi, gettext, icu, glib,
harfbuzz, pango, pcre, pixman and
xml2 sufficed.
Other pre-compiled distributions of clang may be available
from https://github.com/llvm/llvm-project/releases/ (recently
only for arm64 and usually unsigned/not notarized which makes
them hard to use). In particular, these include support for
OpenMP which Apple clang does not. Some of these have
included support for the ASan and UBSan sanitizers.
Suppose one of these distributions is installed under $LOCAL/llvm. Use a file config.site containing
SDK=`xcrun -show-sdk-path` CC="$LOCAL/llvm/bin/clang -isysroot $SDK" CXX="$LOCAL/llvm/bin/clang++ -isysroot $SDK" OBJC=$CC FC=/opt/gfortran/bin/gfortran LDFLAGS="-L$LOCAL/llvm/lib -L$LOCAL/lib" R_LD_LIBRARY_PATH=$LOCAL/llvm/lib:$LOCAL/lib
The care to specify library paths is to ensure that the OpenMP runtime library, here $LOCAL/llvm/lib/libomp.dylib, is found when needed. If this works, you should see the line
checking whether OpenMP SIMD reduction is supported... yes
in the configure output. Also, ‘R_LD_LIBRARY_PATH’ needs
to be set to find the latest version of the C++ run-time libraries
rather than the system ones.
It is normally possible to build R with GCC (built from the sources,
from a gfortran distribution, from Homebrew, …).
When last tested it was not possible to use gcc to build the
quartz() device, so configure --without-aqua may be
required. R was built and tested with the GCC 14.1 compilers in the
arm64 gfortran distribution mentioned above using a
config.site containing
CC=/usr/local/gfortran/bin/gcc CXX=/usr/local/gfortran/bin/g++ FC=/usr/local/gfortran/bin/gfortran CFLAGS="-g -O2 -Wall -pedantic -Wstrict-prototypes" C17FLAGS="-g -O2 -Wall -pedantic -Wno-strict-prototypes" C90FLAGS=$C17FLAGS C99FLAGS=$C17FLAGS CXXFLAGS="-g -O2 -Wall -pedantic" CPPFLAGS='-isystem /opt/R/arm64/include' LDFLAGS=-L/opt/R/arm64/lib
It is usually possible to add some OpenMP support to the Apple
clang compilers: see https://mac.r-project.org/openmp/.
Note that that approach is somewhat fragile as it needs a
libomp.dylib library matching the version of the compiler
used—and for example at the time of writing none was offered for the
current compilers in Xcode CLT 14.3 nor 15.
Pre-compiled versions of many of the Useful libraries and programs are available from https://mac.r-project.org/bin//.
Looking at the top of /Library/Frameworks/R.framework/Resources/etc/Makeconf will show the compilers and configuration options used for the CRAN binary package for R: at the time of writing the non-default options
--enable-memory-profiling --enable-R-framework --x-libraries=/opt/X11/lib --x-includes=/opt/X11/include
were used. (--enable-R-framework implies --enable-R-shlib.)
The main TeX implementation used by the developers is MacTeX81 (https://www.tug.org/mactex/): the full installation is about 8.5GB, but a much smaller version (‘Basic TeX’) is available at https://www.tug.org/mactex/morepackages.html to which you will need to add some packages to build R, e.g. for the 2022 version we needed to add82 helvetic, inconsolata and texinfo which brought this to about 310MB.83 ‘TeX Live Utility’ (available via the MacTeX front page) provides a graphical means to manage TeX packages. These contain executables which run natively on both ‘arm64’ and ‘x86_64’.
Checking packages thoroughly requires Ghostscript (part of the full
MacTeX distribution or separately from
https://www.tug.org/mactex/morepackages.html) and qpdf
(from https://mac.r-project.org/bin//, a version of which is in
the bin directory of a binary installation of R, usually
/Library/Frameworks/R.framework/Resources/bin/qpdf).
R CMD check --as-cran makes use of ‘HTML Tidy’. macOS at the
time of writing has a version in /usr/bin/tidy dating from 2006
which is far too old. Up-to-date versions can be
installed from http://binaries.html-tidy.org/.
One macOS quirk is that the default path has /usr/local/bin after /usr/bin, contrary to common practice on Unix-alikes. This means that if you install tools from the sources they will by default be installed under /usr/local and not supersede the system versions.
Parallel installation of packages will make use of the utility
timeout if available. A ‘universal’ build can be
downloaded from https://www.stats.ox.ac.uk/pub/bdr/timeout: make
it executable (chmod 755 timeout) and put it somewhere on your
path.
The Accelerate
library84
can be used via the configuration option
--with-blas="-framework Accelerate"
to provide potentially higher-performance versions of the BLAS and LAPACK routines.85 This includes a full LAPACK which can be used via --with-lapack: however, the version of LAPACK it contains has often been seriously old (and is not used unless --with-lapack is specified). Some CRAN builds of R can be switched86 to use Accelerate’s BLAS.
As from macOS 13.3, the BLAS and LAPACK libraries under the Accelerate
framework are ‘now inline with reference version
3.9.1’.87 However, this has been done by
naming new entry points and so only accessible via their C
headers. That version can be used for BLAS calls via
configure option --with-newAccelerate: it requires at
least macOS 13.3 and SDK 13.3 (from Xcode CLT 14.3). To use it for both
BLAS and LAPACK calls, configure with
--with-newAccelerate=lapack. These options cannot be used with
others such as --with-blas and --with-lapack.
Threading in Accelerate is controlled by ‘Grand Central
Dispatch’88 and is
said not to need user control. Test nls.R in package stats
has often failed with the Accelerate BLAS on Intel macOS. All versions
of Accelerate show differences from the reference BLAS (and most others)
in the use of NA vs NaN and a substantial number of
R packages fail their checks.
R has been built on ‘arm64’ using OpenBLAS 0.3.24 (sources from https://github.com/OpenMathLib/OpenBLAS/releases) by symlinking /opt/OpenBLAS/lib/libopenblas.dylib to lib/libRblas.dylib (see Shared BLAS).
On macOS, a default build of OpenBLAS uses pthreads (as macOS
does not have OpenMP) with the number of threads controlled by
environment variable OPENBLAS_NUM_THREADS. On an M1 Pro this
defaulted to 10 threads (there are 8 ‘performance’ cores and 2
‘efficiency cores‘) and we saw a 9x speedup over the reference BLAS on a
large SVD (which was slightly faster than Accelerate).
If you plan to use the tcltk package for R, you will need to
install a distribution of Tcl/Tk. There are two alternatives. If you
use R.APP you will want to use X11-based Tcl/Tk (as used on other
Unix-alikes), which is installed under $LOCAL/lib as part of the CRAN binary for
R.89 This may need configure
options
--with-tcltk=$LOCAL/lib
or
--with-tcl-config=$LOCAL/lib/tclConfig.sh --with-tk-config=$LOCAL/lib/tkConfig.sh
Note that this requires a matching XQuartz installation.
There is also a native (‘Aqua’) version of Tcl/Tk which produces widgets in the native macOS style: this will not work with R.APP because of conflicts over the macOS menu, but for those only using command-line R this provides a much more intuitive interface to Tk for experienced Mac users. Earlier versions of macOS came with an Aqua Tcl/Tk distribution but these were often not at all recent versions of Tcl/Tk. It is better to install Tcl/Tk 8.6.x from the sources90 or a binary distribution from https://www.activestate.com/products/tcl/. For the latter, configure R with
--with-tcl-config=/Library/Frameworks/Tcl.framework/tclConfig.sh --with-tk-config=/Library/Frameworks/Tk.framework/tkConfig.sh
If you need to find out which distribution of Tk is in use at run time, use
library(tcltk)
tclvalue(.Tcl("tk windowingsystem")) # "x11" or "aqua"
Note that some Tcl/Tk extensions only support the X11 interface: this
includes Tktable and the CRAN package
tkrplot.
macOS does not comes with an installed Java runtime (JRE) and a
macOS upgrade may remove one if already installed: it is intended to be
installed at first use. Check if a JRE is installed by running
java -version in a Terminal window: if Java is not
installed on an Intel Mac this may prompt you to install it. We
recommend you install a version with long-term support, e.g. 17 or
2191
but not 18–20, 22–24 with a 6-month lifetime.
The currently simplest way to install Java is from
Adoptium92: this installs into an
Apple-standard location and so works with /usr/bin/java. Other
builds are available from
https://www.azul.com/downloads/zulu-community/?os=macos&architecture=arm-64-bit&package=jdk
and from OpenJDK at https://jdk.java.net/, for which
JAVA_HOME may need to be set both when configuring R and at
runtime. Note that Java distribution sites may use unusual designations
for macOS CPUs such as AArch64, x64 or x86
64-bit.
Binary distributions of R are built against a specific version (e.g. 11.0.18 or 17.0.1) of Java so
sudo R CMD javareconf
will likely be needed to be run before using Java-using packages.
To see what compatible versions of Java are currently installed, run the appropriate one of
/usr/libexec/java_home -V -a arm64 /usr/libexec/java_home -V -a x86_64
If needed, set the environment variable JAVA_HOME to choose
between these, both when R is built from the sources and when
R CMD javareconf is run.
Configuring and building R both looks for a JRE and for support for compiling JNI programs (used to install packages rJava and JavaGD); the latter requires a JDK (Java SDK). Most distributions of Java 11 or later are of a full JDK.
The build process tries to fathom out what JRE/JDK to use,
but it may need some help, e.g. by setting environment variable
JAVA_HOME. To select a build from Adoptium set e.g.
JAVA_HOME=/Library/Java/JavaVirtualMachines/temurin-21.jdk/Contents/Home
in config.site. For Java 21 from https://jdk.java.net/ (which might no longer be available), use
JAVA_HOME=/path/to/jdk-21.jdk/Contents/Home
Note that it is necessary to set the environment variable NOAWT to
1 to install many of the Java-using packages.
The CRAN build of R is installed as a framework, which is selected by the option
./configure --enable-R-framework
(This is intended to be used with an Apple toolchain: others may not support frameworks correctly but those from LLVM have done so.)
It is only needed if you want to build R for use with the R.APP console, and implies --enable-R-shlib to build R as a dynamic library. This option configures R to be built and installed as a framework called R.framework. The default installation path for R.framework is /Library/Frameworks but this can be changed at configure time by specifying the flag --enable-R-framework[=DIR] (or --prefix) or at install time via
make prefix=/where/you/want/R.framework/to/go install
Note that installation as a framework is non-standard (especially to a
non-standard location) and Unix utilities may not support it (e.g. the
pkg-config file libR.pc will be put somewhere unknown
to pkg-config).
Building the R.APP GUI console is a separate project, using Xcode. Before compiling R.APP make sure the current version of R is installed in /Library/Frameworks/R.framework and is working at the command-line (this can be a binary install).
The current sources can be checked out by
svn co https://svn.r-project.org/R-packages/trunk/Mac-GUI
and built by loading the R.xcodeproj project (select the
R target and a suitable configuration), or from the command-line
by e.g.
xcodebuild -target R -configuration Release
See also the INSTALL file in the checkout or directly at https://svn.r-project.org/R-packages/trunk/Mac-GUI/INSTALL.
R.APP does not need to be installed in any specific way. Building R.APP results in the R.APP bundle which appears as one R icon. This application bundle can be run from anywhere and it is customary to place it in the /Applications folder.
CRAN macOS binary packages are distributed as tarballs with
suffix .tgz to distinguish them from source tarballs. One can
tar an existing installed package, or use R CMD
INSTALL --build.
However, there are some important details.
CC="clang -mmacos-version-min=11.0" CXX="clang++ -mmacos-version-min=11.0" FC="/opt//gfortran/bin/gfortran -mmacosx-version-min=11.0"
or set the environment variable
export MACOSX_DEPLOYMENT_TARGET=11.0
otool -L or
objdump -macho -dylibs-used. This can include C++ and Fortran
run-time libraries under /opt/R/x86_64/lib or
/opt/R/arm64/lib: one can use install_name_tool to
point these at system versions or those shipped with R, for example
install_name_tool -change /usr/local/llvm/lib/libc++.1.dylib \ /usr/lib/libc++.1.dylib \ pkg.so install_name_tool -change /opt/gfortran/lib/gcc/aarch64-apple-darwin20.0/12.2.0/libgfortran.5.dylib \ /Library/Frameworks/R.framework/Resources/lib/libgfortran.5.dylib \ pkg.so install_name_tool -change /opt/gfortran/lib/gcc/aarch64-apple-darwin20.0/12.2.0/libquadmath.0.dylib \ /Library/Frameworks/R.framework/Resources/lib/libquadmath.0.dylib \ pkg.so
(where the details depend on the compilers and CRAN macOS R release).
SHLIB_CXXLD = /usr/local/llvm/bin/clang PKG_LIBS = /usr/local/llvm/lib/libc++.a /usr/local/llvm/lib/libc++abi.a
in src/Makevars. It would also be possible to static link the
Fortran runtime libraries libgfortran.a and libquadmath.a
should the Fortran compiler have later versions (but gfortran
8–14 all have version 5).
The CRAN binary packages are built with the Apple compiler on the oldest supported version of macOS, which avoids the first two and any issues with C++ libraries.
Should one want to build R for Intel on an ‘arm64’ Big Sur Mac, add the target for the compilers:
CC="clang -arch x86_64 OBJC=$CC CXX="clang++ -arch x86_64" FC="/opt//gfortran/bin/gfortran -arch x86_64 -mtune=native -mmacosx-version-min=11"
and install the Fortran compiler and external software described above for Intel builds (and have /opt/R/x86_64/bin before /opt/R/arm64/bin in your path).
To set the correct architecture (which will be auto-detected as
aarch64), use something like
/path/to/configure --build=x86_64-apple-darwin20
The scripts for the CRAN packaging of R can be found under https://svn.r-project.org/R-dev-web/trunk/QA/Simon/R4/: start with the README file in that directory.
There have been few recent reports on FreeBSD: there is a ‘port’ at
https://svnweb.freebsd.org/ports/head/math/R, currently last
updated for R 4.0.4. Recent versions of FreeBSD use Clang and the
libc++ C++ headers and runtime, but the ‘port’ has been
configured to use GCC.
Use of ICU for collation and the configure option
--with-internal-tzcode are desirable workarounds.
Ingo Feinerer installed R version 3.2.2 on OpenBSD 5.8 arch ‘amd64’ (their name for ‘x86_64’). Details of the build (and patches applied) are at https://cvsweb.openbsd.org/cgi-bin/cvsweb/ports/math/R/, currently updated for R 4.2.3.
The 32-bit version never worked well enough to pass R’s make
check, and residual support from earlier experiments was removed in
R 3.3.0.
The 64-bit version was never supported.
There are a number of sources of problems when installing R on a new hardware/OS platform. These include
Floating Point Arithmetic: R requires arithmetic compliant
with IEC 60559, also known as IEEE 754.
This mandates the use of plus and minus infinity and NaN (not a
number) as well as specific details of rounding. Although almost all
current FPUs can support this, selecting such support can be a pain.
The problem is that there is no agreement on how to set the signalling
behaviour; Sun/Sparc, SGI/IRIX and ‘ix86’ Linux require no
special action, FreeBSD requires a call to (the macro)
fpsetmask(0) and OSF1 required that computation be done with a
-ieee_with_inexact flag etc. With Intel compilers on 32-bit and
64-bit Intel machines, one has to explicitly disable flush-to-zero and
denormals-are-zero modes. Some ARM processors including A12Z and M1
(Apple Silicon) by default use runfast mode, which includes
flush-to-zero and default-nan and hence has to be disabled. With
default-nan mode, the NaN payload used for representation of numeric NA
values is lost even on simple operations with finite values. On a new
platform you must find out the magic recipe and add some code to make it
work. This can often be done via the file config.site which
resides in the top level directory.
Beware of using high levels of optimization, at least initially. On
many compilers these reduce the degree of compliance to the
IEEE model. For example, using -fast on the Oracle
compilers has caused R’s NaN to be set incorrectly, and
gcc’s -ffast-math and clang’s
-Ofast have given incorrect results.
Shared Objects: There seems to be very little agreement
across platforms on what needs to be done to build shared objects.
there are many different combinations of flags for the compilers and
loaders. GNU libtool cannot be used (yet), as it currently
does not fully support Fortran: one would need a shell wrapper for
this). The technique we use is to first interrogate the X window system
about what it does (using xmkmf), and then override this in
situations where we know better (for tools from the GNU
Compiler Collection and/or platforms we know about). This typically
works, but you may have to manually override the results. Scanning the
manual entries for cc and ld usually reveals the
correct incantation. Once you know the recipe you can modify the file
config.site (following the instructions therein) so that the
build will use these options.
It seems that gcc 3.4.x and later on ‘ix86’ Linux
defeat attempts by the LAPACK code to avoid computations entirely in
extended-precision registers, so file src/modules/lapack/dlamc.f
may need to be compiled without optimization or with additional flags.
Set the configure variable SAFE_FFLAGS to the flags to be used for
this file.
If you do manage to get R running on a new platform please let us know about it so we can modify the configuration procedures to include that platform.
If you are having trouble getting R to work on your platform please feel free to use the ‘R-devel’ mailing list to ask questions. We have had a fair amount of practice at porting R to new platforms ...
One thing you might want to add for a new platform is the mapping of
C/C++/Fortran calls to entry point names used for R CMD check.
See https://svn.r-project.org/R-dev-web/trunk/sotools.txt for how
to do so.
| Jump to: | C I M R U |
|---|
| Jump to: | C I M R U |
|---|
| Jump to: | B C F I L M O P R S U V |
|---|
| Jump to: | B C F I L M O P R S U V |
|---|
| Jump to: | B C D J L P R T |
|---|
| Jump to: | B C D J L P R T |
|---|
e.g. GNU
tar version 1.15 or later, or that from the ‘libarchive’
(as used on macOS) or ‘Heirloom Toolchest’ distributions.
for some Subversion clients ‘http:’ may appear to work, but requires continual redirection.
aka ‘Apple Silicon’, known to some as ‘arm64-apple-darwin’.
which use lib rather than lib64 for their primary 64-bit library directories: attempts are made to detect such systems.
not by the version supplied by macOS.
Instructions on how to install the latest version are at https://www.ctan.org/tex-archive/fonts/inconsolata/.
on a
Unix-alike, ‘inconsolata’ is omitted if not found by
configure.
This will be needed if more than one sub-architecture is to be installed.
How to prepare such a directory is described in file src/extra/tzone/Notes in the R sources.
But on Windows problems have been seen with case-changing functions on accented Latin-1 characters.
for
example, -fopenmp, -fiopenmp, -xopenmp or
-qopenmp. This includes for clang and the Intel and
Oracle compilers.
This does not necessarily disable use of
OpenMP – the configure code allows for platforms where OpenMP
is used without a flag. For the flang compiler in late 2017,
the Fortran runtime always used OpenMP.
Then recommended packages installed as part of the R installation do use LTO, but not packages installed later.
A complete CRAN installation reduced from 50 to 35GB.
although there is the possibility to exclude Fortran but that misses some of the benefits.
not NM as we found make
overriding that.
probably also 8.4 and later.
There are reports of segfaults when ‘MiKTeX’
installs additional packages when making NEWS.pdf: re-running
make seems to solve this.
At the time of writing, version 2.8.5 or later.
The installer puts links to R and
Rscript in /usr/local/bin. If these are missing or
that is not on your path, you can run directly the copies in
/Library/Frameworks/R.framework/Resources/bin or link those
yourself to somewhere on your path.
Formerly known as the Trash.
At the
time of writing: use pkgutil --pkgs | grep -i org.r-project to
check.
More precisely, of the Apple package of the same name: this means that ARM and Intel versions can be installed together.
Including GCC 9 on Linux.
On Windows a path containing spaces will be replaced by the ‘short path’ version if that does not contain spaces.
unless they were excluded in the build.
its binding is locked once the startup files have been
read, so users cannot easily change it. See ?.libPaths for how
to make use of the new value.
If a proxy needs to be set, see
?download.file.
for a small number of CRAN packages where this is known to be safe and is needed by the autobuilder this is the default. Look at the source of tools:::.install_packages for the list. It can also be specified in the package’s DESCRIPTION file.
Note that capitalization and versioning may differ from the Open Source project.
using a path containing spaces is likely to cause problems
They need to have been created using -headerpad_max_install_names, which is the default for an R package.
‘X/Open Portability Guide’, which has had several versions.
On some systems setting
LC_ALL or LC_MESSAGES to ‘C’ disables LANGUAGE.
If you try changing from French to Russian except in a UTF-8 locale, you may find messages change to English.
the language written in England: some people living in the USA appropriate this name for their language.
with Americanisms.
e.g. Bessel, beta and gamma functions
including copying MkRules.dist to MkRule.local and selecting the architecture.
also known as IEEE 754
Note
that C11 compilers need not be C99-compliant: R requires support for
double complex and variable-length arrays which are optional in
C11 but are mandatory in C99. C17 (also known as C18 as it was
published in 2018) is a ‘bugfix release’ of C11, clarifying the
standard. However, all known recent compilers in C11 or C17 mode are
C99-compliant, and most default to C17.
Examples are -std=gnu99, -std=c99 and -c99.
However, it is possible to break
the default behaviour of glibc by re-specifying the gconv
modules to be loaded.
specifically, the C99
functionality of headers wchar.h and wctype.h, types
wctans_t and mbstate_t and functions mbrtowc,
mbstowcs, wcrtomb, wcscoll, wcstombs,
wctrans, wctype, and iswctype.
including expm1, hypot, log1p,
nearbyint and va_copy.
including
opendir, readdir, closedir, popen,
stat, glob, access, getcwd and chdir
system calls, select on a Unix-alike, and either putenv or
setenv.
such as
realpath, symlink.
most often distributed as part of xz:
possible names in Linux distributions include
xz-devel/xz-libs and liblzma-dev.
for example to specify static linking with a build which has both shared and static libraries.
Such as
GNU tar 1.15 or later, bsdtar (from
https://github.com/libarchive/libarchive/, used as
tar by FreeBSD and macOS 10.6 and later) or tar from
the Heirloom Toolchest
(https://heirloom.sourceforge.net/tools.html), although the
latter does not support xz compression.
texi2dvi is normally a shell
script. Some of the issues which have been observed with broken
versions of texi2dvi can be circumvented by setting the
environment variable R_TEXI2DVICMD to the value emulation.
If necessary the path to
pkg-config can be specified by setting PKG_CONFIG in
config.site, on the configure command line or in the
environment. There is a compatible re-implementation of
pkg-config called pkgconf which can be used in the
unlikely event that is installed but not linked to
pkg-config.
also known as ttf-mscorefonts-installer in the
Debian/Ubuntu world: see also
https://en.wikipedia.org/wiki/Core_fonts_for_the_Web.
ttf-liberation
in Debian/Ubuntu.
Including that used by Fedora 28 and later
R uses rpc/xdr.h but that includes netconfig.h from the top tirpc directory.
This is true even for the ‘Aqua’ version of Tk on macOS, but distributions of that include a copy of the X11 files needed.
The search order is currently OpenBLAS, BLIS, ATLAS, platform-specific choices (see below) and finally a generic libblas.
Using
the Oracle Developer Studio cc and f95 compilers
for example, Intel MKL not packaged by Fedora.
The only way to see exactly which CPUs the distributed libraries have been tuned for is to read the atlas.spec file.
https://math-atlas.sourceforge.net/atlas_install/
https://math-atlas.sourceforge.net/faq.html#tnum
(and more, e.g. for 64-bit ints and static versions).
Nowadays known as ‘Intel oneAPI Math Kernel Library’ or even ‘oneMKL’.
The issue for macOS has been the use of double-complex routines.
ATLAS, OpenBLAS and Accelerate.
We have measured 15–20% on ‘i686’ Linux and around 10% on ‘x86_64’ Linux.
at the time of revision of this para in early 2024, autoconf-2.72 and automake-1.16.5. Previously autoconf-2.71 was used.
The links there have proved difficult to access, in which case grab the copy made available at https://developer.r-project.org/noweb-2.11b.tgz.
Most clang-based
compilers give 4, but not those distributed by FreeBSD. Intel’s
icx reported 12 in 2023.
for example, X11 font at size 14 could not
be loaded.
For example, glibc: other C libraries such as
musl (as used by Alpine Linux) have been used but are not
routinely tested.
This also needs the OpenMP runtime which has sometimes been distributed separately.
as the ‘Classic’ compiler has been known on Windows.
it will if R has been installed from CRAN since R 4.3.0.
If compiling it from source on
‘arm64’, pcre2 (at least up to version 10.39) needs to
be built without JIT support (the default) as the R build segfaults
if that is enabled, so do run make check on your build.
For licence reasons this is version
5.2 of readline: for those who want a more recent version it is
straightforward to compile it from its sources.
ls -l `xcrun
-show-sdk-path` in a terminal will show you which SDK is selected.
See https://developer.apple.com/documentation/xcode/notarizing_macos_software_before_distribution.
An essentially equivalent TeX installation can be obtained by the Unix TeX Live installation scripts.
E.g. via tlmgr install helvetic
inconsolata texinfo .
Adding all the packages needed to check CRAN increased this to about 600MB.
https://developer.apple.com/documentation/accelerate.
It has been reported that for some
non-Apple toolchains CPPFLAGS needed to contain
-D__ACCELERATE__: not needed for clang from LLVM.
Released 2021-04-01.
E.g., https://en.wikipedia.org/wiki/Grand_Central_Dispatch .
Just that component can be selected from the installer for R: at the ‘Installation Type’ screen select ‘Customise’ and then just the ‘Tcl/Tk 8.6.11’ component.
Configure Tk with --enable-aqua.
The planned nextLTS release is 25 in September 2025. Java 8 aka 1.8.0 is still LTS but some packages require 11 or later.
which website works with Safari but not some other browsers.